solve the pdf concate error

* Added some modules for supporting openrouter model

Added some modules for supporting openrouter model

* Update config.py

* Update .gitignore

* Update bridge_openrouter.py

* Not changed actually

* Refactor logging in bridge_openrouter.py

---------

Co-authored-by: binary-husky <qingxu.fu@outlook.com>

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -69,9 +69,3 @@ body:
     attributes:
       label: Terminal Traceback & Material to Help Reproduce Bugs | 终端traceback（如有） + 帮助我们复现的测试材料样本（如有）
       description: Terminal Traceback & Material to Help Reproduce Bugs | 终端traceback（如有） + 帮助我们复现的测试材料样本（如有）
-
-
-
-
-
-

.github/ISSUE_TEMPLATE/feature_request.yml

@@ -21,8 +21,3 @@ body:
     attributes:
       label: Feature Request | 功能请求
       description: Feature Request | 功能请求
-
-
-
-
-

.github/workflows/build-with-all-capacity-beta.yml

@@ -0,0 +1,44 @@
+# https://docs.github.com/en/actions/publishing-packages/publishing-docker-images#publishing-images-to-github-packages
+name: build-with-all-capacity-beta
+
+on:
+  push:
+    branches:
+      - 'master'
+
+env:
+  REGISTRY: ghcr.io
+  IMAGE_NAME: ${{ github.repository }}_with_all_capacity_beta
+
+jobs:
+  build-and-push-image:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      packages: write
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v3
+
+      - name: Log in to the Container registry
+        uses: docker/login-action@v2
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Extract metadata (tags, labels) for Docker
+        id: meta
+        uses: docker/metadata-action@v4
+        with:
+          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+
+      - name: Build and push Docker image
+        uses: docker/build-push-action@v4
+        with:
+          context: .
+          push: true
+          file: docs/GithubAction+AllCapacityBeta
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}

.github/workflows/build-with-all-capacity.yml

@@ -0,0 +1,44 @@
+# https://docs.github.com/en/actions/publishing-packages/publishing-docker-images#publishing-images-to-github-packages
+name: build-with-all-capacity
+
+on:
+  push:
+    branches:
+      - 'master'
+
+env:
+  REGISTRY: ghcr.io
+  IMAGE_NAME: ${{ github.repository }}_with_all_capacity
+
+jobs:
+  build-and-push-image:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      packages: write
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v3
+
+      - name: Log in to the Container registry
+        uses: docker/login-action@v2
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Extract metadata (tags, labels) for Docker
+        id: meta
+        uses: docker/metadata-action@v4
+        with:
+          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+
+      - name: Build and push Docker image
+        uses: docker/build-push-action@v4
+        with:
+          context: .
+          push: true
+          file: docs/GithubAction+AllCapacity
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}

.github/workflows/stale.yml

@@ -0,0 +1,25 @@
+# This workflow warns and then closes issues and PRs that have had no activity for a specified amount of time.
+#
+# You can adjust the behavior by modifying this file.
+# For more information, see:
+# https://github.com/actions/stale
+
+name: 'Close stale issues and PRs'
+on:
+  schedule:
+    - cron: '*/5 * * * *'
+
+jobs:
+  stale:
+    runs-on: ubuntu-latest
+    permissions:
+      issues: write
+      pull-requests: read
+
+    steps:
+      - uses: actions/stale@v8
+        with:
+          stale-issue-message: 'This issue is stale because it has been open 100 days with no activity. Remove stale label or comment or this will be closed in 1 days.'
+          days-before-stale: 100
+          days-before-close: 1
+          debug-only: true

.gitignore

@@ -131,6 +131,9 @@ dmypy.json
 # Pyre type checker
 .pyre/
 
+# macOS files
+.DS_Store
+
 .vscode
 .idea
 
@@ -146,9 +149,15 @@ debug*
 private*
 crazy_functions/test_project/pdf_and_word
 crazy_functions/test_samples
-request_llm/jittorllms
+request_llms/jittorllms
 multi-language
-request_llm/moss
+request_llms/moss
 media
 flagged
-request_llm/ChatGLM-6b-onnx-u8s8
+request_llms/ChatGLM-6b-onnx-u8s8
+.pre-commit-config.yaml
+test.*
+temp.*
+objdump*
+*.min.*.js
+TODO

Dockerfile

@@ -1,34 +1,39 @@
-# 此Dockerfile适用于“无本地模型”的环境构建，如果需要使用chatglm等本地模型或者latex运行依赖，请参考 docker-compose.yml
-# 如何构建: 先修改 `config.py`， 然后 `docker build -t gpt-academic . `
-# 如何运行(Linux下): `docker run --rm -it --net=host gpt-academic `
-# 如何运行(其他操作系统，选择任意一个固定端口50923): `docker run --rm -it -e WEB_PORT=50923 -p 50923:50923 gpt-academic `
+# 此Dockerfile适用于“无本地模型”的迷你运行环境构建
+# 如果需要使用chatglm等本地模型或者latex运行依赖，请参考 docker-compose.yml
+# - 如何构建: 先修改 `config.py`， 然后 `docker build -t gpt-academic . `
+# - 如何运行(Linux下): `docker run --rm -it --net=host gpt-academic `
+# - 如何运行(其他操作系统，选择任意一个固定端口50923): `docker run --rm -it -e WEB_PORT=50923 -p 50923:50923 gpt-academic `
 FROM python:3.11
 
 
-# 非必要步骤，更换pip源
+# 非必要步骤，更换pip源 （以下三行，可以删除）
 RUN echo '[global]' > /etc/pip.conf && \
     echo 'index-url = https://mirrors.aliyun.com/pypi/simple/' >> /etc/pip.conf && \
     echo 'trusted-host = mirrors.aliyun.com' >> /etc/pip.conf
 
 
-# 进入工作路径
+# 语音输出功能（以下两行，第一行更换阿里源，第二行安装ffmpeg，都可以删除）
+RUN UBUNTU_VERSION=$(awk -F= '/^VERSION_CODENAME=/{print $2}' /etc/os-release); echo "deb https://mirrors.aliyun.com/debian/ $UBUNTU_VERSION main non-free contrib" > /etc/apt/sources.list; apt-get update
+RUN apt-get install ffmpeg -y
+
+
+# 进入工作路径（必要）
 WORKDIR /gpt
 
 
-# 安装大部分依赖，利用Docker缓存加速以后的构建
+# 安装大部分依赖，利用Docker缓存加速以后的构建 （以下两行，可以删除）
 COPY requirements.txt ./
-COPY ./docs/gradio-3.32.2-py3-none-any.whl ./docs/gradio-3.32.2-py3-none-any.whl
 RUN pip3 install -r requirements.txt
 
 
-# 装载项目文件，安装剩余依赖
+# 装载项目文件，安装剩余依赖（必要）
 COPY . .
 RUN pip3 install -r requirements.txt
 
 
-# 非必要步骤，用于预热模块
+# 非必要步骤，用于预热模块（可以删除）
 RUN python3  -c 'from check_proxy import warm_up_modules; warm_up_modules()'
 
 
-# 启动
+# 启动（必要）
 CMD ["python3", "-u", "main.py"]

README.md

@@ -1,70 +1,98 @@
-> **Note**
+> [!IMPORTANT]
+> 2024.6.1: 版本3.80加入插件二级菜单功能（详见wiki）  
+> 2024.5.1: 加入Doc2x翻译PDF论文的功能，[查看详情](https://github.com/binary-husky/gpt_academic/wiki/Doc2x)  
+> 2024.3.11: 全力支持Qwen、GLM、DeepseekCoder等中文大语言模型！ SoVits语音克隆模块，[查看详情](https://www.bilibili.com/video/BV1Rp421S7tF/) 
+> 2024.1.17: 安装依赖时，请选择`requirements.txt`中**指定的版本**。 安装命令：`pip install -r requirements.txt`。本项目完全开源免费，您可通过订阅[在线服务](https://github.com/binary-husky/gpt_academic/wiki/online)的方式鼓励本项目的发展。
+
+<br>
+
+<div align=center>
+<h1 aligh="center">
+<img src="docs/logo.png" width="40"> GPT 学术优化 (GPT Academic)
+</h1>
+
+[![Github][Github-image]][Github-url]
+[![License][License-image]][License-url]
+[![Releases][Releases-image]][Releases-url]
+[![Installation][Installation-image]][Installation-url]
+[![Wiki][Wiki-image]][Wiki-url]
+[![PR][PRs-image]][PRs-url]
+
+[Github-image]: https://img.shields.io/badge/github-12100E.svg?style=flat-square
+[License-image]: https://img.shields.io/github/license/binary-husky/gpt_academic?label=License&style=flat-square&color=orange
+[Releases-image]: https://img.shields.io/github/release/binary-husky/gpt_academic?label=Release&style=flat-square&color=blue
+[Installation-image]: https://img.shields.io/badge/dynamic/json?color=blue&url=https://raw.githubusercontent.com/binary-husky/gpt_academic/master/version&query=$.version&label=Installation&style=flat-square
+[Wiki-image]: https://img.shields.io/badge/wiki-项目文档-black?style=flat-square
+[PRs-image]: https://img.shields.io/badge/PRs-welcome-pink?style=flat-square
+
+[Github-url]: https://github.com/binary-husky/gpt_academic
+[License-url]: https://github.com/binary-husky/gpt_academic/blob/master/LICENSE
+[Releases-url]: https://github.com/binary-husky/gpt_academic/releases
+[Installation-url]: https://github.com/binary-husky/gpt_academic#installation
+[Wiki-url]: https://github.com/binary-husky/gpt_academic/wiki
+[PRs-url]: https://github.com/binary-husky/gpt_academic/pulls
+
+
+</div>
+<br>
+
+**如果喜欢这个项目，请给它一个Star；如果您发明了好用的快捷键或插件，欢迎发pull requests！**
+
+If you like this project, please give it a Star.
+Read this in [English](docs/README.English.md) | [日本語](docs/README.Japanese.md) | [한국어](docs/README.Korean.md) | [Русский](docs/README.Russian.md) | [Français](docs/README.French.md). All translations have been provided by the project itself. To translate this project to arbitrary language with GPT, read and run [`multi_language.py`](multi_language.py) (experimental).
+<br>
+
+> [!NOTE]
+> 1.本项目中每个文件的功能都在[自译解报告](https://github.com/binary-husky/gpt_academic/wiki/GPT‐Academic项目自译解报告)`self_analysis.md`详细说明。随着版本的迭代，您也可以随时自行点击相关函数插件，调用GPT重新生成项目的自我解析报告。常见问题请查阅wiki。
+>    [![常规安装方法](https://img.shields.io/static/v1?label=&message=常规安装方法&color=gray)](#installation)  [![一键安装脚本](https://img.shields.io/static/v1?label=&message=一键安装脚本&color=gray)](https://github.com/binary-husky/gpt_academic/releases)  [![配置说明](https://img.shields.io/static/v1?label=&message=配置说明&color=gray)](https://github.com/binary-husky/gpt_academic/wiki/项目配置说明) [![wiki](https://img.shields.io/static/v1?label=&message=wiki&color=gray)]([https://github.com/binary-husky/gpt_academic/wiki/项目配置说明](https://github.com/binary-husky/gpt_academic/wiki))
 >
-> 2023.7.8: Gradio, Pydantic依赖调整，已修改 `requirements.txt`。请及时**更新代码**，安装依赖时，请严格选择`requirements.txt`中**指定的版本**
->
-> `pip install -r requirements.txt`
-
-
-# <div align=center><img src="docs/logo.png" width="40"> GPT 学术优化 (GPT Academic)</div>
-
-**如果喜欢这个项目，请给它一个Star；如果您发明了好用的快捷键或函数插件，欢迎发pull requests！**
-
-If you like this project, please give it a Star. If you've come up with more useful academic shortcuts or functional plugins, feel free to open an issue or pull request. We also have a README in [English|](docs/README_EN.md)[日本語|](docs/README_JP.md)[한국어|](https://github.com/mldljyh/ko_gpt_academic)[Русский|](docs/README_RS.md)[Français](docs/README_FR.md) translated by this project itself.
-To translate this project to arbitary language with GPT, read and run [`multi_language.py`](multi_language.py) (experimental).
-
-> **Note**
->
-> 1.请注意只有 **高亮(如红色)** 标识的函数插件（按钮）才支持读取文件，部分插件位于插件区的**下拉菜单**中。另外我们以**最高优先级**欢迎和处理任何新插件的PR。
->
-> 2.本项目中每个文件的功能都在自译解[`self_analysis.md`](https://github.com/binary-husky/gpt_academic/wiki/chatgpt-academic%E9%A1%B9%E7%9B%AE%E8%87%AA%E8%AF%91%E8%A7%A3%E6%8A%A5%E5%91%8A)详细说明。随着版本的迭代，您也可以随时自行点击相关函数插件，调用GPT重新生成项目的自我解析报告。常见问题汇总在[`wiki`](https://github.com/binary-husky/gpt_academic/wiki/%E5%B8%B8%E8%A7%81%E9%97%AE%E9%A2%98)当中。[安装方法](#installation)。
-> 
-> 3.本项目兼容并鼓励尝试国产大语言模型ChatGLM和Moss等等。支持多个api-key共存，可在配置文件中填写如`API_KEY="openai-key1,openai-key2,azure-key3,api2d-key4"`。需要临时更换`API_KEY`时，在输入区输入临时的`API_KEY`然后回车键提交后即可生效。
-
-
+> 2.本项目兼容并鼓励尝试国内中文大语言基座模型如通义千问，智谱GLM等。支持多个api-key共存，可在配置文件中填写如`API_KEY="openai-key1,openai-key2,azure-key3,api2d-key4"`。需要临时更换`API_KEY`时，在输入区输入临时的`API_KEY`然后回车键提交即可生效。
 
+<br><br>
 
 <div align="center">
 
 功能（⭐= 近期新增功能） | 描述
 --- | ---
-⭐[接入新模型](https://github.com/binary-husky/gpt_academic/wiki/%E5%A6%82%E4%BD%95%E5%88%87%E6%8D%A2%E6%A8%A1%E5%9E%8B)！ | ⭐阿里达摩院[通义千问](https://modelscope.cn/models/qwen/Qwen-7B-Chat/summary)，上海AI-Lab[书生](https://github.com/InternLM/InternLM)，讯飞[星火](https://xinghuo.xfyun.cn/)
-一键润色 | 支持一键润色、一键查找论文语法错误
-一键中英互译 | 一键中英互译
-一键代码解释 | 显示代码、解释代码、生成代码、给代码加注释
+⭐[接入新模型](https://github.com/binary-husky/gpt_academic/wiki/%E5%A6%82%E4%BD%95%E5%88%87%E6%8D%A2%E6%A8%A1%E5%9E%8B) | 百度[千帆](https://cloud.baidu.com/doc/WENXINWORKSHOP/s/Nlks5zkzu)与文心一言, 通义千问[Qwen](https://modelscope.cn/models/qwen/Qwen-7B-Chat/summary)，上海AI-Lab[书生](https://github.com/InternLM/InternLM)，讯飞[星火](https://xinghuo.xfyun.cn/)，[LLaMa2](https://huggingface.co/meta-llama/Llama-2-7b-chat-hf)，[智谱GLM4](https://open.bigmodel.cn/)，DALLE3, [DeepseekCoder](https://coder.deepseek.com/)
+⭐支持mermaid图像渲染 | 支持让GPT生成[流程图](https://www.bilibili.com/video/BV18c41147H9/)、状态转移图、甘特图、饼状图、GitGraph等等（3.7版本）
+⭐Arxiv论文精细翻译 ([Docker](https://github.com/binary-husky/gpt_academic/pkgs/container/gpt_academic_with_latex)) | [插件] 一键[以超高质量翻译arxiv论文](https://www.bilibili.com/video/BV1dz4y1v77A/)，目前最好的论文翻译工具
+⭐[实时语音对话输入](https://github.com/binary-husky/gpt_academic/blob/master/docs/use_audio.md) | [插件] 异步[监听音频](https://www.bilibili.com/video/BV1AV4y187Uy/)，自动断句，自动寻找回答时机
+⭐AutoGen多智能体插件 | [插件] 借助微软AutoGen，探索多Agent的智能涌现可能！
+⭐虚空终端插件 | [插件] 能够使用自然语言直接调度本项目其他插件
+润色、翻译、代码解释 | 一键润色、翻译、查找论文语法错误、解释代码
 [自定义快捷键](https://www.bilibili.com/video/BV14s4y1E7jN) | 支持自定义快捷键
-模块化设计 | 支持自定义强大的[函数插件](https://github.com/binary-husky/gpt_academic/tree/master/crazy_functions)，插件支持[热更新](https://github.com/binary-husky/gpt_academic/wiki/%E5%87%BD%E6%95%B0%E6%8F%92%E4%BB%B6%E6%8C%87%E5%8D%97)
-[自我程序剖析](https://www.bilibili.com/video/BV1cj411A7VW) | [函数插件] [一键读懂](https://github.com/binary-husky/gpt_academic/wiki/chatgpt-academic%E9%A1%B9%E7%9B%AE%E8%87%AA%E8%AF%91%E8%A7%A3%E6%8A%A5%E5%91%8A)本项目的源代码
-[程序剖析](https://www.bilibili.com/video/BV1cj411A7VW) | [函数插件] 一键可以剖析其他Python/C/C++/Java/Lua/...项目树
-读论文、[翻译](https://www.bilibili.com/video/BV1KT411x7Wn)论文 | [函数插件] 一键解读latex/pdf论文全文并生成摘要
-Latex全文[翻译](https://www.bilibili.com/video/BV1nk4y1Y7Js/)、[润色](https://www.bilibili.com/video/BV1FT411H7c5/) | [函数插件] 一键翻译或润色latex论文
-批量注释生成 | [函数插件] 一键批量生成函数注释
-Markdown[中英互译](https://www.bilibili.com/video/BV1yo4y157jV/) | [函数插件] 看到上面5种语言的[README](https://github.com/binary-husky/gpt_academic/blob/master/docs/README_EN.md)了吗？
-chat分析报告生成 | [函数插件] 运行后自动生成总结汇报
-[PDF论文全文翻译功能](https://www.bilibili.com/video/BV1KT411x7Wn) | [函数插件] PDF论文提取题目&摘要+翻译全文（多线程）
-[Arxiv小助手](https://www.bilibili.com/video/BV1LM4y1279X) | [函数插件] 输入arxiv文章url即可一键翻译摘要+下载PDF
-Latex论文一键校对 | [函数插件] 仿Grammarly对Latex文章进行语法、拼写纠错+输出对照PDF
-[谷歌学术统合小助手](https://www.bilibili.com/video/BV19L411U7ia) | [函数插件] 给定任意谷歌学术搜索页面URL，让gpt帮你[写relatedworks](https://www.bilibili.com/video/BV1GP411U7Az/)
-互联网信息聚合+GPT | [函数插件] 一键[让GPT从互联网获取信息](https://www.bilibili.com/video/BV1om4y127ck)回答问题，让信息永不过时
-⭐Arxiv论文精细翻译 ([Docker](https://github.com/binary-husky/gpt_academic/pkgs/container/gpt_academic_with_latex)) | [函数插件] 一键[以超高质量翻译arxiv论文](https://www.bilibili.com/video/BV1dz4y1v77A/)，目前最好的论文翻译工具
-⭐[实时语音对话输入](https://github.com/binary-husky/gpt_academic/blob/master/docs/use_audio.md) | [函数插件] 异步[监听音频](https://www.bilibili.com/video/BV1AV4y187Uy/)，自动断句，自动寻找回答时机
+模块化设计 | 支持自定义强大的[插件](https://github.com/binary-husky/gpt_academic/tree/master/crazy_functions)，插件支持[热更新](https://github.com/binary-husky/gpt_academic/wiki/%E5%87%BD%E6%95%B0%E6%8F%92%E4%BB%B6%E6%8C%87%E5%8D%97)
+[程序剖析](https://www.bilibili.com/video/BV1cj411A7VW) | [插件] 一键剖析Python/C/C++/Java/Lua/...项目树 或 [自我剖析](https://www.bilibili.com/video/BV1cj411A7VW)
+读论文、[翻译](https://www.bilibili.com/video/BV1KT411x7Wn)论文 | [插件] 一键解读latex/pdf论文全文并生成摘要
+Latex全文[翻译](https://www.bilibili.com/video/BV1nk4y1Y7Js/)、[润色](https://www.bilibili.com/video/BV1FT411H7c5/) | [插件] 一键翻译或润色latex论文
+批量注释生成 | [插件] 一键批量生成函数注释
+Markdown[中英互译](https://www.bilibili.com/video/BV1yo4y157jV/) | [插件] 看到上面5种语言的[README](https://github.com/binary-husky/gpt_academic/blob/master/docs/README.English.md)了吗？就是出自他的手笔
+[PDF论文全文翻译功能](https://www.bilibili.com/video/BV1KT411x7Wn) | [插件] PDF论文提取题目&摘要+翻译全文（多线程）
+[Arxiv小助手](https://www.bilibili.com/video/BV1LM4y1279X) | [插件] 输入arxiv文章url即可一键翻译摘要+下载PDF
+Latex论文一键校对 | [插件] 仿Grammarly对Latex文章进行语法、拼写纠错+输出对照PDF
+[谷歌学术统合小助手](https://www.bilibili.com/video/BV19L411U7ia) | [插件] 给定任意谷歌学术搜索页面URL，让gpt帮你[写relatedworks](https://www.bilibili.com/video/BV1GP411U7Az/)
+互联网信息聚合+GPT | [插件] 一键[让GPT从互联网获取信息](https://www.bilibili.com/video/BV1om4y127ck)回答问题，让信息永不过时
 公式/图片/表格显示 | 可以同时显示公式的[tex形式和渲染形式](https://user-images.githubusercontent.com/96192199/230598842-1d7fcddd-815d-40ee-af60-baf488a199df.png)，支持公式、代码高亮
-多线程函数插件支持 | 支持多线调用chatgpt，一键处理[海量文本](https://www.bilibili.com/video/BV1FT411H7c5/)或程序
 启动暗色[主题](https://github.com/binary-husky/gpt_academic/issues/173) | 在浏览器url后面添加```/?__theme=dark```可以切换dark主题
-[多LLM模型](https://www.bilibili.com/video/BV1wT411p7yf)支持 | 同时被GPT3.5、GPT4、[清华ChatGLM2](https://github.com/THUDM/ChatGLM2-6B)、[复旦MOSS](https://github.com/OpenLMLab/MOSS)同时伺候的感觉一定会很不错吧？
-⭐ChatGLM2微调模型 | 支持加载ChatGLM2微调模型，提供ChatGLM2微调辅助插件
+[多LLM模型](https://www.bilibili.com/video/BV1wT411p7yf)支持 | 同时被GPT3.5、GPT4、[清华ChatGLM2](https://github.com/THUDM/ChatGLM2-6B)、[复旦MOSS](https://github.com/OpenLMLab/MOSS)伺候的感觉一定会很不错吧？
 更多LLM模型接入，支持[huggingface部署](https://huggingface.co/spaces/qingxu98/gpt-academic) | 加入Newbing接口(新必应)，引入清华[Jittorllms](https://github.com/Jittor/JittorLLMs)支持[LLaMA](https://github.com/facebookresearch/llama)和[盘古α](https://openi.org.cn/pangu/)
-⭐[虚空终端](https://github.com/binary-husky/void-terminal)pip包 | 脱离GUI，在Python中直接调用本项目的函数插件（开发中）
+⭐[void-terminal](https://github.com/binary-husky/void-terminal) pip包 | 脱离GUI，在Python中直接调用本项目的所有函数插件（开发中）
 更多新功能展示 (图像生成等) …… | 见本文档结尾处 ……
 </div>
 
 
 - 新界面（修改`config.py`中的LAYOUT选项即可实现“左右布局”和“上下布局”的切换）
 <div align="center">
-<img src="https://user-images.githubusercontent.com/96192199/230361456-61078362-a966-4eb5-b49e-3c62ef18b860.gif" width="700" >
+<img src="https://user-images.githubusercontent.com/96192199/279702205-d81137c3-affd-4cd1-bb5e-b15610389762.gif" width="700" >
+</div>
+
+<div align="center">
+<img src="https://github.com/binary-husky/gpt_academic/assets/96192199/70ff1ec5-e589-4561-a29e-b831079b37fb.gif" width="700" >
 </div>
 
 
-- 所有按钮都通过读取functional.py动态生成，可随意加自定义功能，解放粘贴板
+- 所有按钮都通过读取functional.py动态生成，可随意加自定义功能，解放剪贴板
 <div align="center">
 <img src="https://user-images.githubusercontent.com/96192199/231975334-b4788e91-4887-412f-8b43-2b9c5f41d248.gif" width="700" >
 </div>
@@ -74,66 +102,99 @@ Latex论文一键校对 | [函数插件] 仿Grammarly对Latex文章进行语法
 <img src="https://user-images.githubusercontent.com/96192199/231980294-f374bdcb-3309-4560-b424-38ef39f04ebd.gif" width="700" >
 </div>
 
-- 如果输出包含公式，会同时以tex形式和渲染形式显示，方便复制和阅读
+- 如果输出包含公式，会以tex形式和渲染形式同时显示，方便复制和阅读
 <div align="center">
 <img src="https://user-images.githubusercontent.com/96192199/230598842-1d7fcddd-815d-40ee-af60-baf488a199df.png" width="700" >
 </div>
 
-- 懒得看项目代码？整个工程直接给chatgpt炫嘴里
+- 懒得看项目代码？直接把整个工程炫ChatGPT嘴里
 <div align="center">
 <img src="https://user-images.githubusercontent.com/96192199/226935232-6b6a73ce-8900-4aee-93f9-733c7e6fef53.png" width="700" >
 </div>
 
-- 多种大语言模型混合调用（ChatGLM + OpenAI-GPT3.5 + [API2D](https://api2d.com/)-GPT4）
+- 多种大语言模型混合调用（ChatGLM + OpenAI-GPT3.5 + GPT4）
 <div align="center">
 <img src="https://user-images.githubusercontent.com/96192199/232537274-deca0563-7aa6-4b5d-94a2-b7c453c47794.png" width="700" >
 </div>
 
+<br><br>
+
 # Installation
+
+```mermaid
+flowchart TD
+    A{"安装方法"} --> W1("I. 🔑直接运行 (Windows, Linux or MacOS)")
+    W1 --> W11["1. Python pip包管理依赖"]
+    W1 --> W12["2. Anaconda包管理依赖（推荐⭐）"]
+
+    A --> W2["II. 🐳使用Docker (Windows, Linux or MacOS)"]
+
+    W2 --> k1["1. 部署项目全部能力的大镜像（推荐⭐）"]
+    W2 --> k2["2. 仅在线模型（GPT, GLM4等）镜像"]
+    W2 --> k3["3. 在线模型 + Latex的大镜像"]
+
+    A --> W4["IV. 🚀其他部署方法"]
+    W4 --> C1["1. Windows/MacOS 一键安装运行脚本（推荐⭐）"]
+    W4 --> C2["2. Huggingface, Sealos远程部署"]
+    W4 --> C4["3. ... 其他 ..."]
+```
+
 ### 安装方法I：直接运行 (Windows, Linux or MacOS)
 
 1. 下载项目
-```sh
-git clone --depth=1 https://github.com/binary-husky/gpt_academic.git
-cd gpt_academic
-```
 
-2. 配置API_KEY
+    ```sh
+    git clone --depth=1 https://github.com/binary-husky/gpt_academic.git
+    cd gpt_academic
+    ```
 
-在`config.py`中，配置API KEY等设置，[点击查看特殊网络环境设置方法](https://github.com/binary-husky/gpt_academic/issues/1) 。
+2. 配置API_KEY等变量
 
-(P.S. 程序运行时会优先检查是否存在名为`config_private.py`的私密配置文件，并用其中的配置覆盖`config.py`的同名配置。因此，如果您能理解我们的配置读取逻辑，我们强烈建议您在`config.py`旁边创建一个名为`config_private.py`的新配置文件，并把`config.py`中的配置转移（复制）到`config_private.py`中。`config_private.py`不受git管控，可以让您的隐私信息更加安全。P.S.项目同样支持通过`环境变量`配置大多数选项，环境变量的书写格式参考`docker-compose`文件。读取优先级: `环境变量` > `config_private.py` > `config.py`)
+    在`config.py`中，配置API KEY等变量。[特殊网络环境设置方法](https://github.com/binary-husky/gpt_academic/issues/1)、[Wiki-项目配置说明](https://github.com/binary-husky/gpt_academic/wiki/项目配置说明)。
+
+    「 程序会优先检查是否存在名为`config_private.py`的私密配置文件，并用其中的配置覆盖`config.py`的同名配置。如您能理解以上读取逻辑，我们强烈建议您在`config.py`同路径下创建一个名为`config_private.py`的新配置文件，并使用`config_private.py`配置项目，从而确保自动更新时不会丢失配置 」。
+
+    「 支持通过`环境变量`配置项目，环境变量的书写格式参考`docker-compose.yml`文件或者我们的[Wiki页面](https://github.com/binary-husky/gpt_academic/wiki/项目配置说明)。配置读取优先级: `环境变量` > `config_private.py` > `config.py` 」。
 
 
 3. 安装依赖
-```sh
-# （选择I: 如熟悉python）（python版本3.9以上，越新越好），备注：使用官方pip源或者阿里pip源,临时换源方法：python -m pip install -r requirements.txt -i https://mirrors.aliyun.com/pypi/simple/
-python -m pip install -r requirements.txt
+    ```sh
+    # （选择I: 如熟悉python, python推荐版本 3.9 ~ 3.11）备注：使用官方pip源或者阿里pip源, 临时换源方法：python -m pip install -r requirements.txt -i https://mirrors.aliyun.com/pypi/simple/
+    python -m pip install -r requirements.txt
 
-# （选择II: 如不熟悉python）使用anaconda，步骤也是类似的 (https://www.bilibili.com/video/BV1rc411W7Dr)：
-conda create -n gptac_venv python=3.11    # 创建anaconda环境
-conda activate gptac_venv                 # 激活anaconda环境
-python -m pip install -r requirements.txt # 这个步骤和pip安装一样的步骤
-```
+    # （选择II: 使用Anaconda）步骤也是类似的 (https://www.bilibili.com/video/BV1rc411W7Dr)：
+    conda create -n gptac_venv python=3.11    # 创建anaconda环境
+    conda activate gptac_venv                 # 激活anaconda环境
+    python -m pip install -r requirements.txt # 这个步骤和pip安装一样的步骤
+    ```
 
 
 <details><summary>如果需要支持清华ChatGLM2/复旦MOSS/RWKV作为后端，请点击展开此处</summary>
 <p>
 
-【可选步骤】如果需要支持清华ChatGLM2/复旦MOSS作为后端，需要额外安装更多依赖（前提条件：熟悉Python + 用过Pytorch + 电脑配置够强）：
+【可选步骤】如果需要支持清华ChatGLM3/复旦MOSS作为后端，需要额外安装更多依赖（前提条件：熟悉Python + 用过Pytorch + 电脑配置够强）：
+
 ```sh
-# 【可选步骤I】支持清华ChatGLM2。清华ChatGLM备注：如果遇到"Call ChatGLM fail 不能正常加载ChatGLM的参数" 错误，参考如下： 1：以上默认安装的为torch+cpu版，使用cuda需要卸载torch重新安装torch+cuda； 2：如因本机配置不够无法加载模型，可以修改request_llm/bridge_chatglm.py中的模型精度, 将 AutoTokenizer.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True) 都修改为 AutoTokenizer.from_pretrained("THUDM/chatglm-6b-int4", trust_remote_code=True)
-python -m pip install -r request_llm/requirements_chatglm.txt  
+# 【可选步骤I】支持清华ChatGLM3。清华ChatGLM备注：如果遇到"Call ChatGLM fail 不能正常加载ChatGLM的参数" 错误，参考如下： 1：以上默认安装的为torch+cpu版，使用cuda需要卸载torch重新安装torch+cuda； 2：如因本机配置不够无法加载模型，可以修改request_llm/bridge_chatglm.py中的模型精度, 将 AutoTokenizer.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True) 都修改为 AutoTokenizer.from_pretrained("THUDM/chatglm-6b-int4", trust_remote_code=True)
+python -m pip install -r request_llms/requirements_chatglm.txt
 
 # 【可选步骤II】支持复旦MOSS
-python -m pip install -r request_llm/requirements_moss.txt
-git clone --depth=1 https://github.com/OpenLMLab/MOSS.git request_llm/moss  # 注意执行此行代码时，必须处于项目根路径
+python -m pip install -r request_llms/requirements_moss.txt
+git clone --depth=1 https://github.com/OpenLMLab/MOSS.git request_llms/moss  # 注意执行此行代码时，必须处于项目根路径
 
 # 【可选步骤III】支持RWKV Runner
 参考wiki：https://github.com/binary-husky/gpt_academic/wiki/%E9%80%82%E9%85%8DRWKV-Runner
 
 # 【可选步骤IV】确保config.py配置文件的AVAIL_LLM_MODELS包含了期望的模型，目前支持的全部模型如下(jittorllms系列目前仅支持docker方案)：
-AVAIL_LLM_MODELS = ["gpt-3.5-turbo", "api2d-gpt-3.5-turbo", "gpt-4", "api2d-gpt-4", "chatglm", "newbing", "moss"] # + ["jittorllms_rwkv", "jittorllms_pangualpha", "jittorllms_llama"]
+AVAIL_LLM_MODELS = ["gpt-3.5-turbo", "api2d-gpt-3.5-turbo", "gpt-4", "api2d-gpt-4", "chatglm", "moss"] # + ["jittorllms_rwkv", "jittorllms_pangualpha", "jittorllms_llama"]
+
+# 【可选步骤V】支持本地模型INT8,INT4量化（这里所指的模型本身不是量化版本，目前deepseek-coder支持，后面测试后会加入更多模型量化选择）
+pip install bitsandbyte
+# windows用户安装bitsandbytes需要使用下面bitsandbytes-windows-webui
+python -m pip install bitsandbytes --prefer-binary --extra-index-url=https://jllllll.github.io/bitsandbytes-windows-webui
+pip install -U git+https://github.com/huggingface/transformers.git
+pip install -U git+https://github.com/huggingface/accelerate.git
+pip install peft
 ```
 
 </p>
@@ -142,78 +203,63 @@ AVAIL_LLM_MODELS = ["gpt-3.5-turbo", "api2d-gpt-3.5-turbo", "gpt-4", "api2d-gpt-
 
 
 4. 运行
-```sh
-python main.py
-```
+    ```sh
+    python main.py
+    ```
 
 ### 安装方法II：使用Docker
 
-1. 仅ChatGPT（推荐大多数人选择，等价于docker-compose方案1）
+0. 部署项目的全部能力（这个是包含cuda和latex的大型镜像。但如果您网速慢、硬盘小，则不推荐该方法部署完整项目）
+[![fullcapacity](https://github.com/binary-husky/gpt_academic/actions/workflows/build-with-all-capacity.yml/badge.svg?branch=master)](https://github.com/binary-husky/gpt_academic/actions/workflows/build-with-all-capacity.yml)
+
+    ``` sh
+    # 修改docker-compose.yml，保留方案0并删除其他方案。然后运行：
+    docker-compose up
+    ```
+
+1. 仅ChatGPT + GLM4 + 文心一言+spark等在线模型（推荐大多数人选择）
 [![basic](https://github.com/binary-husky/gpt_academic/actions/workflows/build-without-local-llms.yml/badge.svg?branch=master)](https://github.com/binary-husky/gpt_academic/actions/workflows/build-without-local-llms.yml)
 [![basiclatex](https://github.com/binary-husky/gpt_academic/actions/workflows/build-with-latex.yml/badge.svg?branch=master)](https://github.com/binary-husky/gpt_academic/actions/workflows/build-with-latex.yml)
 [![basicaudio](https://github.com/binary-husky/gpt_academic/actions/workflows/build-with-audio-assistant.yml/badge.svg?branch=master)](https://github.com/binary-husky/gpt_academic/actions/workflows/build-with-audio-assistant.yml)
 
-``` sh
-git clone --depth=1 https://github.com/binary-husky/gpt_academic.git  # 下载项目
-cd gpt_academic                                 # 进入路径
-nano config.py                                      # 用任意文本编辑器编辑config.py, 配置 “Proxy”， “API_KEY” 以及 “WEB_PORT” (例如50923) 等
-docker build -t gpt-academic .                      # 安装
+    ``` sh
+    # 修改docker-compose.yml，保留方案1并删除其他方案。然后运行：
+    docker-compose up
+    ```
 
-#（最后一步-Linux操作系统）用`--net=host`更方便快捷
-docker run --rm -it --net=host gpt-academic
-#（最后一步-MacOS/Windows操作系统）只能用-p选项将容器上的端口(例如50923)暴露给主机上的端口
-docker run --rm -it -e WEB_PORT=50923 -p 50923:50923 gpt-academic
-```
-P.S. 如果需要依赖Latex的插件功能，请见Wiki。另外，您也可以直接使用docker-compose获取Latex功能（修改docker-compose.yml，保留方案4并删除其他方案）。
+P.S. 如果需要依赖Latex的插件功能，请见Wiki。另外，您也可以直接使用方案4或者方案0获取Latex功能。
 
-2. ChatGPT + ChatGLM2 + MOSS（需要熟悉Docker）
+2. ChatGPT + GLM3 + MOSS + LLAMA2 + 通义千问（需要熟悉[Nvidia Docker](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#installing-on-ubuntu-and-debian)运行时）
 [![chatglm](https://github.com/binary-husky/gpt_academic/actions/workflows/build-with-chatglm.yml/badge.svg?branch=master)](https://github.com/binary-husky/gpt_academic/actions/workflows/build-with-chatglm.yml)
 
-``` sh
-# 修改docker-compose.yml，保留方案2并删除其他方案。修改docker-compose.yml中方案2的配置，参考其中注释即可
-docker-compose up
-```
-
-3. ChatGPT + LLAMA + 盘古 + RWKV（需要熟悉Docker）
-[![jittorllms](https://github.com/binary-husky/gpt_academic/actions/workflows/build-with-jittorllms.yml/badge.svg?branch=master)](https://github.com/binary-husky/gpt_academic/actions/workflows/build-with-jittorllms.yml)
-
-``` sh
-# 修改docker-compose.yml，保留方案3并删除其他方案。修改docker-compose.yml中方案3的配置，参考其中注释即可
-docker-compose up
-```
+    ``` sh
+    # 修改docker-compose.yml，保留方案2并删除其他方案。然后运行：
+    docker-compose up
+    ```
 
 
-### 安装方法III：其他部署姿势
-1. 一键运行脚本。
-完全不熟悉python环境的Windows用户可以下载[Release](https://github.com/binary-husky/gpt_academic/releases)中发布的一键运行脚本安装无本地模型的版本。
-脚本的贡献来源是[oobabooga](https://github.com/oobabooga/one-click-installers)。
+### 安装方法III：其他部署方法
+1. **Windows一键运行脚本**。
+完全不熟悉python环境的Windows用户可以下载[Release](https://github.com/binary-husky/gpt_academic/releases)中发布的一键运行脚本安装无本地模型的版本。脚本贡献来源：[oobabooga](https://github.com/oobabooga/one-click-installers)。
 
-2. 使用docker-compose运行。
-请阅读docker-compose.yml后，按照其中的提示操作即可
+2. 使用第三方API、Azure等、文心一言、星火等，见[Wiki页面](https://github.com/binary-husky/gpt_academic/wiki/项目配置说明)
 
-3. 如何使用反代URL
-按照`config.py`中的说明配置API_URL_REDIRECT即可。
+3. 云服务器远程部署避坑指南。
+请访问[云服务器远程部署wiki](https://github.com/binary-husky/gpt_academic/wiki/%E4%BA%91%E6%9C%8D%E5%8A%A1%E5%99%A8%E8%BF%9C%E7%A8%8B%E9%83%A8%E7%BD%B2%E6%8C%87%E5%8D%97)
 
-4. 微软云AzureAPI
-按照`config.py`中的说明配置即可（AZURE_ENDPOINT等四个配置）
-
-5. 远程云服务器部署（需要云服务器知识与经验）。
-请访问[部署wiki-1](https://github.com/binary-husky/gpt_academic/wiki/%E4%BA%91%E6%9C%8D%E5%8A%A1%E5%99%A8%E8%BF%9C%E7%A8%8B%E9%83%A8%E7%BD%B2%E6%8C%87%E5%8D%97)
-
-6. 使用Sealos[一键部署](https://github.com/binary-husky/gpt_academic/issues/993)。
-
-7. 使用WSL2（Windows Subsystem for Linux 子系统）。
-请访问[部署wiki-2](https://github.com/binary-husky/gpt_academic/wiki/%E4%BD%BF%E7%94%A8WSL2%EF%BC%88Windows-Subsystem-for-Linux-%E5%AD%90%E7%B3%BB%E7%BB%9F%EF%BC%89%E9%83%A8%E7%BD%B2)
-
-8. 如何在二级网址（如`http://localhost/subpath`）下运行。
-请访问[FastAPI运行说明](docs/WithFastapi.md)
+4. 在其他平台部署&二级网址部署
+    - 使用Sealos[一键部署](https://github.com/binary-husky/gpt_academic/issues/993)。
+    - 使用WSL2（Windows Subsystem for Linux 子系统）。请访问[部署wiki-2](https://github.com/binary-husky/gpt_academic/wiki/%E4%BD%BF%E7%94%A8WSL2%EF%BC%88Windows-Subsystem-for-Linux-%E5%AD%90%E7%B3%BB%E7%BB%9F%EF%BC%89%E9%83%A8%E7%BD%B2)
+    - 如何在二级网址（如`http://localhost/subpath`）下运行。请访问[FastAPI运行说明](docs/WithFastapi.md)
 
+<br><br>
 
 # Advanced Usage
 ### I：自定义新的便捷按钮（学术快捷键）
-任意文本编辑器打开`core_functional.py`，添加条目如下，然后重启程序即可。（如果按钮已经添加成功并可见，那么前缀、后缀都支持热修改，无需重启程序即可生效。）
-例如
-```
+
+现在已可以通过UI中的`界面外观`菜单中的`自定义菜单`添加新的便捷按钮。如果需要在代码中定义，请使用任意文本编辑器打开`core_functional.py`，添加如下条目即可：
+
+```python
 "超级英译中": {
     # 前缀，会被加在你的输入之前。例如，用来描述你的要求，例如翻译、解释代码、润色等等
     "Prefix": "请翻译把下面一段内容成中文，然后用一个markdown表格逐一解释文中出现的专有名词：\n\n",
@@ -222,19 +268,20 @@ docker-compose up
     "Suffix": "",
 },
 ```
+
 <div align="center">
 <img src="https://user-images.githubusercontent.com/96192199/226899272-477c2134-ed71-4326-810c-29891fe4a508.png" width="500" >
 </div>
 
 ### II：自定义函数插件
-
 编写强大的函数插件来执行任何你想得到的和想不到的任务。
 本项目的插件编写、调试难度很低，只要您具备一定的python基础知识，就可以仿照我们提供的模板实现自己的插件功能。
 详情请参考[函数插件指南](https://github.com/binary-husky/gpt_academic/wiki/%E5%87%BD%E6%95%B0%E6%8F%92%E4%BB%B6%E6%8C%87%E5%8D%97)。
 
+<br><br>
 
-# Latest Update
-### I：新功能动态
+# Updates
+### I：动态
 
 1. 对话保存功能。在函数插件区调用 `保存当前的对话` 即可将当前对话保存为可读+可复原的html文件，
 另外在函数插件区（下拉菜单）调用 `载入对话历史存档` ，即可还原之前的会话。
@@ -249,10 +296,13 @@ Tip：不指定文件直接点击 `载入对话历史存档` 可以查看历史h
 <img src="https://github.com/binary-husky/gpt_academic/assets/96192199/9fdcc391-f823-464f-9322-f8719677043b" height="250" >
 </div>
 
-3. 生成报告。大部分插件都会在执行结束后，生成工作报告
+3. 虚空终端（从自然语言输入中，理解用户意图+自动调用其他插件）
+
+- 步骤一：输入 “ 请调用插件翻译PDF论文，地址为https://openreview.net/pdf?id=rJl0r3R9KX ”
+- 步骤二：点击“虚空终端”
+
 <div align="center">
-<img src="https://user-images.githubusercontent.com/96192199/227503770-fe29ce2c-53fd-47b0-b0ff-93805f0c2ff4.png" height="250" >
-<img src="https://user-images.githubusercontent.com/96192199/227504617-7a497bb3-0a2a-4b50-9a8a-95ae60ea7afd.png" height="250" >
+<img src="https://github.com/binary-husky/gpt_academic/assets/96192199/66f1b044-e9ff-4eed-9126-5d4f3668f1ed" width="500" >
 </div>
 
 4. 模块化功能设计，简单的接口却能支持强大的功能
@@ -272,35 +322,41 @@ Tip：不指定文件直接点击 `载入对话历史存档` 可以查看历史h
 <img src="https://user-images.githubusercontent.com/96192199/236432361-67739153-73e8-43fe-8111-b61296edabd9.png" width="500" >
 </div>
 
-7. 新增MOSS大语言模型支持
-<div align="center">
-<img src="https://user-images.githubusercontent.com/96192199/236639178-92836f37-13af-4fdd-984d-b4450fe30336.png" width="500" >
-</div>
-
-8. OpenAI图像生成
+7. OpenAI图像生成
 <div align="center">
 <img src="https://github.com/binary-husky/gpt_academic/assets/96192199/bc7ab234-ad90-48a0-8d62-f703d9e74665" width="500" >
 </div>
 
-9. OpenAI音频解析与总结
+8. 基于mermaid的流图、脑图绘制
 <div align="center">
-<img src="https://github.com/binary-husky/gpt_academic/assets/96192199/709ccf95-3aee-498a-934a-e1c22d3d5d5b" width="500" >
+<img src="https://github.com/binary-husky/gpt_academic/assets/96192199/c518b82f-bd53-46e2-baf5-ad1b081c1da4" width="500" >
 </div>
 
-10. Latex全文校对纠错
+9. Latex全文校对纠错
 <div align="center">
 <img src="https://github.com/binary-husky/gpt_academic/assets/96192199/651ccd98-02c9-4464-91e1-77a6b7d1b033" height="200" > ===>
 <img src="https://github.com/binary-husky/gpt_academic/assets/96192199/476f66d9-7716-4537-b5c1-735372c25adb" height="200">
 </div>
 
-11. 语言、主题切换
+10. 语言、主题切换
 <div align="center">
 <img src="https://github.com/binary-husky/gpt_academic/assets/96192199/b6799499-b6fb-4f0c-9c8e-1b441872f4e8" width="500" >
 </div>
 
 
+
 ### II：版本:
-- version 3.5(Todo): 使用自然语言调用本项目的所有函数插件（高优先级）
+- version 3.80(TODO): 优化AutoGen插件主题并设计一系列衍生插件
+- version 3.70: 引入Mermaid绘图，实现GPT画脑图等功能   
+- version 3.60: 引入AutoGen作为新一代插件的基石
+- version 3.57: 支持GLM3，星火v3，文心一言v4，修复本地模型的并发BUG
+- version 3.56: 支持动态追加基础功能按钮，新汇报PDF汇总页面
+- version 3.55: 重构前端界面，引入悬浮窗口与菜单栏
+- version 3.54: 新增动态代码解释器（Code Interpreter）（待完善）
+- version 3.53: 支持动态选择不同界面主题，提高稳定性&解决多用户冲突问题
+- version 3.50: 使用自然语言调用本项目的所有函数插件（虚空终端），支持插件分类，改进UI，设计新主题
+- version 3.49: 支持百度千帆平台和文心一言
+- version 3.48: 支持阿里达摩院通义千问，上海AI-Lab书生，讯飞星火
 - version 3.46: 支持完全脱手操作的实时语音对话
 - version 3.45: 支持自定义ChatGLM2微调模型
 - version 3.44: 正式支持Azure，优化界面易用性
@@ -311,25 +367,58 @@ Tip：不指定文件直接点击 `载入对话历史存档` 可以查看历史h
 - version 3.0: 对chatglm和其他小型llm的支持
 - version 2.6: 重构了插件结构，提高了交互性，加入更多插件
 - version 2.5: 自更新，解决总结大工程源代码时文本过长、token溢出的问题
-- version 2.4: (1)新增PDF全文翻译功能; (2)新增输入区切换位置的功能; (3)新增垂直布局选项; (4)多线程函数插件优化。
+- version 2.4: 新增PDF全文翻译功能; 新增输入区切换位置的功能
 - version 2.3: 增强多线程交互性
 - version 2.2: 函数插件支持热重载
 - version 2.1: 可折叠式布局
 - version 2.0: 引入模块化函数插件
 - version 1.0: 基础功能
 
-gpt_academic开发者QQ群-2：610599535
+GPT Academic开发者QQ群：`610599535`
 
 - 已知问题
     - 某些浏览器翻译插件干扰此软件前端的运行
-    - 官方Gradio目前有很多兼容性Bug，请务必使用`requirement.txt`安装Gradio
+    - 官方Gradio目前有很多兼容性问题，请**务必使用`requirement.txt`安装Gradio**
+
+```mermaid
+timeline LR
+    title GPT-Academic项目发展历程
+    section 2.x
+        1.0~2.2: 基础功能: 引入模块化函数插件: 可折叠式布局: 函数插件支持热重载
+        2.3~2.5: 增强多线程交互性: 新增PDF全文翻译功能: 新增输入区切换位置的功能: 自更新
+        2.6: 重构了插件结构: 提高了交互性: 加入更多插件
+    section 3.x
+        3.0~3.1: 对chatglm支持: 对其他小型llm支持: 支持同时问询多个gpt模型: 支持多个apikey负载均衡
+        3.2~3.3: 函数插件支持更多参数接口: 保存对话功能: 解读任意语言代码: 同时询问任意的LLM组合: 互联网信息综合功能
+        3.4: 加入arxiv论文翻译: 加入latex论文批改功能
+        3.44: 正式支持Azure: 优化界面易用性
+        3.46: 自定义ChatGLM2微调模型: 实时语音对话
+        3.49: 支持阿里达摩院通义千问: 上海AI-Lab书生: 讯飞星火: 支持百度千帆平台 & 文心一言
+        3.50: 虚空终端: 支持插件分类: 改进UI: 设计新主题
+        3.53: 动态选择不同界面主题: 提高稳定性: 解决多用户冲突问题
+        3.55: 动态代码解释器: 重构前端界面: 引入悬浮窗口与菜单栏
+        3.56: 动态追加基础功能按钮: 新汇报PDF汇总页面
+        3.57: GLM3, 星火v3: 支持文心一言v4: 修复本地模型的并发BUG
+        3.60: 引入AutoGen
+        3.70: 引入Mermaid绘图: 实现GPT画脑图等功能
+        3.80(TODO): 优化AutoGen插件主题: 设计衍生插件
+
+```
+
 
 ### III：主题
 可以通过修改`THEME`选项（config.py）变更主题
 1. `Chuanhu-Small-and-Beautiful` [网址](https://github.com/GaiZhenbiao/ChuanhuChatGPT/)
 
 
-### IV：参考与学习
+### IV：本项目的开发分支
+
+1. `master` 分支: 主分支，稳定版
+2. `frontier` 分支: 开发分支，测试版
+3. 如何[接入其他大模型](request_llms/README.md)
+4. 访问GPT-Academic的[在线服务并支持我们](https://github.com/binary-husky/gpt_academic/wiki/online)
+
+### V：参考与学习
 
 ```
 代码中参考了很多其他优秀项目中的设计，顺序不分先后：

check_proxy.py

@@ -1,34 +1,45 @@
+from loguru import logger
 
-def check_proxy(proxies):
+def check_proxy(proxies, return_ip=False):
     import requests
     proxies_https = proxies['https'] if proxies is not None else '无'
+    ip = None
     try:
         response = requests.get("https://ipapi.co/json/", proxies=proxies, timeout=4)
         data = response.json()
-        print(f'查询代理的地理位置，返回的结果是{data}')
         if 'country_name' in data:
             country = data['country_name']
             result = f"代理配置 {proxies_https}, 代理所在地：{country}"
+            if 'ip' in data: ip = data['ip']
         elif 'error' in data:
-            alternative = _check_with_backup_source(proxies)
+            alternative, ip = _check_with_backup_source(proxies)
             if alternative is None:
                 result = f"代理配置 {proxies_https}, 代理所在地：未知，IP查询频率受限"
             else:
                 result = f"代理配置 {proxies_https}, 代理所在地：{alternative}"
         else:
             result = f"代理配置 {proxies_https}, 代理数据解析失败：{data}"
-        print(result)
+        if not return_ip:
+            logger.warning(result)
             return result
+        else:
+            return ip
     except:
         result = f"代理配置 {proxies_https}, 代理所在地查询超时，代理可能无效"
-        print(result)
+        if not return_ip:
+            logger.warning(result)
             return result
+        else:
+            return ip
 
 def _check_with_backup_source(proxies):
     import random, string, requests
     random_string = ''.join(random.choices(string.ascii_letters + string.digits, k=32))
-    try: return requests.get(f"http://{random_string}.edns.ip-api.com/json", proxies=proxies, timeout=4).json()['dns']['geo']
-    except: return None
+    try:
+        res_json = requests.get(f"http://{random_string}.edns.ip-api.com/json", proxies=proxies, timeout=4).json()
+        return res_json['dns']['geo'], res_json['dns']['ip']
+    except:
+        return None, None
 
 def backup_and_download(current_version, remote_version):
     """
@@ -46,9 +57,9 @@ def backup_and_download(current_version, remote_version):
         return new_version_dir
     os.makedirs(new_version_dir)
     shutil.copytree('./', backup_dir, ignore=lambda x, y: ['history'])
-    proxies, = get_conf('proxies')
-    r = requests.get(
-        'https://github.com/binary-husky/chatgpt_academic/archive/refs/heads/master.zip', proxies=proxies, stream=True)
+    proxies = get_conf('proxies')
+    try:    r = requests.get('https://github.com/binary-husky/chatgpt_academic/archive/refs/heads/master.zip', proxies=proxies, stream=True)
+    except: r = requests.get('https://public.agent-matrix.com/publish/master.zip', proxies=proxies, stream=True)
     zip_file_path = backup_dir+'/master.zip'
     with open(zip_file_path, 'wb+') as f:
         f.write(r.content)
@@ -72,25 +83,25 @@ def patch_and_restart(path):
     import sys
     import time
     import glob
-    from colorful import print亮黄, print亮绿, print亮红
+    from shared_utils.colorful import log亮黄, log亮绿, log亮红
     # if not using config_private, move origin config.py as config_private.py
     if not os.path.exists('config_private.py'):
-        print亮黄('由于您没有设置config_private.py私密配置，现将您的现有配置移动至config_private.py以防止配置丢失，',
+        log亮黄('由于您没有设置config_private.py私密配置，现将您的现有配置移动至config_private.py以防止配置丢失，',
               '另外您可以随时在history子文件夹下找回旧版的程序。')
         shutil.copyfile('config.py', 'config_private.py')
     path_new_version = glob.glob(path + '/*-master')[0]
     dir_util.copy_tree(path_new_version, './')
-    print亮绿('代码已经更新，即将更新pip包依赖……')
-    for i in reversed(range(5)): time.sleep(1); print(i)
+    log亮绿('代码已经更新，即将更新pip包依赖……')
+    for i in reversed(range(5)): time.sleep(1); log亮绿(i)
     try:
         import subprocess
         subprocess.check_call([sys.executable, '-m', 'pip', 'install', '-r', 'requirements.txt'])
     except:
-        print亮红('pip包依赖安装出现问题，需要手动安装新增的依赖库 `python -m pip install -r requirements.txt`，然后在用常规的`python main.py`的方式启动。')
-    print亮绿('更新完成，您可以随时在history子文件夹下找回旧版的程序，5s之后重启')
-    print亮红('假如重启失败，您可能需要手动安装新增的依赖库 `python -m pip install -r requirements.txt`，然后在用常规的`python main.py`的方式启动。')
-    print(' ------------------------------ -----------------------------------')
-    for i in reversed(range(8)): time.sleep(1); print(i)
+        log亮红('pip包依赖安装出现问题，需要手动安装新增的依赖库 `python -m pip install -r requirements.txt`，然后在用常规的`python main.py`的方式启动。')
+    log亮绿('更新完成，您可以随时在history子文件夹下找回旧版的程序，5s之后重启')
+    log亮红('假如重启失败，您可能需要手动安装新增的依赖库 `python -m pip install -r requirements.txt`，然后在用常规的`python main.py`的方式启动。')
+    log亮绿(' ------------------------------ -----------------------------------')
+    for i in reversed(range(8)): time.sleep(1); log亮绿(i)
     os.execl(sys.executable, sys.executable, *sys.argv)
 
 
@@ -111,11 +122,10 @@ def auto_update(raise_error=False):
     try:
         from toolbox import get_conf
         import requests
-        import time
         import json
-        proxies, = get_conf('proxies')
-        response = requests.get(
-            "https://raw.githubusercontent.com/binary-husky/chatgpt_academic/master/version", proxies=proxies, timeout=5)
+        proxies = get_conf('proxies')
+        try:    response = requests.get("https://raw.githubusercontent.com/binary-husky/chatgpt_academic/master/version", proxies=proxies, timeout=5)
+        except: response = requests.get("https://public.agent-matrix.com/publish/version", proxies=proxies, timeout=5)
         remote_json_data = json.loads(response.text)
         remote_version = remote_json_data['version']
         if remote_json_data["show_feature"]:
@@ -126,10 +136,9 @@ def auto_update(raise_error=False):
             current_version = f.read()
             current_version = json.loads(current_version)['version']
         if (remote_version - current_version) >= 0.01-1e-5:
-            from colorful import print亮黄
-            print亮黄(
-                f'\n新版本可用。新版本:{remote_version}，当前版本:{current_version}。{new_feature}')
-            print('（1）Github更新地址:\nhttps://github.com/binary-husky/chatgpt_academic\n')
+            from shared_utils.colorful import log亮黄
+            log亮黄(f'\n新版本可用。新版本:{remote_version}，当前版本:{current_version}。{new_feature}')
+            logger.info('（1）Github更新地址:\nhttps://github.com/binary-husky/chatgpt_academic\n')
             user_instruction = input('（2）是否一键更新代码（Y+回车=确认，输入其他/无输入+回车=不更新）？')
             if user_instruction in ['Y', 'y']:
                 path = backup_and_download(current_version, remote_version)
@@ -140,9 +149,9 @@ def auto_update(raise_error=False):
                     if raise_error:
                         from toolbox import trimmed_format_exc
                         msg += trimmed_format_exc()
-                    print(msg)
+                    logger.warning(msg)
             else:
-                print('自动更新程序：已禁用')
+                logger.info('自动更新程序：已禁用')
                 return
         else:
             return
@@ -151,19 +160,29 @@ def auto_update(raise_error=False):
         if raise_error:
             from toolbox import trimmed_format_exc
             msg += trimmed_format_exc()
-        print(msg)
+        logger.info(msg)
 
 def warm_up_modules():
-    print('正在执行一些模块的预热...')
-    from request_llm.bridge_all import model_info
+    logger.info('正在执行一些模块的预热 ...')
+    from toolbox import ProxyNetworkActivate
+    from request_llms.bridge_all import model_info
+    with ProxyNetworkActivate("Warmup_Modules"):
         enc = model_info["gpt-3.5-turbo"]['tokenizer']
         enc.encode("模块预热", disallowed_special=())
         enc = model_info["gpt-4"]['tokenizer']
         enc.encode("模块预热", disallowed_special=())
 
+def warm_up_vectordb():
+    logger.info('正在执行一些模块的预热 ...')
+    from toolbox import ProxyNetworkActivate
+    with ProxyNetworkActivate("Warmup_Modules"):
+        import nltk
+        with ProxyNetworkActivate("Warmup_Modules"): nltk.download("punkt")
+
+
 if __name__ == '__main__':
     import os
     os.environ['no_proxy'] = '*'  # 避免代理网络产生意外污染
     from toolbox import get_conf
-    proxies, = get_conf('proxies')
+    proxies = get_conf('proxies')
     check_proxy(proxies)

config.py

@@ -11,17 +11,17 @@
 API_KEY = "此处填API密钥"    # 可同时填写多个API-KEY，用英文逗号分割，例如API_KEY = "sk-openaikey1,sk-openaikey2,fkxxxx-api2dkey3,azure-apikey4"
 
 
-# [step 2]>> 改为True应用代理，如果直接在海外服务器部署，此处不修改
+# [step 2]>> 改为True应用代理，如果直接在海外服务器部署，此处不修改；如果使用本地或无地域限制的大模型时，此处也不需要修改
 USE_PROXY = False
 if USE_PROXY:
     """
+    代理网络的地址，打开你的代理软件查看代理协议(socks5h / http)、地址(localhost)和端口(11284)
     填写格式是 [协议]://  [地址] :[端口]，填写之前不要忘记把USE_PROXY改成True，如果直接在海外服务器部署，此处不修改
             <配置教程&视频教程> https://github.com/binary-husky/gpt_academic/issues/1>
     [协议] 常见协议无非socks5h/http; 例如 v2**y 和 ss* 的默认本地协议是socks5h; 而cl**h 的默认本地协议是http
-    [地址] 懂的都懂，不懂就填localhost或者127.0.0.1肯定错不了（localhost意思是代理软件安装在本机上）
+    [地址] 填localhost或者127.0.0.1（localhost意思是代理软件安装在本机上）
     [端口] 在代理软件的设置里找。虽然不同的代理软件界面不一样，但端口号都应该在最显眼的位置上
     """
-    # 代理网络的地址，打开你的*学*网软件查看代理的协议(socks5h / http)、地址(localhost)和端口(11284)
     proxies = {
         #          [协议]://  [地址]  :[端口]
         "http":  "socks5h://localhost:11284",  # 再例如  "http":  "http://127.0.0.1:7890",
@@ -30,11 +30,44 @@ if USE_PROXY:
 else:
     proxies = None
 
-# ------------------------------------ 以下配置可以优化体验, 但大部分场合下并不需要修改 ------------------------------------
+# [step 3]>> 模型选择是 (注意: LLM_MODEL是默认选中的模型, 它*必须*被包含在AVAIL_LLM_MODELS列表中 )
+LLM_MODEL = "gpt-3.5-turbo-16k" # 可选 ↓↓↓
+AVAIL_LLM_MODELS = ["gpt-4-1106-preview", "gpt-4-turbo-preview", "gpt-4-vision-preview",
+                    "gpt-4o", "gpt-4o-mini", "gpt-4-turbo", "gpt-4-turbo-2024-04-09",
+                    "gpt-3.5-turbo-1106", "gpt-3.5-turbo-16k", "gpt-3.5-turbo", "azure-gpt-3.5",
+                    "gpt-4", "gpt-4-32k", "azure-gpt-4", "glm-4", "glm-4v", "glm-3-turbo",
+                    "gemini-1.5-pro", "chatglm3"
+                    ]
+
+EMBEDDING_MODEL = "text-embedding-3-small"
+
+# --- --- --- ---
+# P.S. 其他可用的模型还包括
+# AVAIL_LLM_MODELS = [
+#   "glm-4-0520", "glm-4-air", "glm-4-airx", "glm-4-flash",
+#   "qianfan", "deepseekcoder",
+#   "spark", "sparkv2", "sparkv3", "sparkv3.5", "sparkv4",
+#   "qwen-turbo", "qwen-plus", "qwen-max", "qwen-local",
+#   "moonshot-v1-128k", "moonshot-v1-32k", "moonshot-v1-8k",
+#   "gpt-3.5-turbo-0613", "gpt-3.5-turbo-16k-0613", "gpt-3.5-turbo-0125", "gpt-4o-2024-05-13"
+#   "claude-3-haiku-20240307","claude-3-sonnet-20240229","claude-3-opus-20240229", "claude-2.1", "claude-instant-1.2",
+#   "moss", "llama2", "chatglm_onnx", "internlm", "jittorllms_pangualpha", "jittorllms_llama",
+#   "deepseek-chat" ,"deepseek-coder",
+#   "gemini-1.5-flash",
+#   "yi-34b-chat-0205","yi-34b-chat-200k","yi-large","yi-medium","yi-spark","yi-large-turbo","yi-large-preview",
+# ]
+# --- --- --- ---
+# 此外，您还可以在接入one-api/vllm/ollama/Openroute时，
+# 使用"one-api-*","vllm-*","ollama-*","openrouter-*"前缀直接使用非标准方式接入的模型，例如
+# AVAIL_LLM_MODELS = ["one-api-claude-3-sonnet-20240229(max_token=100000)", "ollama-phi3(max_token=4096)","openrouter-openai/gpt-4o-mini","openrouter-openai/chatgpt-4o-latest"]
+# --- --- --- ---
+
+
+# --------------- 以下配置可以优化体验 ---------------
 
 # 重新URL重新定向，实现更换API_URL的作用（高危设置! 常规情况下不要修改! 通过修改此设置，您将把您的API-KEY和对话隐私完全暴露给您设定的中间人！）
 # 格式: API_URL_REDIRECT = {"https://api.openai.com/v1/chat/completions": "在这里填写重定向的api.openai.com的URL"}
-# 举例: API_URL_REDIRECT = {"https://api.openai.com/v1/chat/completions": "https://reverse-proxy-url/v1/chat/completions"}
+# 举例: API_URL_REDIRECT = {"https://api.openai.com/v1/chat/completions": "https://reverse-proxy-url/v1/chat/completions", "http://localhost:11434/api/chat": "在这里填写您ollama的URL"}
 API_URL_REDIRECT = {}
 
 
@@ -43,7 +76,17 @@ API_URL_REDIRECT = {}
 DEFAULT_WORKER_NUM = 3
 
 
-# 对话窗的高度
+# 色彩主题, 可选 ["Default", "Chuanhu-Small-and-Beautiful", "High-Contrast"]
+# 更多主题, 请查阅Gradio主题商店: https://huggingface.co/spaces/gradio/theme-gallery 可选 ["Gstaff/Xkcd", "NoCrypt/Miku", ...]
+THEME = "Default"
+AVAIL_THEMES = ["Default", "Chuanhu-Small-and-Beautiful", "High-Contrast", "Gstaff/Xkcd", "NoCrypt/Miku"]
+
+
+# 默认的系统提示词（system prompt）
+INIT_SYS_PROMPT = "Serve me as a writing and programming assistant."
+
+
+# 对话窗的高度 （仅在LAYOUT="TOP-DOWN"时生效）
 CHATBOT_HEIGHT = 1115
 
 
@@ -53,7 +96,10 @@ CODE_HIGHLIGHT = True
 
 # 窗口布局
 LAYOUT = "LEFT-RIGHT"   # "LEFT-RIGHT"（左右布局） # "TOP-DOWN"（上下布局）
-DARK_MODE = True        # 暗色模式 / 亮色模式
+
+
+# 暗色模式 / 亮色模式
+DARK_MODE = True
 
 
 # 发送请求到OpenAI后，等待多久判定为超时
@@ -64,18 +110,40 @@ TIMEOUT_SECONDS = 30
 WEB_PORT = -1
 
 
+# 是否自动打开浏览器页面
+AUTO_OPEN_BROWSER = True
+
+
 # 如果OpenAI不响应（网络卡顿、代理失败、KEY失效），重试的次数限制
 MAX_RETRY = 2
 
 
-# 模型选择是 (注意: LLM_MODEL是默认选中的模型, 它*必须*被包含在AVAIL_LLM_MODELS列表中 )
-LLM_MODEL = "gpt-3.5-turbo" # 可选 ↓↓↓
-AVAIL_LLM_MODELS = ["gpt-3.5-turbo-16k", "gpt-3.5-turbo", "azure-gpt-3.5", "api2d-gpt-3.5-turbo", "gpt-4", "api2d-gpt-4", "chatglm", "moss", "newbing", "stack-claude"]
-# P.S. 其他可用的模型还包括 ["qwen", "gpt-3.5-turbo-0613", "gpt-3.5-turbo-16k-0613", "spark", "chatglm_onnx", "claude-1-100k", "claude-2", "internlm", "jittorllms_rwkv", "jittorllms_pangualpha", "jittorllms_llama"]
+# 插件分类默认选项
+DEFAULT_FN_GROUPS = ['对话', '编程', '学术', '智能体']
 
 
-# ChatGLM(2) Finetune Model Path （如果使用ChatGLM2微调模型，需要把"chatglmft"加入AVAIL_LLM_MODELS中）
-ChatGLM_PTUNING_CHECKPOINT = "" # 例如"/home/hmp/ChatGLM2-6B/ptuning/output/6b-pt-128-1e-2/checkpoint-100"
+# 定义界面上“询问多个GPT模型”插件应该使用哪些模型，请从AVAIL_LLM_MODELS中选择，并在不同模型之间用`&`间隔，例如"gpt-3.5-turbo&chatglm3&azure-gpt-4"
+MULTI_QUERY_LLM_MODELS = "gpt-3.5-turbo&chatglm3"
+
+
+# 选择本地模型变体（只有当AVAIL_LLM_MODELS包含了对应本地模型时，才会起作用）
+# 如果你选择Qwen系列的模型，那么请在下面的QWEN_MODEL_SELECTION中指定具体的模型
+# 也可以是具体的模型路径
+QWEN_LOCAL_MODEL_SELECTION = "Qwen/Qwen-1_8B-Chat-Int8"
+
+
+# 接入通义千问在线大模型 https://dashscope.console.aliyun.com/
+DASHSCOPE_API_KEY = "" # 阿里灵积云API_KEY
+
+
+# 百度千帆（LLM_MODEL="qianfan"）
+BAIDU_CLOUD_API_KEY = ''
+BAIDU_CLOUD_SECRET_KEY = ''
+BAIDU_CLOUD_QIANFAN_MODEL = 'ERNIE-Bot'    # 可选 "ERNIE-Bot-4"(文心大模型4.0), "ERNIE-Bot"(文心一言), "ERNIE-Bot-turbo", "BLOOMZ-7B", "Llama-2-70B-Chat", "Llama-2-13B-Chat", "Llama-2-7B-Chat", "ERNIE-Speed-128K", "ERNIE-Speed-8K", "ERNIE-Lite-8K"
+
+
+# 如果使用ChatGLM2微调模型，请把 LLM_MODEL="chatglmft"，并在此处指定模型路径
+CHATGLM_PTUNING_CHECKPOINT = "" # 例如"/home/hmp/ChatGLM2-6B/ptuning/output/6b-pt-128-1e-2/checkpoint-100"
 
 
 # 本地LLM模型如ChatGLM的执行方式 CPU/GPU
@@ -91,10 +159,6 @@ CONCURRENT_COUNT = 100
 AUTO_CLEAR_TXT = False
 
 
-# 色彩主体，可选 ["Default", "Chuanhu-Small-and-Beautiful"]
-THEME = "Default"
-
-
 # 加一个live2d装饰
 ADD_WAIFU = False
 
@@ -104,33 +168,37 @@ ADD_WAIFU = False
 AUTHENTICATION = []
 
 
-# 如果需要在二级路径下运行（常规情况下，不要修改!!）（需要配合修改main.py才能生效!）
+# 如果需要在二级路径下运行（常规情况下，不要修改!!）
+# （举例 CUSTOM_PATH = "/gpt_academic"，可以让软件运行在 http://ip:port/gpt_academic/ 下。）
 CUSTOM_PATH = "/"
 
 
+# HTTPS 秘钥和证书（不需要修改）
+SSL_KEYFILE = ""
+SSL_CERTFILE = ""
+
+
 # 极少数情况下，openai的官方KEY需要伴随组织编码（格式如org-xxxxxxxxxxxxxxxxxxxxxxxx）使用
 API_ORG = ""
 
 
-# 如果需要使用Slack Claude，使用教程详情见 request_llm/README.md
+# 如果需要使用Slack Claude，使用教程详情见 request_llms/README.md
 SLACK_CLAUDE_BOT_ID = ''
 SLACK_CLAUDE_USER_TOKEN = ''
 
 
-# 如果需要使用AZURE 详情请见额外文档 docs\use_azure.md
+# 如果需要使用AZURE（方法一：单个azure模型部署）详情请见额外文档 docs\use_azure.md
 AZURE_ENDPOINT = "https://你亲手写的api名称.openai.azure.com/"
 AZURE_API_KEY = "填入azure openai api的密钥"    # 建议直接在API_KEY处填写，该选项即将被弃用
 AZURE_ENGINE = "填入你亲手写的部署名"            # 读 docs\use_azure.md
 
 
-# 使用Newbing
-NEWBING_STYLE = "creative"  # ["creative", "balanced", "precise"]
-NEWBING_COOKIES = """
-put your new bing cookies here
-"""
+# 如果需要使用AZURE（方法二：多个azure模型部署+动态切换）详情请见额外文档 docs\use_azure.md
+AZURE_CFG_ARRAY = {}
 
 
-# 阿里云实时语音识别 配置难度较高 仅建议高手用户使用 参考 https://github.com/binary-husky/gpt_academic/blob/master/docs/use_audio.md
+# 阿里云实时语音识别 配置难度较高
+# 参考 https://github.com/binary-husky/gpt_academic/blob/master/docs/use_audio.md
 ENABLE_AUDIO = False
 ALIYUN_TOKEN=""     # 例如 f37f30e0f9934c34a992f6f64f7eba4f
 ALIYUN_APPKEY=""    # 例如 RoPlZrM88DnAFkZK
@@ -138,15 +206,212 @@ ALIYUN_ACCESSKEY="" # （无需填写）
 ALIYUN_SECRET=""    # （无需填写）
 
 
+# GPT-SOVITS 文本转语音服务的运行地址（将语言模型的生成文本朗读出来）
+TTS_TYPE = "EDGE_TTS" # EDGE_TTS / LOCAL_SOVITS_API / DISABLE
+GPT_SOVITS_URL = ""
+EDGE_TTS_VOICE = "zh-CN-XiaoxiaoNeural"
+
+
 # 接入讯飞星火大模型 https://console.xfyun.cn/services/iat
 XFYUN_APPID = "00000000"
 XFYUN_API_SECRET = "bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb"
 XFYUN_API_KEY = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
 
 
+# 接入智谱大模型
+ZHIPUAI_API_KEY = ""
+ZHIPUAI_MODEL = "" # 此选项已废弃，不再需要填写
+
+
 # Claude API KEY
 ANTHROPIC_API_KEY = ""
 
 
+# 月之暗面 API KEY
+MOONSHOT_API_KEY = ""
+
+
+# 零一万物(Yi Model) API KEY
+YIMODEL_API_KEY = ""
+
+
+# 深度求索(DeepSeek) API KEY，默认请求地址为"https://api.deepseek.com/v1/chat/completions"
+DEEPSEEK_API_KEY = ""
+
+
+# 紫东太初大模型 https://ai-maas.wair.ac.cn
+TAICHU_API_KEY = ""
+
+
+# Mathpix 拥有执行PDF的OCR功能，但是需要注册账号
+MATHPIX_APPID = ""
+MATHPIX_APPKEY = ""
+
+
+# DOC2X的PDF解析服务，注册账号并获取API KEY: https://doc2x.noedgeai.com/login
+DOC2X_API_KEY = ""
+
+
 # 自定义API KEY格式
 CUSTOM_API_KEY_PATTERN = ""
+
+
+# Google Gemini API-Key
+GEMINI_API_KEY = ''
+
+
+# HUGGINGFACE的TOKEN，下载LLAMA时起作用 https://huggingface.co/docs/hub/security-tokens
+HUGGINGFACE_ACCESS_TOKEN = "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"
+
+
+# GROBID服务器地址（填写多个可以均衡负载），用于高质量地读取PDF文档
+# 获取方法：复制以下空间https://huggingface.co/spaces/qingxu98/grobid，设为public，然后GROBID_URL = "https://(你的hf用户名如qingxu98)-(你的填写的空间名如grobid).hf.space"
+GROBID_URLS = [
+    "https://qingxu98-grobid.hf.space","https://qingxu98-grobid2.hf.space","https://qingxu98-grobid3.hf.space",
+    "https://qingxu98-grobid4.hf.space","https://qingxu98-grobid5.hf.space", "https://qingxu98-grobid6.hf.space",
+    "https://qingxu98-grobid7.hf.space", "https://qingxu98-grobid8.hf.space",
+]
+
+
+# Searxng互联网检索服务
+SEARXNG_URL = "https://cloud-1.agent-matrix.com/"
+
+
+# 是否允许通过自然语言描述修改本页的配置，该功能具有一定的危险性，默认关闭
+ALLOW_RESET_CONFIG = False
+
+
+# 在使用AutoGen插件时，是否使用Docker容器运行代码
+AUTOGEN_USE_DOCKER = False
+
+
+# 临时的上传文件夹位置，请尽量不要修改
+PATH_PRIVATE_UPLOAD = "private_upload"
+
+
+# 日志文件夹的位置，请尽量不要修改
+PATH_LOGGING = "gpt_log"
+
+
+# 存储翻译好的arxiv论文的路径，请尽量不要修改
+ARXIV_CACHE_DIR = "gpt_log/arxiv_cache"
+
+
+# 除了连接OpenAI之外，还有哪些场合允许使用代理，请尽量不要修改
+WHEN_TO_USE_PROXY = ["Download_LLM", "Download_Gradio_Theme", "Connect_Grobid",
+                     "Warmup_Modules", "Nougat_Download", "AutoGen", "Connect_OpenAI_Embedding"]
+
+
+# 启用插件热加载
+PLUGIN_HOT_RELOAD = False
+
+
+# 自定义按钮的最大数量限制
+NUM_CUSTOM_BASIC_BTN = 4
+
+
+
+"""
+--------------- 配置关联关系说明 ---------------
+
+在线大模型配置关联关系示意图
+│
+├── "gpt-3.5-turbo" 等openai模型
+│   ├── API_KEY
+│   ├── CUSTOM_API_KEY_PATTERN（不常用）
+│   ├── API_ORG（不常用）
+│   └── API_URL_REDIRECT（不常用）
+│
+├── "azure-gpt-3.5" 等azure模型（单个azure模型，不需要动态切换）
+│   ├── API_KEY
+│   ├── AZURE_ENDPOINT
+│   ├── AZURE_API_KEY
+│   ├── AZURE_ENGINE
+│   └── API_URL_REDIRECT
+│
+├── "azure-gpt-3.5" 等azure模型（多个azure模型，需要动态切换，高优先级）
+│   └── AZURE_CFG_ARRAY
+│
+├── "spark" 星火认知大模型 spark & sparkv2
+│   ├── XFYUN_APPID
+│   ├── XFYUN_API_SECRET
+│   └── XFYUN_API_KEY
+│
+├── "claude-3-opus-20240229" 等claude模型
+│   └── ANTHROPIC_API_KEY
+│
+├── "stack-claude"
+│   ├── SLACK_CLAUDE_BOT_ID
+│   └── SLACK_CLAUDE_USER_TOKEN
+│
+├── "qianfan" 百度千帆大模型库
+│   ├── BAIDU_CLOUD_QIANFAN_MODEL
+│   ├── BAIDU_CLOUD_API_KEY
+│   └── BAIDU_CLOUD_SECRET_KEY
+│
+├── "glm-4", "glm-3-turbo", "zhipuai" 智谱AI大模型
+│   └── ZHIPUAI_API_KEY
+│
+├── "yi-34b-chat-0205", "yi-34b-chat-200k" 等零一万物(Yi Model)大模型
+│   └── YIMODEL_API_KEY
+│
+├── "qwen-turbo" 等通义千问大模型
+│   └──  DASHSCOPE_API_KEY
+│
+├── "Gemini"
+│   └──  GEMINI_API_KEY
+│
+└── "one-api-...(max_token=...)" 用一种更方便的方式接入one-api多模型管理界面
+    ├── AVAIL_LLM_MODELS
+    ├── API_KEY
+    └── API_URL_REDIRECT
+
+
+本地大模型示意图
+│
+├── "chatglm3"
+├── "chatglm"
+├── "chatglm_onnx"
+├── "chatglmft"
+├── "internlm"
+├── "moss"
+├── "jittorllms_pangualpha"
+├── "jittorllms_llama"
+├── "deepseekcoder"
+├── "qwen-local"
+├──  RWKV的支持见Wiki
+└── "llama2"
+
+
+用户图形界面布局依赖关系示意图
+│
+├── CHATBOT_HEIGHT 对话窗的高度
+├── CODE_HIGHLIGHT 代码高亮
+├── LAYOUT 窗口布局
+├── DARK_MODE 暗色模式 / 亮色模式
+├── DEFAULT_FN_GROUPS 插件分类默认选项
+├── THEME 色彩主题
+├── AUTO_CLEAR_TXT 是否在提交时自动清空输入框
+├── ADD_WAIFU 加一个live2d装饰
+└── ALLOW_RESET_CONFIG 是否允许通过自然语言描述修改本页的配置，该功能具有一定的危险性
+
+
+插件在线服务配置依赖关系示意图
+│
+├── 互联网检索
+│   └── SEARXNG_URL
+│
+├── 语音功能
+│   ├── ENABLE_AUDIO
+│   ├── ALIYUN_TOKEN
+│   ├── ALIYUN_APPKEY
+│   ├── ALIYUN_ACCESSKEY
+│   └── ALIYUN_SECRET
+│
+└── PDF文档精准解析
+    ├── GROBID_URLS
+    ├── MATHPIX_APPID
+    └── MATHPIX_APPKEY
+
+
+"""

core_functional.py

@@ -3,81 +3,145 @@
 # 'stop' 颜色对应 theme.py 中的 color_er
 import importlib
 from toolbox import clear_line_break
-
+from toolbox import apply_gpt_academic_string_mask_langbased
+from toolbox import build_gpt_academic_masked_string_langbased
+from textwrap import dedent
 
 def get_core_functions():
     return {
-        "英语学术润色": {
-            # 前缀，会被加在你的输入之前。例如，用来描述你的要求，例如翻译、解释代码、润色等等
-            "Prefix":   r"Below is a paragraph from an academic paper. Polish the writing to meet the academic style, " +
-                        r"improve the spelling, grammar, clarity, concision and overall readability. When necessary, rewrite the whole sentence. " +
-                        r"Furthermore, list all modification and explain the reasons to do so in markdown table." + "\n\n",
-            # 后缀，会被加在你的输入之后。例如，配合前缀可以把你的输入内容用引号圈起来
+
+        "学术语料润色": {
+            # [1*] 前缀字符串，会被加在你的输入之前。例如，用来描述你的要求，例如翻译、解释代码、润色等等。
+            #      这里填一个提示词字符串就行了，这里为了区分中英文情景搞复杂了一点
+            "Prefix":   build_gpt_academic_masked_string_langbased(
+                            text_show_english=
+                                r"Below is a paragraph from an academic paper. Polish the writing to meet the academic style, "
+                                r"improve the spelling, grammar, clarity, concision and overall readability. When necessary, rewrite the whole sentence. "
+                                r"Firstly, you should provide the polished paragraph (in English). "
+                                r"Secondly, you should list all your modification and explain the reasons to do so in markdown table.",
+                            text_show_chinese=
+                                r"作为一名中文学术论文写作改进助理，你的任务是改进所提供文本的拼写、语法、清晰、简洁和整体可读性，"
+                                r"同时分解长句，减少重复，并提供改进建议。请先提供文本的更正版本，然后在markdown表格中列出修改的内容，并给出修改的理由:"
+                        ) + "\n\n",
+            # [2*] 后缀字符串，会被加在你的输入之后。例如，配合前缀可以把你的输入内容用引号圈起来
             "Suffix":   r"",
-            # 按钮颜色 (默认 secondary)
+            # [3] 按钮颜色 (可选参数，默认 secondary)
             "Color":    r"secondary",
-            # 按钮是否可见 (默认 True，即可见)
+            # [4] 按钮是否可见 (可选参数，默认 True，即可见)
             "Visible": True,
-            # 是否在触发时清除历史 (默认 False，即不处理之前的对话历史)
-            "AutoClearHistory": False
+            # [5] 是否在触发时清除历史 (可选参数，默认 False，即不处理之前的对话历史)
+            "AutoClearHistory": False,
+            # [6] 文本预处理 （可选参数，默认 None，举例：写个函数移除所有的换行符）
+            "PreProcess": None,
+            # [7] 模型选择 （可选参数。如不设置，则使用当前全局模型；如设置，则用指定模型覆盖全局模型。）
+            # "ModelOverride": "gpt-3.5-turbo", # 主要用途：强制点击此基础功能按钮时，使用指定的模型。
         },
-        "中文学术润色": {
-            "Prefix":   r"作为一名中文学术论文写作改进助理，你的任务是改进所提供文本的拼写、语法、清晰、简洁和整体可读性，" +
-                        r"同时分解长句，减少重复，并提供改进建议。请只提供文本的更正版本，避免包括解释。请编辑以下文本" + "\n\n",
-            "Suffix":   r"",
+
+
+        "总结绘制脑图": {
+            # 前缀，会被加在你的输入之前。例如，用来描述你的要求，例如翻译、解释代码、润色等等
+            "Prefix":   '''"""\n\n''',
+            # 后缀，会被加在你的输入之后。例如，配合前缀可以把你的输入内容用引号圈起来
+            "Suffix":
+                # dedent() 函数用于去除多行字符串的缩进
+                dedent("\n\n"+r'''
+                    """
+
+                    使用mermaid flowchart对以上文本进行总结，概括上述段落的内容以及内在逻辑关系，例如：
+
+                    以下是对以上文本的总结，以mermaid flowchart的形式展示：
+                    ```mermaid
+                    flowchart LR
+                        A["节点名1"] --> B("节点名2")
+                        B --> C{"节点名3"}
+                        C --> D["节点名4"]
+                        C --> |"箭头名1"| E["节点名5"]
+                        C --> |"箭头名2"| F["节点名6"]
+                    ```
+
+                    注意：
+                    （1）使用中文
+                    （2）节点名字使用引号包裹，如["Laptop"]
+                    （3）`|` 和 `"`之间不要存在空格
+                    （4）根据情况选择flowchart LR（从左到右）或者flowchart TD（从上到下）
+                '''),
         },
+
+
         "查找语法错误": {
-            "Prefix":   r"Can you help me ensure that the grammar and the spelling is correct? " +
-                        r"Do not try to polish the text, if no mistake is found, tell me that this paragraph is good." +
-                        r"If you find grammar or spelling mistakes, please list mistakes you find in a two-column markdown table, " +
-                        r"put the original text the first column, " +
-                        r"put the corrected text in the second column and highlight the key words you fixed.""\n"
+            "Prefix":   r"Help me ensure that the grammar and the spelling is correct. "
+                        r"Do not try to polish the text, if no mistake is found, tell me that this paragraph is good. "
+                        r"If you find grammar or spelling mistakes, please list mistakes you find in a two-column markdown table, "
+                        r"put the original text the first column, "
+                        r"put the corrected text in the second column and highlight the key words you fixed. "
+                        r"Finally, please provide the proofreaded text.""\n\n"
                         r"Example:""\n"
                         r"Paragraph: How is you? Do you knows what is it?""\n"
                         r"| Original sentence | Corrected sentence |""\n"
                         r"| :--- | :--- |""\n"
                         r"| How **is** you? | How **are** you? |""\n"
-                        r"| Do you **knows** what **is** **it**? | Do you **know** what **it** **is** ? |""\n"
+                        r"| Do you **knows** what **is** **it**? | Do you **know** what **it** **is** ? |""\n\n"
                         r"Below is a paragraph from an academic paper. "
                         r"You need to report all grammar and spelling mistakes as the example before."
                         + "\n\n",
             "Suffix":   r"",
             "PreProcess": clear_line_break,    # 预处理：清除换行符
         },
+
+
         "中译英": {
             "Prefix":   r"Please translate following sentence to English:" + "\n\n",
             "Suffix":   r"",
         },
-        "学术中英互译": {
-            "Prefix":   r"I want you to act as a scientific English-Chinese translator, " +
-                        r"I will provide you with some paragraphs in one language " +
-                        r"and your task is to accurately and academically translate the paragraphs only into the other language. " +
-                        r"Do not repeat the original provided paragraphs after translation. " +
-                        r"You should use artificial intelligence tools, " +
-                        r"such as natural language processing, and rhetorical knowledge " +
-                        r"and experience about effective writing techniques to reply. " +
-                        r"I'll give you my paragraphs as follows, tell me what language it is written in, and then translate:" + "\n\n",
-            "Suffix": "",
-            "Color": "secondary",
+
+
+        "学术英中互译": {
+            "Prefix":   build_gpt_academic_masked_string_langbased(
+                            text_show_chinese=
+                                r"I want you to act as a scientific English-Chinese translator, "
+                                r"I will provide you with some paragraphs in one language "
+                                r"and your task is to accurately and academically translate the paragraphs only into the other language. "
+                                r"Do not repeat the original provided paragraphs after translation. "
+                                r"You should use artificial intelligence tools, "
+                                r"such as natural language processing, and rhetorical knowledge "
+                                r"and experience about effective writing techniques to reply. "
+                                r"I'll give you my paragraphs as follows, tell me what language it is written in, and then translate:",
+                            text_show_english=
+                                r"你是经验丰富的翻译，请把以下学术文章段落翻译成中文，"
+                                r"并同时充分考虑中文的语法、清晰、简洁和整体可读性，"
+                                r"必要时，你可以修改整个句子的顺序以确保翻译后的段落符合中文的语言习惯。"
+                                r"你需要翻译的文本如下："
+                        ) + "\n\n",
+            "Suffix":   r"",
         },
+
+
         "英译中": {
             "Prefix":   r"翻译成地道的中文：" + "\n\n",
             "Suffix":   r"",
+            "Visible":  False,
         },
+
+
         "找图片": {
-            "Prefix":   r"我需要你找一张网络图片。使用Unsplash API(https://source.unsplash.com/960x640/?<英语关键词>)获取图片URL，" +
+            "Prefix":   r"我需要你找一张网络图片。使用Unsplash API(https://source.unsplash.com/960x640/?<英语关键词>)获取图片URL，"
                         r"然后请使用Markdown格式封装，并且不要有反斜线，不要用代码块。现在，请按以下描述给我发送图片：" + "\n\n",
             "Suffix":   r"",
             "Visible":  False,
         },
+
+
         "解释代码": {
             "Prefix":   r"请解释以下代码：" + "\n```\n",
             "Suffix":   "\n```\n",
         },
+
+
         "参考文献转Bib": {
-            "Prefix":   r"Here are some bibliography items, please transform them into bibtex style." +
-                        r"Note that, reference styles maybe more than one kind, you should transform each item correctly." +
-                        r"Items need to be transformed:",
+            "Prefix":   r"Here are some bibliography items, please transform them into bibtex style."
+                        r"Note that, reference styles maybe more than one kind, you should transform each item correctly."
+                        r"Items need to be transformed:" + "\n\n",
+            "Visible":  False,
             "Suffix":   r"",
         }
     }
@@ -87,8 +151,25 @@ def handle_core_functionality(additional_fn, inputs, history, chatbot):
     import core_functional
     importlib.reload(core_functional)    # 热更新prompt
     core_functional = core_functional.get_core_functions()
-    if "PreProcess" in core_functional[additional_fn]: inputs = core_functional[additional_fn]["PreProcess"](inputs)  # 获取预处理函数（如果有的话）
-    inputs = core_functional[additional_fn]["Prefix"] + inputs + core_functional[additional_fn]["Suffix"]
+    addition = chatbot._cookies['customize_fn_overwrite']
+    if additional_fn in addition:
+        # 自定义功能
+        inputs = addition[additional_fn]["Prefix"] + inputs + addition[additional_fn]["Suffix"]
+        return inputs, history
+    else:
+        # 预制功能
+        if "PreProcess" in core_functional[additional_fn]:
+            if core_functional[additional_fn]["PreProcess"] is not None:
+                inputs = core_functional[additional_fn]["PreProcess"](inputs)  # 获取预处理函数（如果有的话）
+        # 为字符串加上上面定义的前缀和后缀。
+        inputs = apply_gpt_academic_string_mask_langbased(
+            string = core_functional[additional_fn]["Prefix"] + inputs + core_functional[additional_fn]["Suffix"],
+            lang_reference = inputs,
+        )
         if core_functional[additional_fn].get("AutoClearHistory", False):
             history = []
         return inputs, history
+
+if __name__ == "__main__":
+    t = get_core_functions()["总结绘制脑图"]
+    print(t["Prefix"] + t["Suffix"])

crazy_functions/Conversation_To_File.py

@@ -0,0 +1,220 @@
+from toolbox import CatchException, update_ui, promote_file_to_downloadzone, get_log_folder, get_user
+from crazy_functions.plugin_template.plugin_class_template import GptAcademicPluginTemplate, ArgProperty
+import re
+
+f_prefix = 'GPT-Academic对话存档'
+
+def write_chat_to_file(chatbot, history=None, file_name=None):
+    """
+    将对话记录history以Markdown格式写入文件中。如果没有指定文件名，则使用当前时间生成文件名。
+    """
+    import os
+    import time
+    from themes.theme import advanced_css
+
+    if file_name is None:
+        file_name = f_prefix + time.strftime("%Y-%m-%d-%H-%M-%S", time.localtime()) + '.html'
+    fp = os.path.join(get_log_folder(get_user(chatbot), plugin_name='chat_history'), file_name)
+
+    with open(fp, 'w', encoding='utf8') as f:
+        from textwrap import dedent
+        form = dedent("""
+        <!DOCTYPE html><head><meta charset="utf-8"><title>对话存档</title><style>{CSS}</style></head>
+        <body>
+        <div class="test_temp1" style="width:10%; height: 500px; float:left;"></div>
+        <div class="test_temp2" style="width:80%;padding: 40px;float:left;padding-left: 20px;padding-right: 20px;box-shadow: rgba(0, 0, 0, 0.2) 0px 0px 8px 8px;border-radius: 10px;">
+            <div class="chat-body" style="display: flex;justify-content: center;flex-direction: column;align-items: center;flex-wrap: nowrap;">
+                {CHAT_PREVIEW}
+                <div></div>
+                <div></div>
+                <div style="text-align: center;width:80%;padding: 0px;float:left;padding-left:20px;padding-right:20px;box-shadow: rgba(0, 0, 0, 0.05) 0px 0px 1px 2px;border-radius: 1px;">对话（原始数据）</div>
+                {HISTORY_PREVIEW}
+            </div>
+        </div>
+        <div class="test_temp3" style="width:10%; height: 500px; float:left;"></div>
+        </body>
+        """)
+
+        qa_from = dedent("""
+        <div class="QaBox" style="width:80%;padding: 20px;margin-bottom: 20px;box-shadow: rgb(0 255 159 / 50%) 0px 0px 1px 2px;border-radius: 4px;">
+            <div class="Question" style="border-radius: 2px;">{QUESTION}</div>
+            <hr color="blue" style="border-top: dotted 2px #ccc;">
+            <div class="Answer" style="border-radius: 2px;">{ANSWER}</div>
+        </div>
+        """)
+
+        history_from = dedent("""
+        <div class="historyBox" style="width:80%;padding: 0px;float:left;padding-left:20px;padding-right:20px;box-shadow: rgba(0, 0, 0, 0.05) 0px 0px 1px 2px;border-radius: 1px;">
+            <div class="entry" style="border-radius: 2px;">{ENTRY}</div>
+        </div>
+        """)
+        CHAT_PREVIEW_BUF = ""
+        for i, contents in enumerate(chatbot):
+            question, answer = contents[0], contents[1]
+            if question is None: question = ""
+            try: question = str(question)
+            except: question = ""
+            if answer is None: answer = ""
+            try: answer = str(answer)
+            except: answer = ""
+            CHAT_PREVIEW_BUF += qa_from.format(QUESTION=question, ANSWER=answer)
+
+        HISTORY_PREVIEW_BUF = ""
+        for h in history:
+            HISTORY_PREVIEW_BUF += history_from.format(ENTRY=h)
+        html_content = form.format(CHAT_PREVIEW=CHAT_PREVIEW_BUF, HISTORY_PREVIEW=HISTORY_PREVIEW_BUF, CSS=advanced_css)
+        f.write(html_content)
+
+    promote_file_to_downloadzone(fp, rename_file=file_name, chatbot=chatbot)
+    return '对话历史写入：' + fp
+
+def gen_file_preview(file_name):
+    try:
+        with open(file_name, 'r', encoding='utf8') as f:
+            file_content = f.read()
+        # pattern to match the text between <head> and </head>
+        pattern = re.compile(r'<head>.*?</head>', flags=re.DOTALL)
+        file_content = re.sub(pattern, '', file_content)
+        html, history = file_content.split('<hr color="blue"> \n\n 对话数据 (无渲染):\n')
+        history = history.strip('<code>')
+        history = history.strip('</code>')
+        history = history.split("\n>>>")
+        return list(filter(lambda x:x!="", history))[0][:100]
+    except:
+        return ""
+
+def read_file_to_chat(chatbot, history, file_name):
+    with open(file_name, 'r', encoding='utf8') as f:
+        file_content = f.read()
+    from bs4 import BeautifulSoup
+    soup = BeautifulSoup(file_content, 'lxml')
+    # 提取QaBox信息
+    chatbot.clear()
+    qa_box_list = []
+    qa_boxes = soup.find_all("div", class_="QaBox")
+    for box in qa_boxes:
+        question = box.find("div", class_="Question").get_text(strip=False)
+        answer = box.find("div", class_="Answer").get_text(strip=False)
+        qa_box_list.append({"Question": question, "Answer": answer})
+        chatbot.append([question, answer])
+    # 提取historyBox信息
+    history_box_list = []
+    history_boxes = soup.find_all("div", class_="historyBox")
+    for box in history_boxes:
+        entry = box.find("div", class_="entry").get_text(strip=False)
+        history_box_list.append(entry)
+    history = history_box_list
+    chatbot.append([None, f"[Local Message] 载入对话{len(qa_box_list)}条，上下文{len(history)}条。"])
+    return chatbot, history
+
+@CatchException
+def 对话历史存档(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+    """
+    txt             输入栏用户输入的文本，例如需要翻译的一段话，再例如一个包含了待处理文件的路径
+    llm_kwargs      gpt模型参数，如温度和top_p等，一般原样传递下去就行
+    plugin_kwargs   插件模型的参数，暂时没有用武之地
+    chatbot         聊天显示框的句柄，用于显示给用户
+    history         聊天历史，前情提要
+    system_prompt   给gpt的静默提醒
+    user_request    当前用户的请求信息（IP地址等）
+    """
+    file_name = plugin_kwargs.get("file_name", None)
+    if (file_name is not None) and (file_name != "") and (not file_name.endswith('.html')): file_name += '.html'
+    else: file_name = None
+
+    chatbot.append((None, f"[Local Message] {write_chat_to_file(chatbot, history, file_name)}，您可以调用下拉菜单中的“载入对话历史存档”还原当下的对话。"))
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面 # 由于请求gpt需要一段时间，我们先及时地做一次界面更新
+
+
+class Conversation_To_File_Wrap(GptAcademicPluginTemplate):
+    def __init__(self):
+        """
+        请注意`execute`会执行在不同的线程中，因此您在定义和使用类变量时，应当慎之又慎！
+        """
+        pass
+
+    def define_arg_selection_menu(self):
+        """
+        定义插件的二级选项菜单
+
+        第一个参数，名称`file_name`，参数`type`声明这是一个文本框，文本框上方显示`title`，文本框内部显示`description`，`default_value`为默认值；
+        """
+        gui_definition = {
+            "file_name": ArgProperty(title="保存文件名", description="输入对话存档文件名，留空则使用时间作为文件名", default_value="", type="string").model_dump_json(), # 主输入，自动从输入框同步
+        }
+        return gui_definition
+
+    def execute(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+        """
+        执行插件
+        """
+        yield from 对话历史存档(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request)
+
+
+
+
+
+def hide_cwd(str):
+    import os
+    current_path = os.getcwd()
+    replace_path = "."
+    return str.replace(current_path, replace_path)
+
+@CatchException
+def 载入对话历史存档(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+    """
+    txt             输入栏用户输入的文本，例如需要翻译的一段话，再例如一个包含了待处理文件的路径
+    llm_kwargs      gpt模型参数，如温度和top_p等，一般原样传递下去就行
+    plugin_kwargs   插件模型的参数，暂时没有用武之地
+    chatbot         聊天显示框的句柄，用于显示给用户
+    history         聊天历史，前情提要
+    system_prompt   给gpt的静默提醒
+    user_request    当前用户的请求信息（IP地址等）
+    """
+    from crazy_functions.crazy_utils import get_files_from_everything
+    success, file_manifest, _ = get_files_from_everything(txt, type='.html')
+
+    if not success:
+        if txt == "": txt = '空空如也的输入栏'
+        import glob
+        local_history = "<br/>".join([
+            "`"+hide_cwd(f)+f" ({gen_file_preview(f)})"+"`"
+            for f in glob.glob(
+                f'{get_log_folder(get_user(chatbot), plugin_name="chat_history")}/**/{f_prefix}*.html',
+                recursive=True
+            )])
+        chatbot.append([f"正在查找对话历史文件（html格式）: {txt}", f"找不到任何html文件: {txt}。但本地存储了以下历史文件，您可以将任意一个文件路径粘贴到输入区，然后重试：<br/>{local_history}"])
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        return
+
+    try:
+        chatbot, history = read_file_to_chat(chatbot, history, file_manifest[0])
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+    except:
+        chatbot.append([f"载入对话历史文件", f"对话历史文件损坏！"])
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        return
+
+@CatchException
+def 删除所有本地对话历史记录(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+    """
+    txt             输入栏用户输入的文本，例如需要翻译的一段话，再例如一个包含了待处理文件的路径
+    llm_kwargs      gpt模型参数，如温度和top_p等，一般原样传递下去就行
+    plugin_kwargs   插件模型的参数，暂时没有用武之地
+    chatbot         聊天显示框的句柄，用于显示给用户
+    history         聊天历史，前情提要
+    system_prompt   给gpt的静默提醒
+    user_request    当前用户的请求信息（IP地址等）
+    """
+
+    import glob, os
+    local_history = "<br/>".join([
+        "`"+hide_cwd(f)+"`"
+        for f in glob.glob(
+            f'{get_log_folder(get_user(chatbot), plugin_name="chat_history")}/**/{f_prefix}*.html', recursive=True
+        )])
+    for f in glob.glob(f'{get_log_folder(get_user(chatbot), plugin_name="chat_history")}/**/{f_prefix}*.html', recursive=True):
+        os.remove(f)
+    chatbot.append([f"删除所有历史对话文件", f"已删除<br/>{local_history}"])
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+    return

crazy_functions/Image_Generate.py

@@ -0,0 +1,276 @@
+from toolbox import CatchException, update_ui, get_conf, select_api_key, get_log_folder
+from crazy_functions.multi_stage.multi_stage_utils import GptAcademicState
+
+
+def gen_image(llm_kwargs, prompt, resolution="1024x1024", model="dall-e-2", quality=None, style=None):
+    import requests, json, time, os
+    from request_llms.bridge_all import model_info
+
+    proxies = get_conf('proxies')
+    # Set up OpenAI API key and model
+    api_key = select_api_key(llm_kwargs['api_key'], llm_kwargs['llm_model'])
+    chat_endpoint = model_info[llm_kwargs['llm_model']]['endpoint']
+    # 'https://api.openai.com/v1/chat/completions'
+    img_endpoint = chat_endpoint.replace('chat/completions','images/generations')
+    # # Generate the image
+    url = img_endpoint
+    headers = {
+        'Authorization': f"Bearer {api_key}",
+        'Content-Type': 'application/json'
+    }
+    data = {
+        'prompt': prompt,
+        'n': 1,
+        'size': resolution,
+        'model': model,
+        'response_format': 'url'
+    }
+    if quality is not None:
+        data['quality'] = quality
+    if style is not None:
+        data['style'] = style
+    response = requests.post(url, headers=headers, json=data, proxies=proxies)
+    # logger.info(response.content)
+    try:
+        image_url = json.loads(response.content.decode('utf8'))['data'][0]['url']
+    except:
+        raise RuntimeError(response.content.decode())
+    # 文件保存到本地
+    r = requests.get(image_url, proxies=proxies)
+    file_path = f'{get_log_folder()}/image_gen/'
+    os.makedirs(file_path, exist_ok=True)
+    file_name = 'Image' + time.strftime("%Y-%m-%d-%H-%M-%S", time.localtime()) + '.png'
+    with open(file_path+file_name, 'wb+') as f: f.write(r.content)
+
+
+    return image_url, file_path+file_name
+
+
+def edit_image(llm_kwargs, prompt, image_path, resolution="1024x1024", model="dall-e-2"):
+    import requests, json, time, os
+    from request_llms.bridge_all import model_info
+
+    proxies = get_conf('proxies')
+    api_key = select_api_key(llm_kwargs['api_key'], llm_kwargs['llm_model'])
+    chat_endpoint = model_info[llm_kwargs['llm_model']]['endpoint']
+    # 'https://api.openai.com/v1/chat/completions'
+    img_endpoint = chat_endpoint.replace('chat/completions','images/edits')
+    # # Generate the image
+    url = img_endpoint
+    n = 1
+    headers = {
+        'Authorization': f"Bearer {api_key}",
+    }
+    make_transparent(image_path, image_path+'.tsp.png')
+    make_square_image(image_path+'.tsp.png', image_path+'.tspsq.png')
+    resize_image(image_path+'.tspsq.png', image_path+'.ready.png', max_size=1024)
+    image_path = image_path+'.ready.png'
+    with open(image_path, 'rb') as f:
+        file_content = f.read()
+        files = {
+            'image': (os.path.basename(image_path), file_content),
+            # 'mask': ('mask.png', open('mask.png', 'rb'))
+            'prompt':   (None, prompt),
+            "n":        (None, str(n)),
+            'size':     (None, resolution),
+        }
+
+    response = requests.post(url, headers=headers, files=files, proxies=proxies)
+    # logger.info(response.content)
+    try:
+        image_url = json.loads(response.content.decode('utf8'))['data'][0]['url']
+    except:
+        raise RuntimeError(response.content.decode())
+    # 文件保存到本地
+    r = requests.get(image_url, proxies=proxies)
+    file_path = f'{get_log_folder()}/image_gen/'
+    os.makedirs(file_path, exist_ok=True)
+    file_name = 'Image' + time.strftime("%Y-%m-%d-%H-%M-%S", time.localtime()) + '.png'
+    with open(file_path+file_name, 'wb+') as f: f.write(r.content)
+
+
+    return image_url, file_path+file_name
+
+
+@CatchException
+def 图片生成_DALLE2(prompt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+    """
+    txt             输入栏用户输入的文本,例如需要翻译的一段话,再例如一个包含了待处理文件的路径
+    llm_kwargs      gpt模型参数,如温度和top_p等,一般原样传递下去就行
+    plugin_kwargs   插件模型的参数,暂时没有用武之地
+    chatbot         聊天显示框的句柄,用于显示给用户
+    history         聊天历史,前情提要
+    system_prompt   给gpt的静默提醒
+    user_request    当前用户的请求信息（IP地址等）
+    """
+    history = []    # 清空历史,以免输入溢出
+    if prompt.strip() == "":
+        chatbot.append((prompt, "[Local Message] 图像生成提示为空白，请在“输入区”输入图像生成提示。"))
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面 界面更新
+        return
+    chatbot.append(("您正在调用“图像生成”插件。", "[Local Message] 生成图像, 使用前请切换模型到GPT系列。如果中文Prompt效果不理想, 请尝试英文Prompt。正在处理中 ....."))
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面 由于请求gpt需要一段时间,我们先及时地做一次界面更新
+    if ("advanced_arg" in plugin_kwargs) and (plugin_kwargs["advanced_arg"] == ""): plugin_kwargs.pop("advanced_arg")
+    resolution = plugin_kwargs.get("advanced_arg", '1024x1024')
+    image_url, image_path = gen_image(llm_kwargs, prompt, resolution)
+    chatbot.append([prompt,
+        f'图像中转网址: <br/>`{image_url}`<br/>'+
+        f'中转网址预览: <br/><div align="center"><img src="{image_url}"></div>'
+        f'本地文件地址: <br/>`{image_path}`<br/>'+
+        f'本地文件预览: <br/><div align="center"><img src="file={image_path}"></div>'
+    ])
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面 界面更新
+
+
+@CatchException
+def 图片生成_DALLE3(prompt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+    history = []    # 清空历史,以免输入溢出
+    if prompt.strip() == "":
+        chatbot.append((prompt, "[Local Message] 图像生成提示为空白，请在“输入区”输入图像生成提示。"))
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面 界面更新
+        return
+    chatbot.append(("您正在调用“图像生成”插件。", "[Local Message] 生成图像, 使用前请切换模型到GPT系列。如果中文Prompt效果不理想, 请尝试英文Prompt。正在处理中 ....."))
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面 由于请求gpt需要一段时间,我们先及时地做一次界面更新
+    if ("advanced_arg" in plugin_kwargs) and (plugin_kwargs["advanced_arg"] == ""): plugin_kwargs.pop("advanced_arg")
+    resolution_arg = plugin_kwargs.get("advanced_arg", '1024x1024-standard-vivid').lower()
+    parts = resolution_arg.split('-')
+    resolution = parts[0] # 解析分辨率
+    quality = 'standard' # 质量与风格默认值
+    style = 'vivid'
+    # 遍历检查是否有额外参数
+    for part in parts[1:]:
+        if part in ['hd', 'standard']:
+            quality = part
+        elif part in ['vivid', 'natural']:
+            style = part
+    image_url, image_path = gen_image(llm_kwargs, prompt, resolution, model="dall-e-3", quality=quality, style=style)
+    chatbot.append([prompt,
+        f'图像中转网址: <br/>`{image_url}`<br/>'+
+        f'中转网址预览: <br/><div align="center"><img src="{image_url}"></div>'
+        f'本地文件地址: <br/>`{image_path}`<br/>'+
+        f'本地文件预览: <br/><div align="center"><img src="file={image_path}"></div>'
+    ])
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面 界面更新
+
+
+class ImageEditState(GptAcademicState):
+    # 尚未完成
+    def get_image_file(self, x):
+        import os, glob
+        if len(x) == 0:             return False, None
+        if not os.path.exists(x):   return False, None
+        if x.endswith('.png'):      return True, x
+        file_manifest = [f for f in glob.glob(f'{x}/**/*.png', recursive=True)]
+        confirm = (len(file_manifest) >= 1 and file_manifest[0].endswith('.png') and os.path.exists(file_manifest[0]))
+        file = None if not confirm else file_manifest[0]
+        return confirm, file
+
+    def lock_plugin(self, chatbot):
+        chatbot._cookies['lock_plugin'] = 'crazy_functions.Image_Generate->图片修改_DALLE2'
+        self.dump_state(chatbot)
+
+    def unlock_plugin(self, chatbot):
+        self.reset()
+        chatbot._cookies['lock_plugin'] = None
+        self.dump_state(chatbot)
+
+    def get_resolution(self, x):
+        return (x in ['256x256', '512x512', '1024x1024']), x
+
+    def get_prompt(self, x):
+        confirm = (len(x)>=5) and (not self.get_resolution(x)[0]) and (not self.get_image_file(x)[0])
+        return confirm, x
+
+    def reset(self):
+        self.req = [
+            {'value':None, 'description': '请先上传图像（必须是.png格式）, 然后再次点击本插件',                      'verify_fn': self.get_image_file},
+            {'value':None, 'description': '请输入分辨率,可选：256x256, 512x512 或 1024x1024, 然后再次点击本插件',   'verify_fn': self.get_resolution},
+            {'value':None, 'description': '请输入修改需求,建议您使用英文提示词, 然后再次点击本插件',                 'verify_fn': self.get_prompt},
+        ]
+        self.info = ""
+
+    def feed(self, prompt, chatbot):
+        for r in self.req:
+            if r['value'] is None:
+                confirm, res = r['verify_fn'](prompt)
+                if confirm:
+                    r['value'] = res
+                    self.dump_state(chatbot)
+                    break
+        return self
+
+    def next_req(self):
+        for r in self.req:
+            if r['value'] is None:
+                return r['description']
+        return "已经收集到所有信息"
+
+    def already_obtained_all_materials(self):
+        return all([x['value'] is not None for x in self.req])
+
+@CatchException
+def 图片修改_DALLE2(prompt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+    # 尚未完成
+    history = []    # 清空历史
+    state = ImageEditState.get_state(chatbot, ImageEditState)
+    state = state.feed(prompt, chatbot)
+    state.lock_plugin(chatbot)
+    if not state.already_obtained_all_materials():
+        chatbot.append(["图片修改\n\n1. 上传图片（图片中需要修改的位置用橡皮擦擦除为纯白色，即RGB=255,255,255）\n2. 输入分辨率 \n3. 输入修改需求", state.next_req()])
+        yield from update_ui(chatbot=chatbot, history=history)
+        return
+
+    image_path = state.req[0]['value']
+    resolution = state.req[1]['value']
+    prompt = state.req[2]['value']
+    chatbot.append(["图片修改, 执行中", f"图片:`{image_path}`<br/>分辨率:`{resolution}`<br/>修改需求:`{prompt}`"])
+    yield from update_ui(chatbot=chatbot, history=history)
+    image_url, image_path = edit_image(llm_kwargs, prompt, image_path, resolution)
+    chatbot.append([prompt,
+        f'图像中转网址: <br/>`{image_url}`<br/>'+
+        f'中转网址预览: <br/><div align="center"><img src="{image_url}"></div>'
+        f'本地文件地址: <br/>`{image_path}`<br/>'+
+        f'本地文件预览: <br/><div align="center"><img src="file={image_path}"></div>'
+    ])
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面 界面更新
+    state.unlock_plugin(chatbot)
+
+def make_transparent(input_image_path, output_image_path):
+    from PIL import Image
+    image = Image.open(input_image_path)
+    image = image.convert("RGBA")
+    data = image.getdata()
+    new_data = []
+    for item in data:
+        if item[0] == 255 and item[1] == 255 and item[2] == 255:
+            new_data.append((255, 255, 255, 0))
+        else:
+            new_data.append(item)
+    image.putdata(new_data)
+    image.save(output_image_path, "PNG")
+
+def resize_image(input_path, output_path, max_size=1024):
+    from PIL import Image
+    with Image.open(input_path) as img:
+        width, height = img.size
+        if width > max_size or height > max_size:
+            if width >= height:
+                new_width = max_size
+                new_height = int((max_size / width) * height)
+            else:
+                new_height = max_size
+                new_width = int((max_size / height) * width)
+
+            resized_img = img.resize(size=(new_width, new_height))
+            resized_img.save(output_path)
+        else:
+            img.save(output_path)
+
+def make_square_image(input_path, output_path):
+    from PIL import Image
+    with Image.open(input_path) as img:
+        width, height = img.size
+        size = max(width, height)
+        new_img = Image.new("RGBA", (size, size), color="black")
+        new_img.paste(img, ((size - width) // 2, (size - height) // 2))
+        new_img.save(output_path)

crazy_functions/Image_Generate_Wrap.py

@@ -0,0 +1,56 @@
+
+from toolbox import get_conf, update_ui
+from crazy_functions.Image_Generate import 图片生成_DALLE2, 图片生成_DALLE3, 图片修改_DALLE2
+from crazy_functions.plugin_template.plugin_class_template import GptAcademicPluginTemplate, ArgProperty
+
+
+class ImageGen_Wrap(GptAcademicPluginTemplate):
+    def __init__(self):
+        """
+        请注意`execute`会执行在不同的线程中，因此您在定义和使用类变量时，应当慎之又慎！
+        """
+        pass
+
+    def define_arg_selection_menu(self):
+        """
+        定义插件的二级选项菜单
+
+        第一个参数，名称`main_input`，参数`type`声明这是一个文本框，文本框上方显示`title`，文本框内部显示`description`，`default_value`为默认值；
+        第二个参数，名称`advanced_arg`，参数`type`声明这是一个文本框，文本框上方显示`title`，文本框内部显示`description`，`default_value`为默认值；
+
+        """
+        gui_definition = {
+            "main_input":
+                ArgProperty(title="输入图片描述", description="需要生成图像的文本描述，尽量使用英文", default_value="", type="string").model_dump_json(), # 主输入，自动从输入框同步
+            "model_name":
+                ArgProperty(title="模型", options=["DALLE2", "DALLE3"], default_value="DALLE3", description="无", type="dropdown").model_dump_json(),
+            "resolution":
+                ArgProperty(title="分辨率", options=["256x256(限DALLE2)", "512x512(限DALLE2)", "1024x1024", "1792x1024(限DALLE3)", "1024x1792(限DALLE3)"], default_value="1024x1024", description="无", type="dropdown").model_dump_json(),
+            "quality (仅DALLE3生效)":
+                ArgProperty(title="质量", options=["standard", "hd"], default_value="standard", description="无", type="dropdown").model_dump_json(),
+            "style (仅DALLE3生效)":
+                ArgProperty(title="风格", options=["vivid", "natural"], default_value="vivid", description="无", type="dropdown").model_dump_json(),
+
+        }
+        return gui_definition
+
+    def execute(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+        """
+        执行插件
+        """
+        # 分辨率
+        resolution = plugin_kwargs["resolution"].replace("(限DALLE2)", "").replace("(限DALLE3)", "")
+
+        if plugin_kwargs["model_name"] == "DALLE2":
+            plugin_kwargs["advanced_arg"] = resolution
+            yield from 图片生成_DALLE2(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request)
+
+        elif plugin_kwargs["model_name"] == "DALLE3":
+            quality = plugin_kwargs["quality (仅DALLE3生效)"]
+            style = plugin_kwargs["style (仅DALLE3生效)"]
+            plugin_kwargs["advanced_arg"] = f"{resolution}-{quality}-{style}"
+            yield from 图片生成_DALLE3(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request)
+
+        else:
+            chatbot.append([None, "抱歉，找不到该模型"])
+            yield from update_ui(chatbot=chatbot, history=history) # 刷新界面

crazy_functions/Internet_GPT.py

@@ -0,0 +1,278 @@
+import requests
+import random
+import time
+import re
+import json
+from bs4 import BeautifulSoup
+from functools import lru_cache
+from itertools import zip_longest
+from check_proxy import check_proxy
+from toolbox import CatchException, update_ui, get_conf
+from crazy_functions.crazy_utils import request_gpt_model_in_new_thread_with_ui_alive, input_clipping
+from request_llms.bridge_all import model_info
+from request_llms.bridge_all import predict_no_ui_long_connection
+from crazy_functions.prompts.internet import SearchOptimizerPrompt, SearchAcademicOptimizerPrompt
+
+def search_optimizer(
+    query,
+    proxies,
+    history,
+    llm_kwargs,
+    optimizer=1,
+    categories="general",
+    searxng_url=None,
+    engines=None,
+):
+    # ------------- < 第1步：尝试进行搜索优化 > -------------
+    # * 增强优化，会尝试结合历史记录进行搜索优化
+    if optimizer == 2:
+        his = " "
+        if len(history) == 0:
+            pass
+        else:
+            for i, h in enumerate(history):
+                if i % 2 == 0:
+                    his += f"Q: {h}\n"
+                else:
+                    his += f"A: {h}\n"
+        if categories == "general":
+            sys_prompt = SearchOptimizerPrompt.format(query=query, history=his, num=4)
+        elif categories == "science":
+            sys_prompt = SearchAcademicOptimizerPrompt.format(query=query, history=his, num=4)
+    else:
+        his = " "
+        if categories == "general":
+            sys_prompt = SearchOptimizerPrompt.format(query=query, history=his, num=3)
+        elif categories == "science":
+            sys_prompt = SearchAcademicOptimizerPrompt.format(query=query, history=his, num=3)
+    
+    mutable = ["", time.time(), ""]
+    llm_kwargs["temperature"] = 0.8
+    try:
+        querys_json = predict_no_ui_long_connection(
+            inputs=query,
+            llm_kwargs=llm_kwargs,
+            history=[],
+            sys_prompt=sys_prompt,
+            observe_window=mutable,
+        )
+    except Exception:
+        querys_json = "1234"
+    #* 尝试解码优化后的搜索结果
+    querys_json = re.sub(r"```json|```", "", querys_json)
+    try:
+        querys = json.loads(querys_json)
+    except Exception:
+        #* 如果解码失败,降低温度再试一次
+        try:
+            llm_kwargs["temperature"] = 0.4
+            querys_json = predict_no_ui_long_connection(
+                inputs=query,
+                llm_kwargs=llm_kwargs,
+                history=[],
+                sys_prompt=sys_prompt,
+                observe_window=mutable,
+            )
+            querys_json = re.sub(r"```json|```", "", querys_json)
+            querys = json.loads(querys_json)
+        except Exception:
+            #* 如果再次失败，直接返回原始问题
+            querys = [query]
+    links = []
+    success = 0
+    Exceptions = ""
+    for q in querys:
+        try:
+            link = searxng_request(q, proxies, categories, searxng_url, engines=engines)
+            if len(link) > 0:
+                links.append(link[:-5])
+                success += 1
+        except Exception:
+            Exceptions = Exception
+            pass
+    if success == 0:
+        raise ValueError(f"在线搜索失败！\n{Exceptions}")
+    # * 清洗搜索结果，依次放入每组第一，第二个搜索结果，并清洗重复的搜索结果
+    seen_links = set()
+    result = []
+    for tuple in zip_longest(*links, fillvalue=None):
+        for item in tuple:
+            if item is not None:
+                link = item["link"]
+                if link not in seen_links:
+                    seen_links.add(link)
+                    result.append(item)
+    return result
+
+
+@lru_cache
+def get_auth_ip():
+    ip = check_proxy(None, return_ip=True)
+    if ip is None:
+        return '114.114.114.' + str(random.randint(1, 10))
+    return ip
+
+
+def searxng_request(query, proxies, categories='general', searxng_url=None, engines=None):
+    if searxng_url is None:
+        url = get_conf("SEARXNG_URL")
+    else:
+        url = searxng_url
+
+    if engines == "Mixed":
+        engines = None
+
+    if categories == 'general':
+        params = {
+            'q': query,         # 搜索查询
+            'format': 'json',   # 输出格式为JSON
+            'language': 'zh',   # 搜索语言
+            'engines': engines,
+        }
+    elif categories == 'science':
+        params = {
+            'q': query,         # 搜索查询
+            'format': 'json',   # 输出格式为JSON
+            'language': 'zh',   # 搜索语言
+            'categories': 'science'
+        }
+    else:
+        raise ValueError('不支持的检索类型')
+
+    headers = {
+        'Accept-Language': 'zh-CN,zh;q=0.9',
+        'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/58.0.3029.110 Safari/537.36',
+        'X-Forwarded-For': get_auth_ip(),
+        'X-Real-IP': get_auth_ip()
+    }
+    results = []
+    response = requests.post(url, params=params, headers=headers, proxies=proxies, timeout=30)
+    if response.status_code == 200:
+        json_result = response.json()
+        for result in json_result['results']:
+            item = {
+                "title": result.get("title", ""),
+                "source": result.get("engines", "unknown"),
+                "content": result.get("content", ""),
+                "link": result["url"],
+            }
+            results.append(item)
+        return results
+    else:
+        if response.status_code == 429:
+            raise ValueError("Searxng（在线搜索服务）当前使用人数太多，请稍后。")
+        else:
+            raise ValueError("在线搜索失败，状态码: " + str(response.status_code) + '\t' + response.content.decode('utf-8'))
+
+
+def scrape_text(url, proxies) -> str:
+    """Scrape text from a webpage
+
+    Args:
+        url (str): The URL to scrape text from
+
+    Returns:
+        str: The scraped text
+    """
+    headers = {
+        'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/94.0.4606.61 Safari/537.36',
+        'Content-Type': 'text/plain',
+    }
+    try:
+        response = requests.get(url, headers=headers, proxies=proxies, timeout=8)
+        if response.encoding == "ISO-8859-1": response.encoding = response.apparent_encoding
+    except:
+        return "无法连接到该网页"
+    soup = BeautifulSoup(response.text, "html.parser")
+    for script in soup(["script", "style"]):
+        script.extract()
+    text = soup.get_text()
+    lines = (line.strip() for line in text.splitlines())
+    chunks = (phrase.strip() for line in lines for phrase in line.split("  "))
+    text = "\n".join(chunk for chunk in chunks if chunk)
+    return text
+
+
+@CatchException
+def 连接网络回答问题(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+    optimizer_history = history[:-8]
+    history = []    # 清空历史，以免输入溢出
+    chatbot.append((f"请结合互联网信息回答以下问题：{txt}", "检索中..."))
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+
+    # ------------- < 第1步：爬取搜索引擎的结果 > -------------
+    from toolbox import get_conf
+    proxies = get_conf('proxies')
+    categories = plugin_kwargs.get('categories', 'general')
+    searxng_url = plugin_kwargs.get('searxng_url', None)
+    engines = plugin_kwargs.get('engine', None)
+    optimizer = plugin_kwargs.get('optimizer', "关闭")
+    if optimizer == "关闭":
+        urls = searxng_request(txt, proxies, categories, searxng_url, engines=engines)
+    else:
+        urls = search_optimizer(txt, proxies, optimizer_history, llm_kwargs, optimizer, categories, searxng_url, engines)
+    history = []
+    if len(urls) == 0:
+        chatbot.append((f"结论：{txt}",
+                        "[Local Message] 受到限制，无法从searxng获取信息！请尝试更换搜索引擎。"))
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        return
+
+    # ------------- < 第2步：依次访问网页 > -------------
+    max_search_result = 5   # 最多收纳多少个网页的结果
+    if optimizer == "开启(增强)":
+        max_search_result = 8
+    chatbot.append(["联网检索中 ...", None])
+    for index, url in enumerate(urls[:max_search_result]):
+        res = scrape_text(url['link'], proxies)
+        prefix = f"第{index}份搜索结果 [源自{url['source'][0]}搜索] （{url['title'][:25]}）："
+        history.extend([prefix, res])
+        res_squeeze = res.replace('\n', '...')
+        chatbot[-1] = [prefix + "\n\n" + res_squeeze[:500] + "......", None]
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+
+    # ------------- < 第3步：ChatGPT综合 > -------------
+    if (optimizer != "开启(增强)"):
+        i_say = f"从以上搜索结果中抽取信息，然后回答问题：{txt}"
+        i_say, history = input_clipping(    # 裁剪输入，从最长的条目开始裁剪，防止爆token
+            inputs=i_say,
+            history=history,
+            max_token_limit=min(model_info[llm_kwargs['llm_model']]['max_token']*3//4, 8192)
+        )
+        gpt_say = yield from request_gpt_model_in_new_thread_with_ui_alive(
+            inputs=i_say, inputs_show_user=i_say,
+            llm_kwargs=llm_kwargs, chatbot=chatbot, history=history,
+            sys_prompt="请从给定的若干条搜索结果中抽取信息，对最相关的两个搜索结果进行总结，然后回答问题。"
+        )
+        chatbot[-1] = (i_say, gpt_say)
+        history.append(i_say);history.append(gpt_say)
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面 # 界面更新
+
+    #* 或者使用搜索优化器，这样可以保证后续问答能读取到有效的历史记录
+    else:
+        i_say = f"从以上搜索结果中抽取与问题：{txt} 相关的信息:"
+        i_say, history = input_clipping(    # 裁剪输入，从最长的条目开始裁剪，防止爆token
+            inputs=i_say,
+            history=history,
+            max_token_limit=min(model_info[llm_kwargs['llm_model']]['max_token']*3//4, 8192)
+        )
+        gpt_say = yield from request_gpt_model_in_new_thread_with_ui_alive(
+            inputs=i_say, inputs_show_user=i_say,
+            llm_kwargs=llm_kwargs, chatbot=chatbot, history=history,
+            sys_prompt="请从给定的若干条搜索结果中抽取信息，对最相关的三个搜索结果进行总结"
+        )
+        chatbot[-1] = (i_say, gpt_say)
+        history = []
+        history.append(i_say);history.append(gpt_say)
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面 # 界面更新
+
+        # ------------- < 第4步：根据综合回答问题 > -------------
+        i_say = f"请根据以上搜索结果回答问题：{txt}"
+        gpt_say = yield from request_gpt_model_in_new_thread_with_ui_alive(
+            inputs=i_say, inputs_show_user=i_say,
+            llm_kwargs=llm_kwargs, chatbot=chatbot, history=history,
+            sys_prompt="请根据给定的若干条搜索结果回答问题"
+        )
+        chatbot[-1] = (i_say, gpt_say)
+        history.append(i_say);history.append(gpt_say)
+        yield from update_ui(chatbot=chatbot, history=history)

crazy_functions/Internet_GPT_Wrap.py

@@ -0,0 +1,45 @@
+
+from toolbox import get_conf
+from crazy_functions.Internet_GPT import 连接网络回答问题
+from crazy_functions.plugin_template.plugin_class_template import GptAcademicPluginTemplate, ArgProperty
+
+
+class NetworkGPT_Wrap(GptAcademicPluginTemplate):
+    def __init__(self):
+        """
+        请注意`execute`会执行在不同的线程中，因此您在定义和使用类变量时，应当慎之又慎！
+        """
+        pass
+
+    def define_arg_selection_menu(self):
+        """
+        定义插件的二级选项菜单
+
+        第一个参数，名称`main_input`，参数`type`声明这是一个文本框，文本框上方显示`title`，文本框内部显示`description`，`default_value`为默认值；
+        第二个参数，名称`advanced_arg`，参数`type`声明这是一个文本框，文本框上方显示`title`，文本框内部显示`description`，`default_value`为默认值；
+        第三个参数，名称`allow_cache`，参数`type`声明这是一个下拉菜单，下拉菜单上方显示`title`+`description`，下拉菜单的选项为`options`，`default_value`为下拉菜单默认值；
+
+        """
+        gui_definition = {
+            "main_input":
+                ArgProperty(title="输入问题", description="待通过互联网检索的问题，会自动读取输入框内容", default_value="", type="string").model_dump_json(), # 主输入，自动从输入框同步
+            "categories":
+                ArgProperty(title="搜索分类", options=["网页", "学术论文"], default_value="网页", description="无", type="dropdown").model_dump_json(),
+            "engine":
+                ArgProperty(title="选择搜索引擎", options=["Mixed", "bing", "google", "duckduckgo"], default_value="google", description="无", type="dropdown").model_dump_json(),
+            "optimizer":
+                ArgProperty(title="搜索优化", options=["关闭", "开启", "开启(增强)"], default_value="关闭", description="是否使用搜索增强。注意这可能会消耗较多token", type="dropdown").model_dump_json(),
+            "searxng_url":
+                ArgProperty(title="Searxng服务地址", description="输入Searxng的地址", default_value=get_conf("SEARXNG_URL"), type="string").model_dump_json(), # 主输入，自动从输入框同步
+
+        }
+        return gui_definition
+
+    def execute(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+        """
+        执行插件
+        """
+        if plugin_kwargs["categories"] == "网页": plugin_kwargs["categories"] = "general"
+        if plugin_kwargs["categories"] == "学术论文": plugin_kwargs["categories"] = "science"
+        yield from 连接网络回答问题(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request)
+

crazy_functions/Latex_Function.py

@@ -0,0 +1,550 @@
+from toolbox import update_ui, trimmed_format_exc, get_conf, get_log_folder, promote_file_to_downloadzone, check_repeat_upload, map_file_to_sha256
+from toolbox import CatchException, report_exception, update_ui_lastest_msg, zip_result, gen_time_str
+from functools import partial
+from loguru import logger
+
+import glob, os, requests, time, json, tarfile
+
+pj = os.path.join
+ARXIV_CACHE_DIR = get_conf("ARXIV_CACHE_DIR")
+
+
+# =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- 工具函数 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
+# 专业词汇声明  = 'If the term "agent" is used in this section, it should be translated to "智能体". '
+def switch_prompt(pfg, mode, more_requirement):
+    """
+    Generate prompts and system prompts based on the mode for proofreading or translating.
+    Args:
+    - pfg: Proofreader or Translator instance.
+    - mode: A string specifying the mode, either 'proofread' or 'translate_zh'.
+
+    Returns:
+    - inputs_array: A list of strings containing prompts for users to respond to.
+    - sys_prompt_array: A list of strings containing prompts for system prompts.
+    """
+    n_split = len(pfg.sp_file_contents)
+    if mode == 'proofread_en':
+        inputs_array = [r"Below is a section from an academic paper, proofread this section." +
+                        r"Do not modify any latex command such as \section, \cite, \begin, \item and equations. " + more_requirement +
+                        r"Answer me only with the revised text:" +
+                        f"\n\n{frag}" for frag in pfg.sp_file_contents]
+        sys_prompt_array = ["You are a professional academic paper writer." for _ in range(n_split)]
+    elif mode == 'translate_zh':
+        inputs_array = [
+            r"Below is a section from an English academic paper, translate it into Chinese. " + more_requirement +
+            r"Do not modify any latex command such as \section, \cite, \begin, \item and equations. " +
+            r"Answer me only with the translated text:" +
+            f"\n\n{frag}" for frag in pfg.sp_file_contents]
+        sys_prompt_array = ["You are a professional translator." for _ in range(n_split)]
+    else:
+        assert False, "未知指令"
+    return inputs_array, sys_prompt_array
+
+
+def desend_to_extracted_folder_if_exist(project_folder):
+    """
+    Descend into the extracted folder if it exists, otherwise return the original folder.
+
+    Args:
+    - project_folder: A string specifying the folder path.
+
+    Returns:
+    - A string specifying the path to the extracted folder, or the original folder if there is no extracted folder.
+    """
+    maybe_dir = [f for f in glob.glob(f'{project_folder}/*') if os.path.isdir(f)]
+    if len(maybe_dir) == 0: return project_folder
+    if maybe_dir[0].endswith('.extract'): return maybe_dir[0]
+    return project_folder
+
+
+def move_project(project_folder, arxiv_id=None):
+    """
+    Create a new work folder and copy the project folder to it.
+
+    Args:
+    - project_folder: A string specifying the folder path of the project.
+
+    Returns:
+    - A string specifying the path to the new work folder.
+    """
+    import shutil, time
+    time.sleep(2)  # avoid time string conflict
+    if arxiv_id is not None:
+        new_workfolder = pj(ARXIV_CACHE_DIR, arxiv_id, 'workfolder')
+    else:
+        new_workfolder = f'{get_log_folder()}/{gen_time_str()}'
+    try:
+        shutil.rmtree(new_workfolder)
+    except:
+        pass
+
+    # align subfolder if there is a folder wrapper
+    items = glob.glob(pj(project_folder, '*'))
+    items = [item for item in items if os.path.basename(item) != '__MACOSX']
+    if len(glob.glob(pj(project_folder, '*.tex'))) == 0 and len(items) == 1:
+        if os.path.isdir(items[0]): project_folder = items[0]
+
+    shutil.copytree(src=project_folder, dst=new_workfolder)
+    return new_workfolder
+
+
+def arxiv_download(chatbot, history, txt, allow_cache=True):
+    def check_cached_translation_pdf(arxiv_id):
+        translation_dir = pj(ARXIV_CACHE_DIR, arxiv_id, 'translation')
+        if not os.path.exists(translation_dir):
+            os.makedirs(translation_dir)
+        target_file = pj(translation_dir, 'translate_zh.pdf')
+        if os.path.exists(target_file):
+            promote_file_to_downloadzone(target_file, rename_file=None, chatbot=chatbot)
+            target_file_compare = pj(translation_dir, 'comparison.pdf')
+            if os.path.exists(target_file_compare):
+                promote_file_to_downloadzone(target_file_compare, rename_file=None, chatbot=chatbot)
+            return target_file
+        return False
+
+    def is_float(s):
+        try:
+            float(s)
+            return True
+        except ValueError:
+            return False
+
+    if txt.startswith('https://arxiv.org/pdf/'):
+        arxiv_id = txt.split('/')[-1]   # 2402.14207v2.pdf
+        txt = arxiv_id.split('v')[0]  # 2402.14207
+
+    if ('.' in txt) and ('/' not in txt) and is_float(txt):  # is arxiv ID
+        txt = 'https://arxiv.org/abs/' + txt.strip()
+    if ('.' in txt) and ('/' not in txt) and is_float(txt[:10]):  # is arxiv ID
+        txt = 'https://arxiv.org/abs/' + txt[:10]
+
+    if not txt.startswith('https://arxiv.org'):
+        return txt, None    # 是本地文件，跳过下载
+
+    # <-------------- inspect format ------------->
+    chatbot.append([f"检测到arxiv文档连接", '尝试下载 ...'])
+    yield from update_ui(chatbot=chatbot, history=history)
+    time.sleep(1)  # 刷新界面
+
+    url_ = txt  # https://arxiv.org/abs/1707.06690
+
+    if not txt.startswith('https://arxiv.org/abs/'):
+        msg = f"解析arxiv网址失败, 期望格式例如: https://arxiv.org/abs/1707.06690。实际得到格式: {url_}。"
+        yield from update_ui_lastest_msg(msg, chatbot=chatbot, history=history)  # 刷新界面
+        return msg, None
+    # <-------------- set format ------------->
+    arxiv_id = url_.split('/abs/')[-1]
+    if 'v' in arxiv_id: arxiv_id = arxiv_id[:10]
+    cached_translation_pdf = check_cached_translation_pdf(arxiv_id)
+    if cached_translation_pdf and allow_cache: return cached_translation_pdf, arxiv_id
+
+    url_tar = url_.replace('/abs/', '/e-print/')
+    translation_dir = pj(ARXIV_CACHE_DIR, arxiv_id, 'e-print')
+    extract_dst = pj(ARXIV_CACHE_DIR, arxiv_id, 'extract')
+    os.makedirs(translation_dir, exist_ok=True)
+
+    # <-------------- download arxiv source file ------------->
+    dst = pj(translation_dir, arxiv_id + '.tar')
+    if os.path.exists(dst):
+        yield from update_ui_lastest_msg("调用缓存", chatbot=chatbot, history=history)  # 刷新界面
+    else:
+        yield from update_ui_lastest_msg("开始下载", chatbot=chatbot, history=history)  # 刷新界面
+        proxies = get_conf('proxies')
+        r = requests.get(url_tar, proxies=proxies)
+        with open(dst, 'wb+') as f:
+            f.write(r.content)
+    # <-------------- extract file ------------->
+    yield from update_ui_lastest_msg("下载完成", chatbot=chatbot, history=history)  # 刷新界面
+    from toolbox import extract_archive
+    extract_archive(file_path=dst, dest_dir=extract_dst)
+    return extract_dst, arxiv_id
+
+
+def pdf2tex_project(pdf_file_path, plugin_kwargs):
+    if plugin_kwargs["method"] == "MATHPIX":
+        # Mathpix API credentials
+        app_id, app_key = get_conf('MATHPIX_APPID', 'MATHPIX_APPKEY')
+        headers = {"app_id": app_id, "app_key": app_key}
+
+        # Step 1: Send PDF file for processing
+        options = {
+            "conversion_formats": {"tex.zip": True},
+            "math_inline_delimiters": ["$", "$"],
+            "rm_spaces": True
+        }
+
+        response = requests.post(url="https://api.mathpix.com/v3/pdf",
+                                headers=headers,
+                                data={"options_json": json.dumps(options)},
+                                files={"file": open(pdf_file_path, "rb")})
+
+        if response.ok:
+            pdf_id = response.json()["pdf_id"]
+            logger.info(f"PDF processing initiated. PDF ID: {pdf_id}")
+
+            # Step 2: Check processing status
+            while True:
+                conversion_response = requests.get(f"https://api.mathpix.com/v3/pdf/{pdf_id}", headers=headers)
+                conversion_data = conversion_response.json()
+
+                if conversion_data["status"] == "completed":
+                    logger.info("PDF processing completed.")
+                    break
+                elif conversion_data["status"] == "error":
+                    logger.info("Error occurred during processing.")
+                else:
+                    logger.info(f"Processing status: {conversion_data['status']}")
+                    time.sleep(5)  # wait for a few seconds before checking again
+
+            # Step 3: Save results to local files
+            output_dir = os.path.join(os.path.dirname(pdf_file_path), 'mathpix_output')
+            if not os.path.exists(output_dir):
+                os.makedirs(output_dir)
+
+            url = f"https://api.mathpix.com/v3/pdf/{pdf_id}.tex"
+            response = requests.get(url, headers=headers)
+            file_name_wo_dot = '_'.join(os.path.basename(pdf_file_path).split('.')[:-1])
+            output_name = f"{file_name_wo_dot}.tex.zip"
+            output_path = os.path.join(output_dir, output_name)
+            with open(output_path, "wb") as output_file:
+                output_file.write(response.content)
+            logger.info(f"tex.zip file saved at: {output_path}")
+
+            import zipfile
+            unzip_dir = os.path.join(output_dir, file_name_wo_dot)
+            with zipfile.ZipFile(output_path, 'r') as zip_ref:
+                zip_ref.extractall(unzip_dir)
+
+            return unzip_dir
+
+        else:
+            logger.error(f"Error sending PDF for processing. Status code: {response.status_code}")
+            return None
+    else:
+        from crazy_functions.pdf_fns.parse_pdf_via_doc2x import 解析PDF_DOC2X_转Latex
+        unzip_dir = 解析PDF_DOC2X_转Latex(pdf_file_path)
+        return unzip_dir
+
+
+
+
+# =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= 插件主程序1 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
+
+
+@CatchException
+def Latex英文纠错加PDF对比(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+    # <-------------- information about this plugin ------------->
+    chatbot.append(["函数插件功能？",
+                    "对整个Latex项目进行纠错, 用latex编译为PDF对修正处做高亮。函数插件贡献者: Binary-Husky。注意事项: 目前对机器学习类文献转化效果最好，其他类型文献转化效果未知。仅在Windows系统进行了测试，其他操作系统表现未知。"])
+    yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+
+    # <-------------- more requirements ------------->
+    if ("advanced_arg" in plugin_kwargs) and (plugin_kwargs["advanced_arg"] == ""): plugin_kwargs.pop("advanced_arg")
+    more_req = plugin_kwargs.get("advanced_arg", "")
+    _switch_prompt_ = partial(switch_prompt, more_requirement=more_req)
+
+    # <-------------- check deps ------------->
+    try:
+        import glob, os, time, subprocess
+        subprocess.Popen(['pdflatex', '-version'])
+        from .latex_fns.latex_actions import Latex精细分解与转化, 编译Latex
+    except Exception as e:
+        chatbot.append([f"解析项目: {txt}",
+                        f"尝试执行Latex指令失败。Latex没有安装, 或者不在环境变量PATH中。安装方法https://tug.org/texlive/。报错信息\n\n```\n\n{trimmed_format_exc()}\n\n```\n\n"])
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+
+    # <-------------- clear history and read input ------------->
+    history = []
+    if os.path.exists(txt):
+        project_folder = txt
+    else:
+        if txt == "": txt = '空空如也的输入栏'
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到本地项目或无权访问: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+    file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.tex', recursive=True)]
+    if len(file_manifest) == 0:
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何.tex文件: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+
+    # <-------------- if is a zip/tar file ------------->
+    project_folder = desend_to_extracted_folder_if_exist(project_folder)
+
+    # <-------------- move latex project away from temp folder ------------->
+    from shared_utils.fastapi_server import validate_path_safety
+    validate_path_safety(project_folder, chatbot.get_user())
+    project_folder = move_project(project_folder, arxiv_id=None)
+
+    # <-------------- if merge_translate_zh is already generated, skip gpt req ------------->
+    if not os.path.exists(project_folder + '/merge_proofread_en.tex'):
+        yield from Latex精细分解与转化(file_manifest, project_folder, llm_kwargs, plugin_kwargs,
+                                       chatbot, history, system_prompt, mode='proofread_en',
+                                       switch_prompt=_switch_prompt_)
+
+    # <-------------- compile PDF ------------->
+    success = yield from 编译Latex(chatbot, history, main_file_original='merge',
+                                   main_file_modified='merge_proofread_en',
+                                   work_folder_original=project_folder, work_folder_modified=project_folder,
+                                   work_folder=project_folder)
+
+    # <-------------- zip PDF ------------->
+    zip_res = zip_result(project_folder)
+    if success:
+        chatbot.append((f"成功啦", '请查收结果（压缩包）...'))
+        yield from update_ui(chatbot=chatbot, history=history);
+        time.sleep(1)  # 刷新界面
+        promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
+    else:
+        chatbot.append((f"失败了",
+                        '虽然PDF生成失败了, 但请查收结果（压缩包）, 内含已经翻译的Tex文档, 也是可读的, 您可以到Github Issue区, 用该压缩包+Conversation_To_File进行反馈 ...'))
+        yield from update_ui(chatbot=chatbot, history=history);
+        time.sleep(1)  # 刷新界面
+        promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
+
+    # <-------------- we are done ------------->
+    return success
+
+
+# =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= 插件主程序2 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
+
+@CatchException
+def Latex翻译中文并重新编译PDF(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+    # <-------------- information about this plugin ------------->
+    chatbot.append([
+        "函数插件功能？",
+        "对整个Latex项目进行翻译, 生成中文PDF。函数插件贡献者: Binary-Husky。注意事项: 此插件Windows支持最佳，Linux下必须使用Docker安装，详见项目主README.md。目前对机器学习类文献转化效果最好，其他类型文献转化效果未知。"])
+    yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+
+    # <-------------- more requirements ------------->
+    if ("advanced_arg" in plugin_kwargs) and (plugin_kwargs["advanced_arg"] == ""): plugin_kwargs.pop("advanced_arg")
+    more_req = plugin_kwargs.get("advanced_arg", "")
+    no_cache = more_req.startswith("--no-cache")
+    if no_cache: more_req.lstrip("--no-cache")
+    allow_cache = not no_cache
+    _switch_prompt_ = partial(switch_prompt, more_requirement=more_req)
+
+    # <-------------- check deps ------------->
+    try:
+        import glob, os, time, subprocess
+        subprocess.Popen(['pdflatex', '-version'])
+        from .latex_fns.latex_actions import Latex精细分解与转化, 编译Latex
+    except Exception as e:
+        chatbot.append([f"解析项目: {txt}",
+                        f"尝试执行Latex指令失败。Latex没有安装, 或者不在环境变量PATH中。安装方法https://tug.org/texlive/。报错信息\n\n```\n\n{trimmed_format_exc()}\n\n```\n\n"])
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+
+    # <-------------- clear history and read input ------------->
+    history = []
+    try:
+        txt, arxiv_id = yield from arxiv_download(chatbot, history, txt, allow_cache)
+    except tarfile.ReadError as e:
+        yield from update_ui_lastest_msg(
+            "无法自动下载该论文的Latex源码，请前往arxiv打开此论文下载页面，点other Formats，然后download source手动下载latex源码包。接下来调用本地Latex翻译插件即可。",
+            chatbot=chatbot, history=history)
+        return
+
+    if txt.endswith('.pdf'):
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"发现已经存在翻译好的PDF文档")
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+
+    if os.path.exists(txt):
+        project_folder = txt
+    else:
+        if txt == "": txt = '空空如也的输入栏'
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到本地项目或无法处理: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+
+    file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.tex', recursive=True)]
+    if len(file_manifest) == 0:
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何.tex文件: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+
+    # <-------------- if is a zip/tar file ------------->
+    project_folder = desend_to_extracted_folder_if_exist(project_folder)
+
+    # <-------------- move latex project away from temp folder ------------->
+    from shared_utils.fastapi_server import validate_path_safety
+    validate_path_safety(project_folder, chatbot.get_user())
+    project_folder = move_project(project_folder, arxiv_id)
+
+    # <-------------- if merge_translate_zh is already generated, skip gpt req ------------->
+    if not os.path.exists(project_folder + '/merge_translate_zh.tex'):
+        yield from Latex精细分解与转化(file_manifest, project_folder, llm_kwargs, plugin_kwargs,
+                                       chatbot, history, system_prompt, mode='translate_zh',
+                                       switch_prompt=_switch_prompt_)
+
+    # <-------------- compile PDF ------------->
+    success = yield from 编译Latex(chatbot, history, main_file_original='merge',
+                                   main_file_modified='merge_translate_zh', mode='translate_zh',
+                                   work_folder_original=project_folder, work_folder_modified=project_folder,
+                                   work_folder=project_folder)
+
+    # <-------------- zip PDF ------------->
+    zip_res = zip_result(project_folder)
+    if success:
+        chatbot.append((f"成功啦", '请查收结果（压缩包）...'))
+        yield from update_ui(chatbot=chatbot, history=history);
+        time.sleep(1)  # 刷新界面
+        promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
+    else:
+        chatbot.append((f"失败了",
+                        '虽然PDF生成失败了, 但请查收结果（压缩包）, 内含已经翻译的Tex文档, 您可以到Github Issue区, 用该压缩包进行反馈。如系统是Linux，请检查系统字体（见Github wiki） ...'))
+        yield from update_ui(chatbot=chatbot, history=history);
+        time.sleep(1)  # 刷新界面
+        promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
+
+    # <-------------- we are done ------------->
+    return success
+
+
+#  =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- 插件主程序3  =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
+
+@CatchException
+def PDF翻译中文并重新编译PDF(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+    # <-------------- information about this plugin ------------->
+    chatbot.append([
+        "函数插件功能？",
+        "将PDF转换为Latex项目，翻译为中文后重新编译为PDF。函数插件贡献者: Marroh。注意事项: 此插件Windows支持最佳，Linux下必须使用Docker安装，详见项目主README.md。目前对机器学习类文献转化效果最好，其他类型文献转化效果未知。"])
+    yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+
+    # <-------------- more requirements ------------->
+    if ("advanced_arg" in plugin_kwargs) and (plugin_kwargs["advanced_arg"] == ""): plugin_kwargs.pop("advanced_arg")
+    more_req = plugin_kwargs.get("advanced_arg", "")
+    no_cache = more_req.startswith("--no-cache")
+    if no_cache: more_req.lstrip("--no-cache")
+    allow_cache = not no_cache
+    _switch_prompt_ = partial(switch_prompt, more_requirement=more_req)
+
+    # <-------------- check deps ------------->
+    try:
+        import glob, os, time, subprocess
+        subprocess.Popen(['pdflatex', '-version'])
+        from .latex_fns.latex_actions import Latex精细分解与转化, 编译Latex
+    except Exception as e:
+        chatbot.append([f"解析项目: {txt}",
+                        f"尝试执行Latex指令失败。Latex没有安装, 或者不在环境变量PATH中。安装方法https://tug.org/texlive/。报错信息\n\n```\n\n{trimmed_format_exc()}\n\n```\n\n"])
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+
+    # <-------------- clear history and read input ------------->
+    if os.path.exists(txt):
+        project_folder = txt
+    else:
+        if txt == "": txt = '空空如也的输入栏'
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到本地项目或无法处理: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+
+    file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.pdf', recursive=True)]
+    if len(file_manifest) == 0:
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何.pdf文件: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+    if len(file_manifest) != 1:
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"不支持同时处理多个pdf文件: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+
+    if plugin_kwargs.get("method", "") == 'MATHPIX':
+        app_id, app_key = get_conf('MATHPIX_APPID', 'MATHPIX_APPKEY')
+        if len(app_id) == 0 or len(app_key) == 0:
+            report_exception(chatbot, history, a="缺失 MATHPIX_APPID 和 MATHPIX_APPKEY。", b=f"请配置 MATHPIX_APPID 和 MATHPIX_APPKEY")
+            yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+            return
+    if plugin_kwargs.get("method", "") == 'DOC2X':
+        app_id, app_key = "", ""
+        DOC2X_API_KEY = get_conf('DOC2X_API_KEY')
+        if len(DOC2X_API_KEY) == 0:
+            report_exception(chatbot, history, a="缺失 DOC2X_API_KEY。", b=f"请配置 DOC2X_API_KEY")
+            yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+            return
+
+    hash_tag = map_file_to_sha256(file_manifest[0])
+
+    # # <-------------- check repeated pdf ------------->
+    # chatbot.append([f"检查PDF是否被重复上传", "正在检查..."])
+    # yield from update_ui(chatbot=chatbot, history=history)
+    # repeat, project_folder = check_repeat_upload(file_manifest[0], hash_tag)
+
+    # if repeat:
+    #     yield from update_ui_lastest_msg(f"发现重复上传，请查收结果（压缩包）...", chatbot=chatbot, history=history)
+    #     try:
+    #         translate_pdf = [f for f in glob.glob(f'{project_folder}/**/merge_translate_zh.pdf', recursive=True)][0]
+    #         promote_file_to_downloadzone(translate_pdf, rename_file=None, chatbot=chatbot)
+    #         comparison_pdf = [f for f in glob.glob(f'{project_folder}/**/comparison.pdf', recursive=True)][0]
+    #         promote_file_to_downloadzone(comparison_pdf, rename_file=None, chatbot=chatbot)
+    #         zip_res = zip_result(project_folder)
+    #         promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
+    #         return
+    #     except:
+    #         report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"发现重复上传，但是无法找到相关文件")
+    #         yield from update_ui(chatbot=chatbot, history=history)
+    # else:
+    #     yield from update_ui_lastest_msg(f"未发现重复上传", chatbot=chatbot, history=history)
+
+    # <-------------- convert pdf into tex ------------->
+    chatbot.append([f"解析项目: {txt}", "正在将PDF转换为tex项目，请耐心等待..."])
+    yield from update_ui(chatbot=chatbot, history=history)
+    project_folder = pdf2tex_project(file_manifest[0], plugin_kwargs)
+    if project_folder is None:
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"PDF转换为tex项目失败")
+        yield from update_ui(chatbot=chatbot, history=history)
+        return False
+
+    # <-------------- translate latex file into Chinese ------------->
+    yield from update_ui_lastest_msg("正在tex项目将翻译为中文...", chatbot=chatbot, history=history)
+    file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.tex', recursive=True)]
+    if len(file_manifest) == 0:
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何.tex文件: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history)  # 刷新界面
+        return
+
+    # <-------------- if is a zip/tar file ------------->
+    project_folder = desend_to_extracted_folder_if_exist(project_folder)
+
+    # <-------------- move latex project away from temp folder ------------->
+    from shared_utils.fastapi_server import validate_path_safety
+    validate_path_safety(project_folder, chatbot.get_user())
+    project_folder = move_project(project_folder)
+
+    # <-------------- set a hash tag for repeat-checking ------------->
+    with open(pj(project_folder, hash_tag + '.tag'), 'w') as f:
+        f.write(hash_tag)
+        f.close()
+
+
+    # <-------------- if merge_translate_zh is already generated, skip gpt req ------------->
+    if not os.path.exists(project_folder + '/merge_translate_zh.tex'):
+        yield from Latex精细分解与转化(file_manifest, project_folder, llm_kwargs, plugin_kwargs,
+                                    chatbot, history, system_prompt, mode='translate_zh',
+                                    switch_prompt=_switch_prompt_)
+
+    # <-------------- compile PDF ------------->
+    yield from update_ui_lastest_msg("正在将翻译好的项目tex项目编译为PDF...", chatbot=chatbot, history=history)
+    success = yield from 编译Latex(chatbot, history, main_file_original='merge',
+                                main_file_modified='merge_translate_zh', mode='translate_zh',
+                                work_folder_original=project_folder, work_folder_modified=project_folder,
+                                work_folder=project_folder)
+
+    # <-------------- zip PDF ------------->
+    zip_res = zip_result(project_folder)
+    if success:
+        chatbot.append((f"成功啦", '请查收结果（压缩包）...'))
+        yield from update_ui(chatbot=chatbot, history=history);
+        time.sleep(1)  # 刷新界面
+        promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
+    else:
+        chatbot.append((f"失败了",
+                        '虽然PDF生成失败了, 但请查收结果（压缩包）, 内含已经翻译的Tex文档, 您可以到Github Issue区, 用该压缩包进行反馈。如系统是Linux，请检查系统字体（见Github wiki） ...'))
+        yield from update_ui(chatbot=chatbot, history=history);
+        time.sleep(1)  # 刷新界面
+        promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
+
+    # <-------------- we are done ------------->
+    return success

crazy_functions/Latex_Function_Wrap.py

@@ -0,0 +1,78 @@
+
+from crazy_functions.Latex_Function import Latex翻译中文并重新编译PDF, PDF翻译中文并重新编译PDF
+from crazy_functions.plugin_template.plugin_class_template import GptAcademicPluginTemplate, ArgProperty
+
+
+class Arxiv_Localize(GptAcademicPluginTemplate):
+    def __init__(self):
+        """
+        请注意`execute`会执行在不同的线程中，因此您在定义和使用类变量时，应当慎之又慎！
+        """
+        pass
+
+    def define_arg_selection_menu(self):
+        """
+        定义插件的二级选项菜单
+
+        第一个参数，名称`main_input`，参数`type`声明这是一个文本框，文本框上方显示`title`，文本框内部显示`description`，`default_value`为默认值；
+        第二个参数，名称`advanced_arg`，参数`type`声明这是一个文本框，文本框上方显示`title`，文本框内部显示`description`，`default_value`为默认值；
+        第三个参数，名称`allow_cache`，参数`type`声明这是一个下拉菜单，下拉菜单上方显示`title`+`description`，下拉菜单的选项为`options`，`default_value`为下拉菜单默认值；
+
+        """
+        gui_definition = {
+            "main_input":
+                ArgProperty(title="ArxivID", description="输入Arxiv的ID或者网址", default_value="", type="string").model_dump_json(), # 主输入，自动从输入框同步
+            "advanced_arg":
+                ArgProperty(title="额外的翻译提示词",
+                            description=r"如果有必要, 请在此处给出自定义翻译命令, 解决部分词汇翻译不准确的问题。 "
+                                        r"例如当单词'agent'翻译不准确时, 请尝试把以下指令复制到高级参数区: "
+                                        r'If the term "agent" is used in this section, it should be translated to "智能体". ',
+                            default_value="", type="string").model_dump_json(), # 高级参数输入区，自动同步
+            "allow_cache":
+                ArgProperty(title="是否允许从缓存中调取结果", options=["允许缓存", "从头执行"], default_value="允许缓存", description="无", type="dropdown").model_dump_json(),
+        }
+        return gui_definition
+
+    def execute(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+        """
+        执行插件
+        """
+        allow_cache = plugin_kwargs["allow_cache"]
+        advanced_arg = plugin_kwargs["advanced_arg"]
+
+        if allow_cache == "从头执行": plugin_kwargs["advanced_arg"] = "--no-cache " + plugin_kwargs["advanced_arg"]
+        yield from Latex翻译中文并重新编译PDF(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request)
+
+
+
+class PDF_Localize(GptAcademicPluginTemplate):
+    def __init__(self):
+        """
+        请注意`execute`会执行在不同的线程中，因此您在定义和使用类变量时，应当慎之又慎！
+        """
+        pass
+
+    def define_arg_selection_menu(self):
+        """
+        定义插件的二级选项菜单
+        """
+        gui_definition = {
+            "main_input":
+                ArgProperty(title="PDF文件路径", description="未指定路径，请上传文件后，再点击该插件", default_value="", type="string").model_dump_json(), # 主输入，自动从输入框同步
+            "advanced_arg":
+                ArgProperty(title="额外的翻译提示词",
+                            description=r"如果有必要, 请在此处给出自定义翻译命令, 解决部分词汇翻译不准确的问题。 "
+                                        r"例如当单词'agent'翻译不准确时, 请尝试把以下指令复制到高级参数区: "
+                                        r'If the term "agent" is used in this section, it should be translated to "智能体". ',
+                            default_value="", type="string").model_dump_json(), # 高级参数输入区，自动同步
+            "method":
+                ArgProperty(title="采用哪种方法执行转换", options=["MATHPIX", "DOC2X"], default_value="DOC2X", description="无", type="dropdown").model_dump_json(),
+
+        }
+        return gui_definition
+
+    def execute(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+        """
+        执行插件
+        """
+        yield from PDF翻译中文并重新编译PDF(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request)

crazy_functions/Latex全文润色.py

@@ -1,6 +1,6 @@
-from toolbox import update_ui, trimmed_format_exc
-from toolbox import CatchException, report_execption, write_results_to_file, zip_folder
-
+from toolbox import update_ui, trimmed_format_exc, promote_file_to_downloadzone, get_log_folder
+from toolbox import CatchException, report_exception, write_history_to_file, zip_folder
+from loguru import logger
 
 class PaperFileGroup():
     def __init__(self):
@@ -11,7 +11,7 @@ class PaperFileGroup():
         self.sp_file_tag = []
 
         # count_token
-        from request_llm.bridge_all import model_info
+        from request_llms.bridge_all import model_info
         enc = model_info["gpt-3.5-turbo"]['tokenizer']
         def get_token_num(txt): return len(enc.encode(txt, disallowed_special=()))
         self.get_token_num = get_token_num
@@ -26,14 +26,14 @@ class PaperFileGroup():
                 self.sp_file_index.append(index)
                 self.sp_file_tag.append(self.file_paths[index])
             else:
-                from .crazy_utils import breakdown_txt_to_satisfy_token_limit_for_pdf
-                segments = breakdown_txt_to_satisfy_token_limit_for_pdf(file_content, self.get_token_num, max_token_limit)
+                from crazy_functions.pdf_fns.breakdown_txt import breakdown_text_to_satisfy_token_limit
+                segments = breakdown_text_to_satisfy_token_limit(file_content, max_token_limit)
                 for j, segment in enumerate(segments):
                     self.sp_file_contents.append(segment)
                     self.sp_file_index.append(index)
                     self.sp_file_tag.append(self.file_paths[index] + f".part-{j}.tex")
 
-        print('Segmentation: done')
+        logger.info('Segmentation: done')
     def merge_result(self):
         self.file_result = ["" for _ in range(len(self.file_paths))]
         for r, k in zip(self.sp_file_result, self.sp_file_index):
@@ -51,12 +51,12 @@ class PaperFileGroup():
         import os, time
         folder = os.path.dirname(self.file_paths[0])
         t = time.strftime("%Y-%m-%d-%H-%M-%S", time.localtime())
-        zip_folder(folder, './gpt_log/', f'{t}-polished.zip')
+        zip_folder(folder, get_log_folder(), f'{t}-polished.zip')
 
 
 def 多文件润色(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, language='en', mode='polish'):
     import time, os, re
-    from .crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
+    from crazy_functions.crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
 
 
     #  <-------- 读取Latex文件，删除其中的所有注释 ---------->
@@ -81,8 +81,8 @@ def 多文件润色(file_manifest, project_folder, llm_kwargs, plugin_kwargs, ch
     #  <-------- 多线程润色开始 ---------->
     if language == 'en':
         if mode == 'polish':
-            inputs_array = ["Below is a section from an academic paper, polish this section to meet the academic standard, " + 
-                            "improve the grammar, clarity and overall readability, do not modify any latex command such as \section, \cite and equations:" + 
+            inputs_array = [r"Below is a section from an academic paper, polish this section to meet the academic standard, " +
+                            r"improve the grammar, clarity and overall readability, do not modify any latex command such as \section, \cite and equations:" +
                             f"\n\n{frag}" for frag in pfg.sp_file_contents]
         else:
             inputs_array = [r"Below is a section from an academic paper, proofread this section." +
@@ -93,10 +93,10 @@ def 多文件润色(file_manifest, project_folder, llm_kwargs, plugin_kwargs, ch
         sys_prompt_array = ["You are a professional academic paper writer." for _ in range(n_split)]
     elif language == 'zh':
         if mode == 'polish':
-            inputs_array = [f"以下是一篇学术论文中的一段内容，请将此部分润色以满足学术标准，提高语法、清晰度和整体可读性，不要修改任何LaTeX命令，例如\section，\cite和方程式：" + 
+            inputs_array = [r"以下是一篇学术论文中的一段内容，请将此部分润色以满足学术标准，提高语法、清晰度和整体可读性，不要修改任何LaTeX命令，例如\section，\cite和方程式：" +
                             f"\n\n{frag}" for frag in pfg.sp_file_contents]
         else:
-            inputs_array = [f"以下是一篇学术论文中的一段内容，请对这部分内容进行语法矫正。不要修改任何LaTeX命令，例如\section，\cite和方程式：" + 
+            inputs_array = [r"以下是一篇学术论文中的一段内容，请对这部分内容进行语法矫正。不要修改任何LaTeX命令，例如\section，\cite和方程式：" +
                             f"\n\n{frag}" for frag in pfg.sp_file_contents]
         inputs_show_user_array = [f"润色 {f}" for f in pfg.sp_file_tag]
         sys_prompt_array=["你是一位专业的中文学术论文作家。" for _ in range(n_split)]
@@ -122,29 +122,31 @@ def 多文件润色(file_manifest, project_folder, llm_kwargs, plugin_kwargs, ch
         pfg.write_result()
         pfg.zip_result()
     except:
-        print(trimmed_format_exc())
+        logger.error(trimmed_format_exc())
 
     #  <-------- 整理结果，退出 ---------->
     create_report_file_name = time.strftime("%Y-%m-%d-%H-%M-%S", time.localtime()) + f"-chatgpt.polish.md"
-    res = write_results_to_file(gpt_response_collection, file_name=create_report_file_name)
+    res = write_history_to_file(gpt_response_collection, file_basename=create_report_file_name)
+    promote_file_to_downloadzone(res, chatbot=chatbot)
+
     history = gpt_response_collection
     chatbot.append((f"{fp}完成了吗？", res))
     yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
 
 
 @CatchException
-def Latex英文润色(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def Latex英文润色(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     # 基本信息：功能、贡献者
     chatbot.append([
         "函数插件功能？",
-        "对整个Latex项目进行润色。函数插件贡献者: Binary-Husky"])
+        "对整个Latex项目进行润色。函数插件贡献者: Binary-Husky。（注意，此插件不调用Latex，如果有Latex环境，请使用「Latex英文纠错+高亮修正位置(需Latex)插件」"])
     yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
 
     # 尝试导入依赖，如果缺少依赖，则给出安装建议
     try:
         import tiktoken
     except:
-        report_execption(chatbot, history,
+        report_exception(chatbot, history,
                          a=f"解析项目: {txt}",
                          b=f"导入软件依赖失败。使用该模块需要额外依赖，安装方法```pip install --upgrade tiktoken```。")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
@@ -155,12 +157,12 @@ def Latex英文润色(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_p
         project_folder = txt
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.tex', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 多文件润色(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, language='en')
@@ -171,7 +173,7 @@ def Latex英文润色(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_p
 
 
 @CatchException
-def Latex中文润色(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def Latex中文润色(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     # 基本信息：功能、贡献者
     chatbot.append([
         "函数插件功能？",
@@ -182,7 +184,7 @@ def Latex中文润色(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_p
     try:
         import tiktoken
     except:
-        report_execption(chatbot, history,
+        report_exception(chatbot, history,
                          a=f"解析项目: {txt}",
                          b=f"导入软件依赖失败。使用该模块需要额外依赖，安装方法```pip install --upgrade tiktoken```。")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
@@ -193,12 +195,12 @@ def Latex中文润色(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_p
         project_folder = txt
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.tex', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 多文件润色(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, language='zh')
@@ -207,7 +209,7 @@ def Latex中文润色(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_p
 
 
 @CatchException
-def Latex英文纠错(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def Latex英文纠错(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     # 基本信息：功能、贡献者
     chatbot.append([
         "函数插件功能？",
@@ -218,7 +220,7 @@ def Latex英文纠错(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_p
     try:
         import tiktoken
     except:
-        report_execption(chatbot, history,
+        report_exception(chatbot, history,
                          a=f"解析项目: {txt}",
                          b=f"导入软件依赖失败。使用该模块需要额外依赖，安装方法```pip install --upgrade tiktoken```。")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
@@ -229,12 +231,12 @@ def Latex英文纠错(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_p
         project_folder = txt
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.tex', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 多文件润色(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, language='en', mode='proofread')

crazy_functions/Latex全文翻译.py

@@ -1,6 +1,6 @@
-from toolbox import update_ui
-from toolbox import CatchException, report_execption, write_results_to_file
-fast_debug = False
+from toolbox import update_ui, promote_file_to_downloadzone
+from toolbox import CatchException, report_exception, write_history_to_file
+from loguru import logger
 
 class PaperFileGroup():
     def __init__(self):
@@ -11,7 +11,7 @@ class PaperFileGroup():
         self.sp_file_tag = []
 
         # count_token
-        from request_llm.bridge_all import model_info
+        from request_llms.bridge_all import model_info
         enc = model_info["gpt-3.5-turbo"]['tokenizer']
         def get_token_num(txt): return len(enc.encode(txt, disallowed_special=()))
         self.get_token_num = get_token_num
@@ -26,18 +26,18 @@ class PaperFileGroup():
                 self.sp_file_index.append(index)
                 self.sp_file_tag.append(self.file_paths[index])
             else:
-                from .crazy_utils import breakdown_txt_to_satisfy_token_limit_for_pdf
-                segments = breakdown_txt_to_satisfy_token_limit_for_pdf(file_content, self.get_token_num, max_token_limit)
+                from crazy_functions.pdf_fns.breakdown_txt import breakdown_text_to_satisfy_token_limit
+                segments = breakdown_text_to_satisfy_token_limit(file_content, max_token_limit)
                 for j, segment in enumerate(segments):
                     self.sp_file_contents.append(segment)
                     self.sp_file_index.append(index)
                     self.sp_file_tag.append(self.file_paths[index] + f".part-{j}.tex")
 
-        print('Segmentation: done')
+        logger.info('Segmentation: done')
 
 def 多文件翻译(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, language='en'):
     import time, os, re
-    from .crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
+    from crazy_functions.crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
 
     #  <-------- 读取Latex文件，删除其中的所有注释 ---------->
     pfg = PaperFileGroup()
@@ -95,7 +95,8 @@ def 多文件翻译(file_manifest, project_folder, llm_kwargs, plugin_kwargs, ch
 
     #  <-------- 整理结果，退出 ---------->
     create_report_file_name = time.strftime("%Y-%m-%d-%H-%M-%S", time.localtime()) + f"-chatgpt.polish.md"
-    res = write_results_to_file(gpt_response_collection, file_name=create_report_file_name)
+    res = write_history_to_file(gpt_response_collection, create_report_file_name)
+    promote_file_to_downloadzone(res, chatbot=chatbot)
     history = gpt_response_collection
     chatbot.append((f"{fp}完成了吗？", res))
     yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
@@ -105,7 +106,7 @@ def 多文件翻译(file_manifest, project_folder, llm_kwargs, plugin_kwargs, ch
 
 
 @CatchException
-def Latex英译中(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def Latex英译中(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     # 基本信息：功能、贡献者
     chatbot.append([
         "函数插件功能？",
@@ -116,7 +117,7 @@ def Latex英译中(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prom
     try:
         import tiktoken
     except:
-        report_execption(chatbot, history,
+        report_exception(chatbot, history,
                          a=f"解析项目: {txt}",
                          b=f"导入软件依赖失败。使用该模块需要额外依赖，安装方法```pip install --upgrade tiktoken```。")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
@@ -127,12 +128,12 @@ def Latex英译中(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prom
         project_folder = txt
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.tex', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 多文件翻译(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, language='en->zh')
@@ -142,7 +143,7 @@ def Latex英译中(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prom
 
 
 @CatchException
-def Latex中译英(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def Latex中译英(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     # 基本信息：功能、贡献者
     chatbot.append([
         "函数插件功能？",
@@ -153,7 +154,7 @@ def Latex中译英(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prom
     try:
         import tiktoken
     except:
-        report_execption(chatbot, history,
+        report_exception(chatbot, history,
                          a=f"解析项目: {txt}",
                          b=f"导入软件依赖失败。使用该模块需要额外依赖，安装方法```pip install --upgrade tiktoken```。")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
@@ -164,12 +165,12 @@ def Latex中译英(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prom
         project_folder = txt
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.tex', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 多文件翻译(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, language='zh->en')

crazy_functions/Latex输出PDF结果.py

@@ -1,300 +0,0 @@
-from toolbox import update_ui, trimmed_format_exc, get_conf, objdump, objload, promote_file_to_downloadzone
-from toolbox import CatchException, report_execption, update_ui_lastest_msg, zip_result, gen_time_str
-from functools import partial
-import glob, os, requests, time
-pj = os.path.join
-ARXIV_CACHE_DIR = os.path.expanduser(f"~/arxiv_cache/")
-
-# =================================== 工具函数 ===============================================
-专业词汇声明  = 'If the term "agent" is used in this section, it should be translated to "智能体". '
-def switch_prompt(pfg, mode, more_requirement):
-    """
-    Generate prompts and system prompts based on the mode for proofreading or translating.
-    Args:
-    - pfg: Proofreader or Translator instance.
-    - mode: A string specifying the mode, either 'proofread' or 'translate_zh'.
-
-    Returns:
-    - inputs_array: A list of strings containing prompts for users to respond to.
-    - sys_prompt_array: A list of strings containing prompts for system prompts.
-    """
-    n_split = len(pfg.sp_file_contents)
-    if mode == 'proofread_en':
-        inputs_array = [r"Below is a section from an academic paper, proofread this section." + 
-                        r"Do not modify any latex command such as \section, \cite, \begin, \item and equations. " + more_requirement +
-                        r"Answer me only with the revised text:" + 
-                        f"\n\n{frag}" for frag in pfg.sp_file_contents]
-        sys_prompt_array = ["You are a professional academic paper writer." for _ in range(n_split)]
-    elif mode == 'translate_zh':
-        inputs_array = [r"Below is a section from an English academic paper, translate it into Chinese. " + more_requirement + 
-                        r"Do not modify any latex command such as \section, \cite, \begin, \item and equations. " + 
-                        r"Answer me only with the translated text:" + 
-                        f"\n\n{frag}" for frag in pfg.sp_file_contents]
-        sys_prompt_array = ["You are a professional translator." for _ in range(n_split)]
-    else:
-        assert False, "未知指令"
-    return inputs_array, sys_prompt_array
-
-def desend_to_extracted_folder_if_exist(project_folder):
-    """ 
-    Descend into the extracted folder if it exists, otherwise return the original folder.
-
-    Args:
-    - project_folder: A string specifying the folder path.
-
-    Returns:
-    - A string specifying the path to the extracted folder, or the original folder if there is no extracted folder.
-    """
-    maybe_dir = [f for f in glob.glob(f'{project_folder}/*') if os.path.isdir(f)]
-    if len(maybe_dir) == 0: return project_folder
-    if maybe_dir[0].endswith('.extract'): return maybe_dir[0]
-    return project_folder
-
-def move_project(project_folder, arxiv_id=None):
-    """ 
-    Create a new work folder and copy the project folder to it.
-
-    Args:
-    - project_folder: A string specifying the folder path of the project.
-
-    Returns:
-    - A string specifying the path to the new work folder.
-    """
-    import shutil, time
-    time.sleep(2)   # avoid time string conflict
-    if arxiv_id is not None:
-        new_workfolder = pj(ARXIV_CACHE_DIR, arxiv_id, 'workfolder')
-    else:
-        new_workfolder = f'gpt_log/{gen_time_str()}'
-    try:
-        shutil.rmtree(new_workfolder)
-    except:
-        pass
-
-    # align subfolder if there is a folder wrapper
-    items = glob.glob(pj(project_folder,'*'))
-    if len(glob.glob(pj(project_folder,'*.tex'))) == 0 and len(items) == 1:
-        if os.path.isdir(items[0]): project_folder = items[0]
-
-    shutil.copytree(src=project_folder, dst=new_workfolder)
-    return new_workfolder
-
-def arxiv_download(chatbot, history, txt):
-    def check_cached_translation_pdf(arxiv_id):
-        translation_dir = pj(ARXIV_CACHE_DIR, arxiv_id, 'translation')
-        if not os.path.exists(translation_dir):
-            os.makedirs(translation_dir)
-        target_file = pj(translation_dir, 'translate_zh.pdf')
-        if os.path.exists(target_file):
-            promote_file_to_downloadzone(target_file, rename_file=None, chatbot=chatbot)
-            return target_file
-        return False
-    def is_float(s):
-        try:
-            float(s)
-            return True
-        except ValueError:
-            return False
-    if ('.' in txt) and ('/' not in txt) and is_float(txt): # is arxiv ID
-        txt = 'https://arxiv.org/abs/' + txt.strip()
-    if ('.' in txt) and ('/' not in txt) and is_float(txt[:10]): # is arxiv ID
-        txt = 'https://arxiv.org/abs/' + txt[:10]
-    if not txt.startswith('https://arxiv.org'): 
-        return txt, None
-    
-    # <-------------- inspect format ------------->
-    chatbot.append([f"检测到arxiv文档连接", '尝试下载 ...']) 
-    yield from update_ui(chatbot=chatbot, history=history)
-    time.sleep(1) # 刷新界面
-
-    url_ = txt   # https://arxiv.org/abs/1707.06690
-    if not txt.startswith('https://arxiv.org/abs/'): 
-        msg = f"解析arxiv网址失败, 期望格式例如: https://arxiv.org/abs/1707.06690。实际得到格式: {url_}"
-        yield from update_ui_lastest_msg(msg, chatbot=chatbot, history=history) # 刷新界面
-        return msg, None
-    # <-------------- set format ------------->
-    arxiv_id = url_.split('/abs/')[-1]
-    if 'v' in arxiv_id: arxiv_id = arxiv_id[:10]
-    cached_translation_pdf = check_cached_translation_pdf(arxiv_id)
-    if cached_translation_pdf: return cached_translation_pdf, arxiv_id
-
-    url_tar = url_.replace('/abs/', '/e-print/')
-    translation_dir = pj(ARXIV_CACHE_DIR, arxiv_id, 'e-print')
-    extract_dst = pj(ARXIV_CACHE_DIR, arxiv_id, 'extract')
-    os.makedirs(translation_dir, exist_ok=True)
-    
-    # <-------------- download arxiv source file ------------->
-    dst = pj(translation_dir, arxiv_id+'.tar')
-    if os.path.exists(dst):
-        yield from update_ui_lastest_msg("调用缓存", chatbot=chatbot, history=history)  # 刷新界面
-    else:
-        yield from update_ui_lastest_msg("开始下载", chatbot=chatbot, history=history)  # 刷新界面
-        proxies, = get_conf('proxies')
-        r = requests.get(url_tar, proxies=proxies)
-        with open(dst, 'wb+') as f:
-            f.write(r.content)
-    # <-------------- extract file ------------->
-    yield from update_ui_lastest_msg("下载完成", chatbot=chatbot, history=history)  # 刷新界面
-    from toolbox import extract_archive
-    extract_archive(file_path=dst, dest_dir=extract_dst)
-    return extract_dst, arxiv_id
-# ========================================= 插件主程序1 =====================================================    
-
-
-@CatchException
-def Latex英文纠错加PDF对比(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
-    # <-------------- information about this plugin ------------->
-    chatbot.append([ "函数插件功能？",
-        "对整个Latex项目进行纠错, 用latex编译为PDF对修正处做高亮。函数插件贡献者: Binary-Husky。注意事项: 目前仅支持GPT3.5/GPT4，其他模型转化效果未知。目前对机器学习类文献转化效果最好，其他类型文献转化效果未知。仅在Windows系统进行了测试，其他操作系统表现未知。"])
-    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-    
-    # <-------------- more requirements ------------->
-    if ("advanced_arg" in plugin_kwargs) and (plugin_kwargs["advanced_arg"] == ""): plugin_kwargs.pop("advanced_arg")
-    more_req = plugin_kwargs.get("advanced_arg", "")
-    _switch_prompt_ = partial(switch_prompt, more_requirement=more_req)
-
-    # <-------------- check deps ------------->
-    try:
-        import glob, os, time, subprocess
-        subprocess.Popen(['pdflatex', '-version'])
-        from .latex_fns.latex_actions import Latex精细分解与转化, 编译Latex
-    except Exception as e:
-        chatbot.append([ f"解析项目: {txt}",
-            f"尝试执行Latex指令失败。Latex没有安装, 或者不在环境变量PATH中。安装方法https://tug.org/texlive/。报错信息\n\n```\n\n{trimmed_format_exc()}\n\n```\n\n"])
-        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-        return
-    
-
-    # <-------------- clear history and read input ------------->
-    history = []
-    if os.path.exists(txt):
-        project_folder = txt
-    else:
-        if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
-        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-        return
-    file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.tex', recursive=True)]
-    if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
-        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-        return
-    
-
-    # <-------------- if is a zip/tar file ------------->
-    project_folder = desend_to_extracted_folder_if_exist(project_folder)
-
-
-    # <-------------- move latex project away from temp folder ------------->
-    project_folder = move_project(project_folder, arxiv_id=None)
-
-
-    # <-------------- if merge_translate_zh is already generated, skip gpt req ------------->
-    if not os.path.exists(project_folder + '/merge_proofread_en.tex'):
-        yield from Latex精细分解与转化(file_manifest, project_folder, llm_kwargs, plugin_kwargs, 
-                                chatbot, history, system_prompt, mode='proofread_en', switch_prompt=_switch_prompt_)
-
-
-    # <-------------- compile PDF ------------->
-    success = yield from 编译Latex(chatbot, history, main_file_original='merge', main_file_modified='merge_proofread_en', 
-                             work_folder_original=project_folder, work_folder_modified=project_folder, work_folder=project_folder)
-    
-
-    # <-------------- zip PDF ------------->
-    zip_res = zip_result(project_folder)
-    if success:
-        chatbot.append((f"成功啦", '请查收结果（压缩包）...'))
-        yield from update_ui(chatbot=chatbot, history=history); time.sleep(1) # 刷新界面
-        promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
-    else:
-        chatbot.append((f"失败了", '虽然PDF生成失败了, 但请查收结果（压缩包）, 内含已经翻译的Tex文档, 也是可读的, 您可以到Github Issue区, 用该压缩包+对话历史存档进行反馈 ...'))
-        yield from update_ui(chatbot=chatbot, history=history); time.sleep(1) # 刷新界面
-        promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
-
-    # <-------------- we are done ------------->
-    return success
-
-
-# ========================================= 插件主程序2 =====================================================    
-
-@CatchException
-def Latex翻译中文并重新编译PDF(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
-    # <-------------- information about this plugin ------------->
-    chatbot.append([
-        "函数插件功能？",
-        "对整个Latex项目进行翻译, 生成中文PDF。函数插件贡献者: Binary-Husky。注意事项: 此插件Windows支持最佳，Linux下必须使用Docker安装，详见项目主README.md。目前仅支持GPT3.5/GPT4，其他模型转化效果未知。目前对机器学习类文献转化效果最好，其他类型文献转化效果未知。"])
-    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-
-    # <-------------- more requirements ------------->
-    if ("advanced_arg" in plugin_kwargs) and (plugin_kwargs["advanced_arg"] == ""): plugin_kwargs.pop("advanced_arg")
-    more_req = plugin_kwargs.get("advanced_arg", "")
-    _switch_prompt_ = partial(switch_prompt, more_requirement=more_req)
-
-    # <-------------- check deps ------------->
-    try:
-        import glob, os, time, subprocess
-        subprocess.Popen(['pdflatex', '-version'])
-        from .latex_fns.latex_actions import Latex精细分解与转化, 编译Latex
-    except Exception as e:
-        chatbot.append([ f"解析项目: {txt}",
-            f"尝试执行Latex指令失败。Latex没有安装, 或者不在环境变量PATH中。安装方法https://tug.org/texlive/。报错信息\n\n```\n\n{trimmed_format_exc()}\n\n```\n\n"])
-        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-        return
-    
-
-    # <-------------- clear history and read input ------------->
-    history = []
-    txt, arxiv_id = yield from arxiv_download(chatbot, history, txt)
-    if txt.endswith('.pdf'):
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"发现已经存在翻译好的PDF文档")
-        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-        return
-    
-
-    if os.path.exists(txt):
-        project_folder = txt
-    else:
-        if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
-        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-        return
-    
-    file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.tex', recursive=True)]
-    if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.tex文件: {txt}")
-        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-        return
-    
-
-    # <-------------- if is a zip/tar file ------------->
-    project_folder = desend_to_extracted_folder_if_exist(project_folder)
-
-
-    # <-------------- move latex project away from temp folder ------------->
-    project_folder = move_project(project_folder, arxiv_id)
-
-
-    # <-------------- if merge_translate_zh is already generated, skip gpt req ------------->
-    if not os.path.exists(project_folder + '/merge_translate_zh.tex'):
-        yield from Latex精细分解与转化(file_manifest, project_folder, llm_kwargs, plugin_kwargs, 
-                                chatbot, history, system_prompt, mode='translate_zh', switch_prompt=_switch_prompt_)
-
-
-    # <-------------- compile PDF ------------->
-    success = yield from 编译Latex(chatbot, history, main_file_original='merge', main_file_modified='merge_translate_zh', mode='translate_zh', 
-                             work_folder_original=project_folder, work_folder_modified=project_folder, work_folder=project_folder)
-
-    # <-------------- zip PDF ------------->
-    zip_res = zip_result(project_folder)
-    if success:
-        chatbot.append((f"成功啦", '请查收结果（压缩包）...'))
-        yield from update_ui(chatbot=chatbot, history=history); time.sleep(1) # 刷新界面
-        promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
-    else:
-        chatbot.append((f"失败了", '虽然PDF生成失败了, 但请查收结果（压缩包）, 内含已经翻译的Tex文档, 也是可读的, 您可以到Github Issue区, 用该压缩包+对话历史存档进行反馈 ...'))
-        yield from update_ui(chatbot=chatbot, history=history); time.sleep(1) # 刷新界面
-        promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
-
-
-    # <-------------- we are done ------------->
-    return success

crazy_functions/Markdown_Translate.py

@@ -1,7 +1,8 @@
-import glob, time, os, re
-from toolbox import update_ui, trimmed_format_exc, gen_time_str, disable_auto_promotion
-from toolbox import CatchException, report_execption, write_history_to_file
-from toolbox import promote_file_to_downloadzone, get_log_folder
+import glob, shutil, os, re
+from loguru import logger
+from toolbox import update_ui, trimmed_format_exc, gen_time_str
+from toolbox import CatchException, report_exception, get_log_folder
+from toolbox import write_history_to_file, promote_file_to_downloadzone
 fast_debug = False
 
 class PaperFileGroup():
@@ -13,12 +14,12 @@ class PaperFileGroup():
         self.sp_file_tag = []
 
         # count_token
-        from request_llm.bridge_all import model_info
+        from request_llms.bridge_all import model_info
         enc = model_info["gpt-3.5-turbo"]['tokenizer']
         def get_token_num(txt): return len(enc.encode(txt, disallowed_special=()))
         self.get_token_num = get_token_num
 
-    def run_file_split(self, max_token_limit=1900):
+    def run_file_split(self, max_token_limit=2048):
         """
         将长文本分离开来
         """
@@ -28,13 +29,13 @@ class PaperFileGroup():
                 self.sp_file_index.append(index)
                 self.sp_file_tag.append(self.file_paths[index])
             else:
-                from .crazy_utils import breakdown_txt_to_satisfy_token_limit_for_pdf
-                segments = breakdown_txt_to_satisfy_token_limit_for_pdf(file_content, self.get_token_num, max_token_limit)
+                from crazy_functions.pdf_fns.breakdown_txt import breakdown_text_to_satisfy_token_limit
+                segments = breakdown_text_to_satisfy_token_limit(file_content, max_token_limit)
                 for j, segment in enumerate(segments):
                     self.sp_file_contents.append(segment)
                     self.sp_file_index.append(index)
                     self.sp_file_tag.append(self.file_paths[index] + f".part-{j}.md")
-        print('Segmentation: done')
+        logger.info('Segmentation: done')
 
     def merge_result(self):
         self.file_result = ["" for _ in range(len(self.file_paths))]
@@ -51,7 +52,7 @@ class PaperFileGroup():
         return manifest
 
 def 多文件翻译(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, language='en'):
-    from .crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
+    from crazy_functions.crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
 
     #  <-------- 读取Markdown文件，删除其中的所有注释 ---------->
     pfg = PaperFileGroup()
@@ -64,25 +65,25 @@ def 多文件翻译(file_manifest, project_folder, llm_kwargs, plugin_kwargs, ch
             pfg.file_contents.append(file_content)
 
     #  <-------- 拆分过长的Markdown文件 ---------->
-    pfg.run_file_split(max_token_limit=1500)
+    pfg.run_file_split(max_token_limit=2048)
     n_split = len(pfg.sp_file_contents)
 
     #  <-------- 多线程翻译开始 ---------->
     if language == 'en->zh':
-        inputs_array = ["This is a Markdown file, translate it into Chinese, do not modify any existing Markdown commands:" + 
+        inputs_array = ["This is a Markdown file, translate it into Chinese, do NOT modify any existing Markdown commands, do NOT use code wrapper (```), ONLY answer me with translated results:" +
                         f"\n\n{frag}" for frag in pfg.sp_file_contents]
         inputs_show_user_array = [f"翻译 {f}" for f in pfg.sp_file_tag]
-        sys_prompt_array = ["You are a professional academic paper translator." for _ in range(n_split)]
+        sys_prompt_array = ["You are a professional academic paper translator." + plugin_kwargs.get("additional_prompt", "") for _ in range(n_split)]
     elif language == 'zh->en':
-        inputs_array = [f"This is a Markdown file, translate it into English, do not modify any existing Markdown commands:" + 
+        inputs_array = [f"This is a Markdown file, translate it into English, do NOT modify any existing Markdown commands, do NOT use code wrapper (```), ONLY answer me with translated results:" +
                         f"\n\n{frag}" for frag in pfg.sp_file_contents]
         inputs_show_user_array = [f"翻译 {f}" for f in pfg.sp_file_tag]
-        sys_prompt_array = ["You are a professional academic paper translator." for _ in range(n_split)]
+        sys_prompt_array = ["You are a professional academic paper translator." + plugin_kwargs.get("additional_prompt", "") for _ in range(n_split)]
     else:
-        inputs_array = [f"This is a Markdown file, translate it into {language}, do not modify any existing Markdown commands, only answer me with translated results:" + 
+        inputs_array = [f"This is a Markdown file, translate it into {language}, do NOT modify any existing Markdown commands, do NOT use code wrapper (```), ONLY answer me with translated results:" +
                         f"\n\n{frag}" for frag in pfg.sp_file_contents]
         inputs_show_user_array = [f"翻译 {f}" for f in pfg.sp_file_tag]
-        sys_prompt_array = ["You are a professional academic paper translator." for _ in range(n_split)]
+        sys_prompt_array = ["You are a professional academic paper translator." + plugin_kwargs.get("additional_prompt", "") for _ in range(n_split)]
 
     gpt_response_collection = yield from request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
         inputs_array=inputs_array,
@@ -99,9 +100,14 @@ def 多文件翻译(file_manifest, project_folder, llm_kwargs, plugin_kwargs, ch
         for i_say, gpt_say in zip(gpt_response_collection[0::2], gpt_response_collection[1::2]):
             pfg.sp_file_result.append(gpt_say)
         pfg.merge_result()
-        pfg.write_result(language)
+        output_file_arr = pfg.write_result(language)
+        for output_file in output_file_arr:
+            promote_file_to_downloadzone(output_file, chatbot=chatbot)
+            if 'markdown_expected_output_path' in plugin_kwargs:
+                expected_f_name = plugin_kwargs['markdown_expected_output_path']
+                shutil.copyfile(output_file, expected_f_name)
     except:
-        print(trimmed_format_exc())
+        logger.error(trimmed_format_exc())
 
     #  <-------- 整理结果，退出 ---------->
     create_report_file_name = gen_time_str() + f"-chatgpt.md"
@@ -118,10 +124,10 @@ def get_files_from_everything(txt, preference=''):
     if txt.startswith('http'):
         import requests
         from toolbox import get_conf
-        proxies, = get_conf('proxies')
+        proxies = get_conf('proxies')
         # 网络的远程文件
         if preference == 'Github':
-            print('正在从github下载资源 ...')
+            logger.info('正在从github下载资源 ...')
             if not txt.endswith('.md'):
                 # Make a request to the GitHub API to retrieve the repository information
                 url = txt.replace("https://github.com/", "https://api.github.com/repos/") + '/readme'
@@ -145,25 +151,26 @@ def get_files_from_everything(txt, preference=''):
         project_folder = txt
         file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.md', recursive=True)]
     else:
+        project_folder = None
+        file_manifest = []
         success = False
 
     return success, file_manifest, project_folder
 
 
 @CatchException
-def Markdown英译中(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def Markdown英译中(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     # 基本信息：功能、贡献者
     chatbot.append([
         "函数插件功能？",
         "对整个Markdown项目进行翻译。函数插件贡献者: Binary-Husky"])
     yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-    disable_auto_promotion(chatbot)
 
     # 尝试导入依赖，如果缺少依赖，则给出安装建议
     try:
         import tiktoken
     except:
-        report_execption(chatbot, history,
+        report_exception(chatbot, history,
                          a=f"解析项目: {txt}",
                          b=f"导入软件依赖失败。使用该模块需要额外依赖，安装方法```pip install --upgrade tiktoken```。")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
@@ -175,12 +182,12 @@ def Markdown英译中(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_p
     if not success:
         # 什么都没有
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
 
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.md文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.md文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
 
@@ -191,19 +198,18 @@ def Markdown英译中(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_p
 
 
 @CatchException
-def Markdown中译英(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def Markdown中译英(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     # 基本信息：功能、贡献者
     chatbot.append([
         "函数插件功能？",
         "对整个Markdown项目进行翻译。函数插件贡献者: Binary-Husky"])
     yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-    disable_auto_promotion(chatbot)
 
     # 尝试导入依赖，如果缺少依赖，则给出安装建议
     try:
         import tiktoken
     except:
-        report_execption(chatbot, history,
+        report_exception(chatbot, history,
                          a=f"解析项目: {txt}",
                          b=f"导入软件依赖失败。使用该模块需要额外依赖，安装方法```pip install --upgrade tiktoken```。")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
@@ -213,30 +219,29 @@ def Markdown中译英(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_p
     if not success:
         # 什么都没有
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.md文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.md文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 多文件翻译(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, language='zh->en')
 
 
 @CatchException
-def Markdown翻译指定语言(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def Markdown翻译指定语言(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     # 基本信息：功能、贡献者
     chatbot.append([
         "函数插件功能？",
         "对整个Markdown项目进行翻译。函数插件贡献者: Binary-Husky"])
     yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
-    disable_auto_promotion(chatbot)
 
     # 尝试导入依赖，如果缺少依赖，则给出安装建议
     try:
         import tiktoken
     except:
-        report_execption(chatbot, history,
+        report_exception(chatbot, history,
                          a=f"解析项目: {txt}",
                          b=f"导入软件依赖失败。使用该模块需要额外依赖，安装方法```pip install --upgrade tiktoken```。")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
@@ -246,11 +251,11 @@ def Markdown翻译指定语言(txt, llm_kwargs, plugin_kwargs, chatbot, history,
     if not success:
         # 什么都没有
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.md文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.md文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
 

crazy_functions/PDF_Translate.py

@@ -0,0 +1,83 @@
+from toolbox import CatchException, check_packages, get_conf
+from toolbox import update_ui, update_ui_lastest_msg, disable_auto_promotion
+from toolbox import trimmed_format_exc_markdown
+from crazy_functions.crazy_utils import get_files_from_everything
+from crazy_functions.pdf_fns.parse_pdf import get_avail_grobid_url
+from crazy_functions.pdf_fns.parse_pdf_via_doc2x import 解析PDF_基于DOC2X
+from crazy_functions.pdf_fns.parse_pdf_legacy import 解析PDF_简单拆解
+from crazy_functions.pdf_fns.parse_pdf_grobid import 解析PDF_基于GROBID
+from shared_utils.colorful import *
+
+@CatchException
+def 批量翻译PDF文档(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+
+    disable_auto_promotion(chatbot)
+    # 基本信息：功能、贡献者
+    chatbot.append([None, "插件功能：批量翻译PDF文档。函数插件贡献者: Binary-Husky"])
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+
+    # 尝试导入依赖，如果缺少依赖，则给出安装建议
+    try:
+        check_packages(["fitz", "tiktoken", "scipdf"])
+    except:
+        chatbot.append([None, f"导入软件依赖失败。使用该模块需要额外依赖，安装方法```pip install --upgrade pymupdf tiktoken scipdf_parser```。"])
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        return
+
+    # 清空历史，以免输入溢出
+    history = []
+    success, file_manifest, project_folder = get_files_from_everything(txt, type='.pdf')
+
+    # 检测输入参数，如没有给定输入参数，直接退出
+    if (not success) and txt == "": txt = '空空如也的输入栏。提示：请先上传文件（把PDF文件拖入对话）。'
+
+    # 如果没找到任何文件
+    if len(file_manifest) == 0:
+        chatbot.append([None, f"找不到任何.pdf拓展名的文件: {txt}"])
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        return
+
+    # 开始正式执行任务
+    method = plugin_kwargs.get("pdf_parse_method", None)
+    if method == "DOC2X":
+        # ------- 第一种方法，效果最好，但是需要DOC2X服务 -------
+        DOC2X_API_KEY = get_conf("DOC2X_API_KEY")
+        if len(DOC2X_API_KEY) != 0:
+            try:
+                yield from 解析PDF_基于DOC2X(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, DOC2X_API_KEY, user_request)
+                return
+            except:
+                chatbot.append([None, f"DOC2X服务不可用，现在将执行效果稍差的旧版代码。{trimmed_format_exc_markdown()}"])
+                yield from update_ui(chatbot=chatbot, history=history)
+
+    if method == "GROBID":
+        # ------- 第二种方法，效果次优 -------
+        grobid_url = get_avail_grobid_url()
+        if grobid_url is not None:
+            yield from 解析PDF_基于GROBID(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, grobid_url)
+            return
+
+    if method == "ClASSIC":
+        # ------- 第三种方法，早期代码，效果不理想 -------
+        yield from update_ui_lastest_msg("GROBID服务不可用，请检查config中的GROBID_URL。作为替代，现在将执行效果稍差的旧版代码。", chatbot, history, delay=3)
+        yield from 解析PDF_简单拆解(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
+        return
+
+    if method is None:
+        # ------- 以上三种方法都试一遍 -------
+        DOC2X_API_KEY = get_conf("DOC2X_API_KEY")
+        if len(DOC2X_API_KEY) != 0:
+            try:
+                yield from 解析PDF_基于DOC2X(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, DOC2X_API_KEY, user_request)
+                return
+            except:
+                chatbot.append([None, f"DOC2X服务不可用，正在尝试GROBID。{trimmed_format_exc_markdown()}"])
+                yield from update_ui(chatbot=chatbot, history=history)
+        grobid_url = get_avail_grobid_url()
+        if grobid_url is not None:
+            yield from 解析PDF_基于GROBID(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, grobid_url)
+            return
+        yield from update_ui_lastest_msg("GROBID服务不可用，请检查config中的GROBID_URL。作为替代，现在将执行效果稍差的旧版代码。", chatbot, history, delay=3)
+        yield from 解析PDF_简单拆解(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
+        return
+

crazy_functions/PDF_Translate_Wrap.py

@@ -0,0 +1,33 @@
+from crazy_functions.plugin_template.plugin_class_template import GptAcademicPluginTemplate, ArgProperty
+from .PDF_Translate import 批量翻译PDF文档
+
+
+class PDF_Tran(GptAcademicPluginTemplate):
+    def __init__(self):
+        """
+        请注意`execute`会执行在不同的线程中，因此您在定义和使用类变量时，应当慎之又慎！
+        """
+        pass
+
+    def define_arg_selection_menu(self):
+        """
+        定义插件的二级选项菜单
+        """
+        gui_definition = {
+            "main_input":
+                ArgProperty(title="PDF文件路径", description="未指定路径，请上传文件后，再点击该插件", default_value="", type="string").model_dump_json(), # 主输入，自动从输入框同步
+            "additional_prompt":
+                ArgProperty(title="额外提示词", description="例如：对专有名词、翻译语气等方面的要求", default_value="", type="string").model_dump_json(), # 高级参数输入区，自动同步
+            "pdf_parse_method":
+                ArgProperty(title="PDF解析方法", options=["DOC2X", "GROBID", "ClASSIC"], description="无", default_value="GROBID", type="dropdown").model_dump_json(),
+        }
+        return gui_definition
+
+    def execute(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+        """
+        执行插件
+        """
+        main_input = plugin_kwargs["main_input"]
+        additional_prompt = plugin_kwargs["additional_prompt"]
+        pdf_parse_method = plugin_kwargs["pdf_parse_method"]
+        yield from 批量翻译PDF文档(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request)

crazy_functions/Rag_Interface.py

@@ -0,0 +1,95 @@
+from toolbox import CatchException, update_ui, get_conf, get_log_folder, update_ui_lastest_msg
+from crazy_functions.crazy_utils import input_clipping
+from crazy_functions.crazy_utils import request_gpt_model_in_new_thread_with_ui_alive
+
+VECTOR_STORE_TYPE = "Milvus"
+
+if VECTOR_STORE_TYPE == "Milvus":
+    try:
+        from crazy_functions.rag_fns.milvus_worker import MilvusRagWorker as LlamaIndexRagWorker
+    except:
+        VECTOR_STORE_TYPE = "Simple"
+
+if VECTOR_STORE_TYPE == "Simple":
+    from crazy_functions.rag_fns.llama_index_worker import LlamaIndexRagWorker
+
+
+RAG_WORKER_REGISTER = {}
+
+MAX_HISTORY_ROUND = 5
+MAX_CONTEXT_TOKEN_LIMIT = 4096
+REMEMBER_PREVIEW = 1000
+
+@CatchException
+def Rag问答(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+
+    # 1. we retrieve rag worker from global context
+    user_name = chatbot.get_user()
+    checkpoint_dir = get_log_folder(user_name, plugin_name='experimental_rag')
+    if user_name in RAG_WORKER_REGISTER:
+        rag_worker = RAG_WORKER_REGISTER[user_name]
+    else:
+        rag_worker = RAG_WORKER_REGISTER[user_name] = LlamaIndexRagWorker(
+            user_name, 
+            llm_kwargs, 
+            checkpoint_dir=checkpoint_dir, 
+            auto_load_checkpoint=True)
+    current_context = f"{VECTOR_STORE_TYPE} @ {checkpoint_dir}"
+    tip = "提示：输入“清空向量数据库”可以清空RAG向量数据库"
+    if txt == "清空向量数据库":
+        chatbot.append([txt, f'正在清空 ({current_context}) ...'])
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        rag_worker.purge()
+        yield from update_ui_lastest_msg('已清空', chatbot, history, delay=0) # 刷新界面
+        return
+
+    chatbot.append([txt, f'正在召回知识 ({current_context}) ...'])
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+
+    # 2. clip history to reduce token consumption
+    #   2-1. reduce chat round
+    txt_origin = txt
+
+    if len(history) > MAX_HISTORY_ROUND * 2:
+        history = history[-(MAX_HISTORY_ROUND * 2):]
+    txt_clip, history, flags = input_clipping(txt, history, max_token_limit=MAX_CONTEXT_TOKEN_LIMIT, return_clip_flags=True)
+    input_is_clipped_flag = (flags["original_input_len"] != flags["clipped_input_len"])
+
+    #   2-2. if input is clipped, add input to vector store before retrieve
+    if input_is_clipped_flag:
+        yield from update_ui_lastest_msg('检测到长输入, 正在向量化 ...', chatbot, history, delay=0) # 刷新界面
+        # save input to vector store
+        rag_worker.add_text_to_vector_store(txt_origin)
+        yield from update_ui_lastest_msg('向量化完成 ...', chatbot, history, delay=0) # 刷新界面
+        if len(txt_origin) > REMEMBER_PREVIEW:
+            HALF = REMEMBER_PREVIEW//2
+            i_say_to_remember = txt[:HALF] + f" ...\n...(省略{len(txt_origin)-REMEMBER_PREVIEW}字)...\n... " + txt[-HALF:]
+            if (flags["original_input_len"] - flags["clipped_input_len"]) > HALF:
+                txt_clip = txt_clip  + f" ...\n...(省略{len(txt_origin)-len(txt_clip)-HALF}字)...\n... " + txt[-HALF:]
+            else:
+                pass
+            i_say = txt_clip
+        else:
+            i_say_to_remember = i_say = txt_clip
+    else:
+        i_say_to_remember = i_say = txt_clip
+
+    # 3. we search vector store and build prompts
+    nodes = rag_worker.retrieve_from_store_with_query(i_say)
+    prompt = rag_worker.build_prompt(query=i_say, nodes=nodes)
+
+    # 4. it is time to query llms
+    if len(chatbot) != 0: chatbot.pop(-1) # pop temp chat, because we are going to add them again inside `request_gpt_model_in_new_thread_with_ui_alive`
+    model_say = yield from request_gpt_model_in_new_thread_with_ui_alive(
+        inputs=prompt, inputs_show_user=i_say,
+        llm_kwargs=llm_kwargs, chatbot=chatbot, history=history,
+        sys_prompt=system_prompt,
+        retry_times_at_unknown_error=0
+    )
+
+    # 5. remember what has been asked / answered
+    yield from update_ui_lastest_msg(model_say + '</br></br>' + f'对话记忆中, 请稍等 ({current_context}) ...', chatbot, history, delay=0.5) # 刷新界面
+    rag_worker.remember_qa(i_say_to_remember, model_say)
+    history.extend([i_say, model_say])
+
+    yield from update_ui_lastest_msg(model_say, chatbot, history, delay=0, msg=tip) # 刷新界面

crazy_functions/Social_Helper.py

@@ -0,0 +1,167 @@
+import pickle, os, random
+from toolbox import CatchException, update_ui, get_conf, get_log_folder, update_ui_lastest_msg
+from crazy_functions.crazy_utils import input_clipping
+from crazy_functions.crazy_utils import request_gpt_model_in_new_thread_with_ui_alive
+from request_llms.bridge_all import predict_no_ui_long_connection
+from crazy_functions.json_fns.select_tool import structure_output, select_tool
+from pydantic import BaseModel, Field
+from loguru import logger
+from typing import List
+
+
+SOCIAL_NETWOK_WORKER_REGISTER = {}
+
+class SocialNetwork():
+    def __init__(self):
+        self.people = []
+
+class SaveAndLoad():
+    def __init__(self, user_name, llm_kwargs, auto_load_checkpoint=True, checkpoint_dir=None) -> None:
+        self.user_name = user_name
+        self.checkpoint_dir = checkpoint_dir
+        if auto_load_checkpoint:
+            self.social_network = self.load_from_checkpoint(checkpoint_dir)
+        else:
+            self.social_network = SocialNetwork()
+
+    def does_checkpoint_exist(self, checkpoint_dir=None):
+        import os, glob
+        if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        if not os.path.exists(checkpoint_dir): return False
+        if len(glob.glob(os.path.join(checkpoint_dir, "social_network.pkl"))) == 0: return False
+        return True
+
+    def save_to_checkpoint(self, checkpoint_dir=None):
+        if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        with open(os.path.join(checkpoint_dir, 'social_network.pkl'), "wb+") as f:
+            pickle.dump(self.social_network, f)
+        return
+
+    def load_from_checkpoint(self, checkpoint_dir=None):
+        if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        if self.does_checkpoint_exist(checkpoint_dir=checkpoint_dir):
+            with open(os.path.join(checkpoint_dir, 'social_network.pkl'), "rb") as f:
+                social_network = pickle.load(f)
+                return social_network
+        else:
+            return SocialNetwork()
+
+
+class Friend(BaseModel):
+    friend_name: str = Field(description="name of a friend")
+    friend_description: str = Field(description="description of a friend (everything about this friend)")
+    friend_relationship: str = Field(description="The relationship with a friend (e.g. friend, family, colleague)")
+
+class FriendList(BaseModel):
+    friends_list: List[Friend] = Field(description="The list of friends")
+
+
+class SocialNetworkWorker(SaveAndLoad):
+    def ai_socail_advice(self, prompt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, run_gpt_fn, intention_type):
+        pass
+
+    def ai_remove_friend(self, prompt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, run_gpt_fn, intention_type):
+        pass
+
+    def ai_list_friends(self, prompt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, run_gpt_fn, intention_type):
+        pass
+
+    def ai_add_multi_friends(self, prompt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, run_gpt_fn, intention_type):
+        friend, err_msg = structure_output(
+            txt=prompt,
+            prompt="根据提示, 解析多个联系人的身份信息\n\n",
+            err_msg=f"不能理解该联系人",
+            run_gpt_fn=run_gpt_fn,
+            pydantic_cls=FriendList
+        )
+        if friend.friends_list:
+            for f in friend.friends_list: 
+                self.add_friend(f)
+            msg = f"成功添加{len(friend.friends_list)}个联系人: {str(friend.friends_list)}"
+            yield from update_ui_lastest_msg(lastmsg=msg, chatbot=chatbot, history=history, delay=0)
+
+
+    def run(self, txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+        prompt = txt
+        run_gpt_fn = lambda inputs, sys_prompt: predict_no_ui_long_connection(inputs=inputs, llm_kwargs=llm_kwargs, history=[], sys_prompt=sys_prompt, observe_window=[])
+        self.tools_to_select = {
+            "SocialAdvice":{
+                "explain_to_llm": "如果用户希望获取社交指导，调用SocialAdvice生成一些社交建议",
+                "callback": self.ai_socail_advice,
+            },
+            "AddFriends":{
+                "explain_to_llm": "如果用户给出了联系人，调用AddMultiFriends把联系人添加到数据库",
+                "callback": self.ai_add_multi_friends,
+            },
+            "RemoveFriend":{
+                "explain_to_llm": "如果用户希望移除某个联系人，调用RemoveFriend",
+                "callback": self.ai_remove_friend,
+            },
+            "ListFriends":{
+                "explain_to_llm": "如果用户列举联系人，调用ListFriends",
+                "callback": self.ai_list_friends,
+            }
+        }
+
+        try:
+            Explaination = '\n'.join([f'{k}: {v["explain_to_llm"]}' for k, v in self.tools_to_select.items()])
+            class UserSociaIntention(BaseModel):
+                intention_type: str = Field(
+                    description=
+                        f"The type of user intention. You must choose from {self.tools_to_select.keys()}.\n\n" 
+                        f"Explaination:\n{Explaination}", 
+                    default="SocialAdvice"
+                )
+            pydantic_cls_instance, err_msg = select_tool(
+                prompt=txt,
+                run_gpt_fn=run_gpt_fn,
+                pydantic_cls=UserSociaIntention
+            )
+        except Exception as e:
+            yield from update_ui_lastest_msg(
+                lastmsg=f"无法理解用户意图 {err_msg}", 
+                chatbot=chatbot, 
+                history=history, 
+                delay=0
+            )
+            return
+
+        intention_type = pydantic_cls_instance.intention_type
+        intention_callback = self.tools_to_select[pydantic_cls_instance.intention_type]['callback']
+        yield from intention_callback(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, run_gpt_fn, intention_type)
+
+
+    def add_friend(self, friend):
+        # check whether the friend is already in the social network
+        for f in self.social_network.people:
+            if f.friend_name == friend.friend_name:
+                f.friend_description = friend.friend_description
+                f.friend_relationship = friend.friend_relationship
+                logger.info(f"Repeated friend, update info: {friend}")
+                return
+        logger.info(f"Add a new friend: {friend}")
+        self.social_network.people.append(friend)
+        return
+
+
+@CatchException
+def I人助手(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+
+    # 1. we retrieve worker from global context
+    user_name = chatbot.get_user()
+    checkpoint_dir=get_log_folder(user_name, plugin_name='experimental_rag')
+    if user_name in SOCIAL_NETWOK_WORKER_REGISTER:
+        social_network_worker = SOCIAL_NETWOK_WORKER_REGISTER[user_name]
+    else:
+        social_network_worker = SOCIAL_NETWOK_WORKER_REGISTER[user_name] = SocialNetworkWorker(
+            user_name, 
+            llm_kwargs, 
+            checkpoint_dir=checkpoint_dir, 
+            auto_load_checkpoint=True
+        )
+
+    # 2. save
+    yield from social_network_worker.run(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request)
+    social_network_worker.save_to_checkpoint(checkpoint_dir)
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+

crazy_functions/SourceCode_Analyse.py

@@ -1,12 +1,13 @@
-from toolbox import update_ui
-from toolbox import CatchException, report_execption, write_results_to_file
-from .crazy_utils import input_clipping
+from toolbox import update_ui, promote_file_to_downloadzone
+from toolbox import CatchException, report_exception, write_history_to_file
+from shared_utils.fastapi_server import validate_path_safety
+from crazy_functions.crazy_utils import input_clipping
 
 def 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt):
     import os, copy
-    from .crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
-    from .crazy_utils import request_gpt_model_in_new_thread_with_ui_alive
-    msg = '正常'
+    from crazy_functions.crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
+    from crazy_functions.crazy_utils import request_gpt_model_in_new_thread_with_ui_alive
+
     summary_batch_isolation = True
     inputs_array = []
     inputs_show_user_array = []
@@ -22,7 +23,7 @@ def 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs,
             file_content = f.read()
         prefix = "接下来请你逐文件分析下面的工程" if index==0 else ""
         i_say = prefix + f'请对下面的程序文件做一个概述文件名是{os.path.relpath(fp, project_folder)}，文件代码是 ```{file_content}```'
-        i_say_show_user = prefix + f'[{index}/{len(file_manifest)}] 请对下面的程序文件做一个概述: {os.path.abspath(fp)}'
+        i_say_show_user = prefix + f'[{index+1}/{len(file_manifest)}] 请对下面的程序文件做一个概述: {fp}'
         # 装载请求内容
         inputs_array.append(i_say)
         inputs_show_user_array.append(i_say_show_user)
@@ -43,7 +44,8 @@ def 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs,
     # 全部文件解析完成，结果写入文件，准备对工程源代码进行汇总分析
     report_part_1 = copy.deepcopy(gpt_response_collection)
     history_to_return = report_part_1
-    res = write_results_to_file(report_part_1)
+    res = write_history_to_file(report_part_1)
+    promote_file_to_downloadzone(res, chatbot=chatbot)
     chatbot.append(("完成？", "逐个文件分析已完成。" + res + "\n\n正在开始汇总。"))
     yield from update_ui(chatbot=chatbot, history=history_to_return) # 刷新界面
 
@@ -81,7 +83,8 @@ def 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs,
             history=this_iteration_history_feed,   # 迭代之前的分析
             sys_prompt="你是一个程序架构分析师，正在分析一个项目的源代码。" + sys_prompt_additional)
 
-        summary = "请用一句话概括这些文件的整体功能"
+        diagram_code = make_diagram(this_iteration_files, result, this_iteration_history_feed)
+        summary = "请用一句话概括这些文件的整体功能。\n\n" + diagram_code
         summary_result = yield from request_gpt_model_in_new_thread_with_ui_alive(
             inputs=summary,
             inputs_show_user=summary,
@@ -97,73 +100,97 @@ def 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs,
 
     ############################## <END> ##################################
     history_to_return.extend(report_part_2)
-    res = write_results_to_file(history_to_return)
+    res = write_history_to_file(history_to_return)
+    promote_file_to_downloadzone(res, chatbot=chatbot)
     chatbot.append(("完成了吗？", res))
     yield from update_ui(chatbot=chatbot, history=history_to_return) # 刷新界面
 
+def make_diagram(this_iteration_files, result, this_iteration_history_feed):
+    from crazy_functions.diagram_fns.file_tree import build_file_tree_mermaid_diagram
+    return build_file_tree_mermaid_diagram(this_iteration_history_feed[0::2], this_iteration_history_feed[1::2], "项目示意图")
 
 @CatchException
-def 解析项目本身(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 解析项目本身(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     history = []    # 清空历史，以免输入溢出
     import glob
-    file_manifest = [f for f in glob.glob('./*.py') if ('test_project' not in f) and ('gpt_log' not in f)] + \
-                    [f for f in glob.glob('./crazy_functions/*.py') if ('test_project' not in f) and ('gpt_log' not in f)]+ \
-                    [f for f in glob.glob('./request_llm/*.py') if ('test_project' not in f) and ('gpt_log' not in f)]
+    file_manifest = [f for f in glob.glob('./*.py')] + \
+                    [f for f in glob.glob('./*/*.py')]
     project_folder = './'
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何python文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何python文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
 
 @CatchException
-def 解析一个Python项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 解析一个Python项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     history = []    # 清空历史，以免输入溢出
     import glob, os
     if os.path.exists(txt):
         project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.py', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何python文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何python文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
 
-
 @CatchException
-def 解析一个C项目的头文件(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 解析一个Matlab项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     history = []    # 清空历史，以免输入溢出
     import glob, os
     if os.path.exists(txt):
         project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析Matlab项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        return
+    file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.m', recursive=True)]
+    if len(file_manifest) == 0:
+        report_exception(chatbot, history, a = f"解析Matlab项目: {txt}", b = f"找不到任何`.m`源文件: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        return
+    yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
+
+@CatchException
+def 解析一个C项目的头文件(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+    history = []    # 清空历史，以免输入溢出
+    import glob, os
+    if os.path.exists(txt):
+        project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
+    else:
+        if txt == "": txt = '空空如也的输入栏'
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.h', recursive=True)]  + \
                     [f for f in glob.glob(f'{project_folder}/**/*.hpp', recursive=True)] #+ \
                     # [f for f in glob.glob(f'{project_folder}/**/*.c', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.h头文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.h头文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
 
 @CatchException
-def 解析一个C项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 解析一个C项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     history = []    # 清空历史，以免输入溢出
     import glob, os
     if os.path.exists(txt):
         project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.h', recursive=True)]  + \
@@ -171,21 +198,22 @@ def 解析一个C项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system
                     [f for f in glob.glob(f'{project_folder}/**/*.hpp', recursive=True)] + \
                     [f for f in glob.glob(f'{project_folder}/**/*.c', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.h头文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何.h头文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
 
 
 @CatchException
-def 解析一个Java项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 解析一个Java项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     history = []  # 清空历史，以免输入溢出
     import glob, os
     if os.path.exists(txt):
         project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a=f"解析项目: {txt}", b=f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.java', recursive=True)] + \
@@ -193,21 +221,22 @@ def 解析一个Java项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, sys
                     [f for f in glob.glob(f'{project_folder}/**/*.xml', recursive=True)] + \
                     [f for f in glob.glob(f'{project_folder}/**/*.sh', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何java文件: {txt}")
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何java文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
 
 
 @CatchException
-def 解析一个前端项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 解析一个前端项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     history = []  # 清空历史，以免输入溢出
     import glob, os
     if os.path.exists(txt):
         project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a=f"解析项目: {txt}", b=f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.ts', recursive=True)] + \
@@ -222,21 +251,22 @@ def 解析一个前端项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, s
                     [f for f in glob.glob(f'{project_folder}/**/*.css', recursive=True)] + \
                     [f for f in glob.glob(f'{project_folder}/**/*.jsx', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何前端相关文件: {txt}")
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何前端相关文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
 
 
 @CatchException
-def 解析一个Golang项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 解析一个Golang项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     history = []  # 清空历史，以免输入溢出
     import glob, os
     if os.path.exists(txt):
         project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a=f"解析项目: {txt}", b=f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.go', recursive=True)] + \
@@ -244,40 +274,42 @@ def 解析一个Golang项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, s
                     [f for f in glob.glob(f'{project_folder}/**/go.sum', recursive=True)] + \
                     [f for f in glob.glob(f'{project_folder}/**/go.work', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何golang文件: {txt}")
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何golang文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
 
 @CatchException
-def 解析一个Rust项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 解析一个Rust项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     history = []  # 清空历史，以免输入溢出
     import glob, os
     if os.path.exists(txt):
         project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a=f"解析项目: {txt}", b=f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.rs', recursive=True)] + \
                     [f for f in glob.glob(f'{project_folder}/**/*.toml', recursive=True)] + \
                     [f for f in glob.glob(f'{project_folder}/**/*.lock', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何golang文件: {txt}")
+        report_exception(chatbot, history, a=f"解析项目: {txt}", b=f"找不到任何golang文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
 
 @CatchException
-def 解析一个Lua项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 解析一个Lua项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     history = []    # 清空历史，以免输入溢出
     import glob, os
     if os.path.exists(txt):
         project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.lua', recursive=True)] + \
@@ -285,34 +317,35 @@ def 解析一个Lua项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, syst
                     [f for f in glob.glob(f'{project_folder}/**/*.json', recursive=True)] + \
                     [f for f in glob.glob(f'{project_folder}/**/*.toml', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何lua文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何lua文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
 
 
 @CatchException
-def 解析一个CSharp项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 解析一个CSharp项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     history = []    # 清空历史，以免输入溢出
     import glob, os
     if os.path.exists(txt):
         project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.cs', recursive=True)] + \
                     [f for f in glob.glob(f'{project_folder}/**/*.csproj', recursive=True)]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何CSharp文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何CSharp文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)
 
 
 @CatchException
-def 解析任意code项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 解析任意code项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     txt_pattern = plugin_kwargs.get("advanced_arg")
     txt_pattern = txt_pattern.replace("，", ",")
     # 将要匹配的模式(例如: *.c, *.cpp, *.py, config.toml)
@@ -322,18 +355,22 @@ def 解析任意code项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, sys
     pattern_except_suffix = [_.lstrip(" ^*.,").rstrip(" ,") for _ in txt_pattern.split(" ") if _ != "" and _.strip().startswith("^*.")]
     pattern_except_suffix += ['zip', 'rar', '7z', 'tar', 'gz'] # 避免解析压缩文件
     # 将要忽略匹配的文件名(例如: ^README.md)
-    pattern_except_name = [_.lstrip(" ^*,").rstrip(" ,").replace(".", "\.") for _ in txt_pattern.split(" ") if _ != "" and _.strip().startswith("^") and not _.strip().startswith("^*.")]
+    pattern_except_name = [_.lstrip(" ^*,").rstrip(" ,").replace(".", r"\.") # 移除左边通配符，移除右侧逗号，转义点号
+                           for _ in txt_pattern.split(" ") # 以空格分割
+                           if (_ != "" and _.strip().startswith("^") and not _.strip().startswith("^*."))   # ^开始，但不是^*.开始
+                           ]
     # 生成正则表达式
-    pattern_except = '/[^/]+\.(' + "|".join(pattern_except_suffix) + ')$'
+    pattern_except = r'/[^/]+\.(' + "|".join(pattern_except_suffix) + ')$'
     pattern_except += '|/(' + "|".join(pattern_except_name) + ')$' if pattern_except_name != [] else ''
 
     history.clear()
     import glob, os, re
     if os.path.exists(txt):
         project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
     else:
         if txt == "": txt = '空空如也的输入栏'
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     # 若上传压缩文件, 先寻找到解压的文件夹路径, 从而避免解析压缩文件
@@ -346,7 +383,7 @@ def 解析任意code项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, sys
     file_manifest = [f for pattern in pattern_include for f in glob.glob(f'{extract_folder_path}/**/{pattern}', recursive=True) if "" != extract_folder_path and \
                       os.path.isfile(f) and (not re.search(pattern_except, f) or pattern.endswith('.' + re.search(pattern_except, f).group().split('.')[-1]))]
     if len(file_manifest) == 0:
-        report_execption(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何文件: {txt}")
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何文件: {txt}")
         yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
         return
     yield from 解析源代码新(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)

crazy_functions/SourceCode_Comment.py

@@ -0,0 +1,138 @@
+import os, copy, time
+from toolbox import CatchException, report_exception, update_ui, zip_result, promote_file_to_downloadzone, update_ui_lastest_msg, get_conf, generate_file_link
+from shared_utils.fastapi_server import validate_path_safety
+from crazy_functions.crazy_utils import input_clipping
+from crazy_functions.crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
+from crazy_functions.crazy_utils import request_gpt_model_in_new_thread_with_ui_alive
+from crazy_functions.agent_fns.python_comment_agent import PythonCodeComment
+from crazy_functions.diagram_fns.file_tree import FileNode
+from shared_utils.advanced_markdown_format import markdown_convertion_for_file
+
+def 注释源代码(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt):
+
+    summary_batch_isolation = True
+    inputs_array = []
+    inputs_show_user_array = []
+    history_array = []
+    sys_prompt_array = []
+
+    assert len(file_manifest) <= 512, "源文件太多（超过512个）, 请缩减输入文件的数量。或者，您也可以选择删除此行警告，并修改代码拆分file_manifest列表，从而实现分批次处理。"
+
+    # 建立文件树
+    file_tree_struct = FileNode("root", build_manifest=True)
+    for file_path in file_manifest:
+        file_tree_struct.add_file(file_path, file_path)
+
+    # <第一步，逐个文件分析，多线程>
+    for index, fp in enumerate(file_manifest):
+        # 读取文件
+        with open(fp, 'r', encoding='utf-8', errors='replace') as f:
+            file_content = f.read()
+        prefix = ""
+        i_say = prefix + f'Please conclude the following source code at {os.path.relpath(fp, project_folder)} with only one sentence, the code is:\n```{file_content}```'
+        i_say_show_user = prefix + f'[{index+1}/{len(file_manifest)}] 请用一句话对下面的程序文件做一个整体概述: {fp}'
+        # 装载请求内容
+        MAX_TOKEN_SINGLE_FILE = 2560
+        i_say, _ = input_clipping(inputs=i_say, history=[], max_token_limit=MAX_TOKEN_SINGLE_FILE)
+        inputs_array.append(i_say)
+        inputs_show_user_array.append(i_say_show_user)
+        history_array.append([])
+        sys_prompt_array.append("You are a software architecture analyst analyzing a source code project. Do not dig into details, tell me what the code is doing in general. Your answer must be short, simple and clear.")
+    # 文件读取完成，对每一个源代码文件，生成一个请求线程，发送到大模型进行分析
+    gpt_response_collection = yield from request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
+        inputs_array = inputs_array,
+        inputs_show_user_array = inputs_show_user_array,
+        history_array = history_array,
+        sys_prompt_array = sys_prompt_array,
+        llm_kwargs = llm_kwargs,
+        chatbot = chatbot,
+        show_user_at_complete = True
+    )
+
+    # <第二步，逐个文件分析，生成带注释文件>
+    from concurrent.futures import ThreadPoolExecutor
+    executor = ThreadPoolExecutor(max_workers=get_conf('DEFAULT_WORKER_NUM'))
+    def _task_multi_threading(i_say, gpt_say, fp, file_tree_struct):
+        pcc = PythonCodeComment(llm_kwargs, language='English')
+        pcc.read_file(path=fp, brief=gpt_say)
+        revised_path, revised_content = pcc.begin_comment_source_code(None, None)
+        file_tree_struct.manifest[fp].revised_path = revised_path
+        file_tree_struct.manifest[fp].revised_content = revised_content
+        # <将结果写回源文件>
+        with open(fp, 'w', encoding='utf-8') as f:
+            f.write(file_tree_struct.manifest[fp].revised_content)
+        # <生成对比html>
+        with open("crazy_functions/agent_fns/python_comment_compare.html", 'r', encoding='utf-8') as f:
+            html_template = f.read()
+        warp = lambda x: "```python\n\n" + x + "\n\n```"
+        from themes.theme import advanced_css
+        html_template = html_template.replace("ADVANCED_CSS", advanced_css)
+        html_template = html_template.replace("REPLACE_CODE_FILE_LEFT", pcc.get_markdown_block_in_html(markdown_convertion_for_file(warp(pcc.original_content))))
+        html_template = html_template.replace("REPLACE_CODE_FILE_RIGHT", pcc.get_markdown_block_in_html(markdown_convertion_for_file(warp(revised_content))))
+        compare_html_path = fp + '.compare.html'
+        file_tree_struct.manifest[fp].compare_html = compare_html_path
+        with open(compare_html_path, 'w', encoding='utf-8') as f:
+            f.write(html_template)
+        # print('done 1')
+
+    chatbot.append([None, f"正在处理:"])
+    futures = []
+    for i_say, gpt_say, fp in zip(gpt_response_collection[0::2], gpt_response_collection[1::2], file_manifest):
+        future = executor.submit(_task_multi_threading, i_say, gpt_say, fp, file_tree_struct)
+        futures.append(future)
+
+    cnt = 0
+    while True:
+        cnt += 1
+        time.sleep(3)
+        worker_done = [h.done() for h in futures]
+        remain = len(worker_done) - sum(worker_done)
+
+        # <展示已经完成的部分>
+        preview_html_list = []
+        for done, fp in zip(worker_done, file_manifest):
+            if not done: continue
+            preview_html_list.append(file_tree_struct.manifest[fp].compare_html)
+        file_links = generate_file_link(preview_html_list)
+
+        yield from update_ui_lastest_msg(
+            f"剩余源文件数量: {remain}.\n\n" + 
+            f"已完成的文件: {sum(worker_done)}.\n\n" + 
+            file_links +
+            "\n\n" +
+            ''.join(['.']*(cnt % 10 + 1)
+        ), chatbot=chatbot, history=history, delay=0)
+        yield from update_ui(chatbot=chatbot, history=[]) # 刷新界面
+        if all(worker_done):
+            executor.shutdown()
+            break
+
+    # <第四步，压缩结果>
+    zip_res = zip_result(project_folder)
+    promote_file_to_downloadzone(file=zip_res, chatbot=chatbot)
+
+    # <END>
+    chatbot.append((None, "所有源文件均已处理完毕。"))
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+
+
+
+@CatchException
+def 注释Python项目(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+    history = []    # 清空历史，以免输入溢出
+    import glob, os
+    if os.path.exists(txt):
+        project_folder = txt
+        validate_path_safety(project_folder, chatbot.get_user())
+    else:
+        if txt == "": txt = '空空如也的输入栏'
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到本地项目或无权访问: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        return
+    file_manifest = [f for f in glob.glob(f'{project_folder}/**/*.py', recursive=True)]
+    if len(file_manifest) == 0:
+        report_exception(chatbot, history, a = f"解析项目: {txt}", b = f"找不到任何python文件: {txt}")
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        return
+
+    yield from 注释源代码(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt)

crazy_functions/agent_fns/auto_agent.py

@@ -0,0 +1,23 @@
+from toolbox import CatchException, update_ui, gen_time_str, trimmed_format_exc, ProxyNetworkActivate
+from toolbox import report_exception, get_log_folder, update_ui_lastest_msg, Singleton
+from crazy_functions.agent_fns.pipe import PluginMultiprocessManager, PipeCom
+from crazy_functions.agent_fns.general import AutoGenGeneral
+
+
+
+class AutoGenMath(AutoGenGeneral):
+
+    def define_agents(self):
+        from autogen import AssistantAgent, UserProxyAgent
+        return [
+            {
+                "name": "assistant",            # name of the agent.
+                "cls":  AssistantAgent,         # class of the agent.
+            },
+            {
+                "name": "user_proxy",           # name of the agent.
+                "cls":  UserProxyAgent,         # class of the agent.
+                "human_input_mode": "ALWAYS",   # always ask for human input.
+                "llm_config": False,            # disables llm-based auto reply.
+            },
+        ]

crazy_functions/agent_fns/echo_agent.py

@@ -0,0 +1,20 @@
+from crazy_functions.agent_fns.pipe import PluginMultiprocessManager, PipeCom
+from loguru import logger
+
+class EchoDemo(PluginMultiprocessManager):
+    def subprocess_worker(self, child_conn):
+        # ⭐⭐ 子进程
+        self.child_conn = child_conn
+        while True:
+            msg = self.child_conn.recv() # PipeCom
+            if msg.cmd == "user_input":
+                # wait futher user input
+                self.child_conn.send(PipeCom("show", msg.content))
+                wait_success = self.subprocess_worker_wait_user_feedback(wait_msg="我准备好处理下一个问题了.")
+                if not wait_success:
+                    # wait timeout, terminate this subprocess_worker
+                    break
+            elif msg.cmd == "terminate":
+                self.child_conn.send(PipeCom("done", ""))
+                break
+        logger.info('[debug] subprocess_worker terminated')

crazy_functions/agent_fns/general.py

@@ -0,0 +1,138 @@
+from toolbox import trimmed_format_exc, get_conf, ProxyNetworkActivate
+from crazy_functions.agent_fns.pipe import PluginMultiprocessManager, PipeCom
+from request_llms.bridge_all import predict_no_ui_long_connection
+import time
+
+def gpt_academic_generate_oai_reply(
+    self,
+    messages,
+    sender,
+    config,
+):
+    llm_config = self.llm_config if config is None else config
+    if llm_config is False:
+        return False, None
+    if messages is None:
+        messages = self._oai_messages[sender]
+
+    inputs = messages[-1]['content']
+    history = []
+    for message in messages[:-1]:
+        history.append(message['content'])
+    context=messages[-1].pop("context", None)
+    assert context is None, "预留参数 context 未实现"
+
+    reply = predict_no_ui_long_connection(
+        inputs=inputs,
+        llm_kwargs=llm_config,
+        history=history,
+        sys_prompt=self._oai_system_message[0]['content'],
+        console_slience=True
+    )
+    assumed_done = reply.endswith('\nTERMINATE')
+    return True, reply
+
+class AutoGenGeneral(PluginMultiprocessManager):
+    def gpt_academic_print_override(self, user_proxy, message, sender):
+        # ⭐⭐ run in subprocess
+        try:
+            print_msg = sender.name + "\n\n---\n\n" + message["content"]
+        except:
+            print_msg = sender.name + "\n\n---\n\n" + message
+        self.child_conn.send(PipeCom("show", print_msg))
+
+    def gpt_academic_get_human_input(self, user_proxy, message):
+        # ⭐⭐ run in subprocess
+        patience = 300
+        begin_waiting_time = time.time()
+        self.child_conn.send(PipeCom("interact", message))
+        while True:
+            time.sleep(0.5)
+            if self.child_conn.poll():
+                wait_success = True
+                break
+            if time.time() - begin_waiting_time > patience:
+                self.child_conn.send(PipeCom("done", ""))
+                wait_success = False
+                break
+        if wait_success:
+            return self.child_conn.recv().content
+        else:
+            raise TimeoutError("等待用户输入超时")
+
+    def define_agents(self):
+        raise NotImplementedError
+
+    def exe_autogen(self, input):
+        # ⭐⭐ run in subprocess
+        input = input.content
+        code_execution_config = {"work_dir": self.autogen_work_dir, "use_docker": self.use_docker}
+        agents = self.define_agents()
+        user_proxy = None
+        assistant = None
+        for agent_kwargs in agents:
+            agent_cls = agent_kwargs.pop('cls')
+            kwargs = {
+                'llm_config':self.llm_kwargs,
+                'code_execution_config':code_execution_config
+            }
+            kwargs.update(agent_kwargs)
+            agent_handle = agent_cls(**kwargs)
+            agent_handle._print_received_message = lambda a,b: self.gpt_academic_print_override(agent_kwargs, a, b)
+            for d in agent_handle._reply_func_list:
+                if hasattr(d['reply_func'],'__name__') and d['reply_func'].__name__ == 'generate_oai_reply':
+                    d['reply_func'] = gpt_academic_generate_oai_reply
+            if agent_kwargs['name'] == 'user_proxy':
+                agent_handle.get_human_input = lambda a: self.gpt_academic_get_human_input(user_proxy, a)
+                user_proxy = agent_handle
+            if agent_kwargs['name'] == 'assistant': assistant = agent_handle
+        try:
+            if user_proxy is None or assistant is None: raise Exception("用户代理或助理代理未定义")
+            with ProxyNetworkActivate("AutoGen"):
+                user_proxy.initiate_chat(assistant, message=input)
+        except Exception as e:
+            tb_str = '```\n' + trimmed_format_exc() + '```'
+            self.child_conn.send(PipeCom("done", "AutoGen 执行失败: \n\n" + tb_str))
+
+    def subprocess_worker(self, child_conn):
+        # ⭐⭐ run in subprocess
+        self.child_conn = child_conn
+        while True:
+            msg = self.child_conn.recv()  # PipeCom
+            self.exe_autogen(msg)
+
+
+class AutoGenGroupChat(AutoGenGeneral):
+    def exe_autogen(self, input):
+        # ⭐⭐ run in subprocess
+        import autogen
+
+        input = input.content
+        with ProxyNetworkActivate("AutoGen"):
+            code_execution_config = {"work_dir": self.autogen_work_dir, "use_docker": self.use_docker}
+            agents = self.define_agents()
+            agents_instances = []
+            for agent_kwargs in agents:
+                agent_cls = agent_kwargs.pop("cls")
+                kwargs = {"code_execution_config": code_execution_config}
+                kwargs.update(agent_kwargs)
+                agent_handle = agent_cls(**kwargs)
+                agent_handle._print_received_message = lambda a, b: self.gpt_academic_print_override(agent_kwargs, a, b)
+                agents_instances.append(agent_handle)
+                if agent_kwargs["name"] == "user_proxy":
+                    user_proxy = agent_handle
+                    user_proxy.get_human_input = lambda a: self.gpt_academic_get_human_input(user_proxy, a)
+            try:
+                groupchat = autogen.GroupChat(agents=agents_instances, messages=[], max_round=50)
+                manager = autogen.GroupChatManager(groupchat=groupchat, **self.define_group_chat_manager_config())
+                manager._print_received_message = lambda a, b: self.gpt_academic_print_override(agent_kwargs, a, b)
+                manager.get_human_input = lambda a: self.gpt_academic_get_human_input(manager, a)
+                if user_proxy is None:
+                    raise Exception("user_proxy is not defined")
+                user_proxy.initiate_chat(manager, message=input)
+            except Exception:
+                tb_str = "```\n" + trimmed_format_exc() + "```"
+                self.child_conn.send(PipeCom("done", "AutoGen exe failed: \n\n" + tb_str))
+
+    def define_group_chat_manager_config(self):
+        raise NotImplementedError

crazy_functions/agent_fns/persistent.py

@@ -0,0 +1,16 @@
+from toolbox import Singleton
+@Singleton
+class GradioMultiuserManagerForPersistentClasses():
+    def __init__(self):
+        self.mapping = {}
+
+    def already_alive(self, key):
+        return (key in self.mapping) and (self.mapping[key].is_alive())
+
+    def set(self, key, x):
+        self.mapping[key] = x
+        return self.mapping[key]
+
+    def get(self, key):
+        return self.mapping[key]
+

crazy_functions/agent_fns/pipe.py

@@ -0,0 +1,195 @@
+from toolbox import get_log_folder, update_ui, gen_time_str, get_conf, promote_file_to_downloadzone
+from crazy_functions.agent_fns.watchdog import WatchDog
+from loguru import logger
+import time, os
+
+class PipeCom:
+    def __init__(self, cmd, content) -> None:
+        self.cmd = cmd
+        self.content = content
+
+
+class PluginMultiprocessManager:
+    def __init__(self, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+        # ⭐ run in main process
+        self.autogen_work_dir = os.path.join(get_log_folder("autogen"), gen_time_str())
+        self.previous_work_dir_files = {}
+        self.llm_kwargs = llm_kwargs
+        self.plugin_kwargs = plugin_kwargs
+        self.chatbot = chatbot
+        self.history = history
+        self.system_prompt = system_prompt
+        # self.user_request = user_request
+        self.alive = True
+        self.use_docker = get_conf("AUTOGEN_USE_DOCKER")
+        self.last_user_input = ""
+        # create a thread to monitor self.heartbeat, terminate the instance if no heartbeat for a long time
+        timeout_seconds = 5 * 60
+        self.heartbeat_watchdog = WatchDog(timeout=timeout_seconds, bark_fn=self.terminate, interval=5)
+        self.heartbeat_watchdog.begin_watch()
+
+    def feed_heartbeat_watchdog(self):
+        # feed this `dog`, so the dog will not `bark` (bark_fn will terminate the instance)
+        self.heartbeat_watchdog.feed()
+
+    def is_alive(self):
+        return self.alive
+
+    def launch_subprocess_with_pipe(self):
+        # ⭐ run in main process
+        from multiprocessing import Process, Pipe
+
+        parent_conn, child_conn = Pipe()
+        self.p = Process(target=self.subprocess_worker, args=(child_conn,))
+        self.p.daemon = True
+        self.p.start()
+        return parent_conn
+
+    def terminate(self):
+        self.p.terminate()
+        self.alive = False
+        logger.info("[debug] instance terminated")
+
+    def subprocess_worker(self, child_conn):
+        # ⭐⭐ run in subprocess
+        raise NotImplementedError
+
+    def send_command(self, cmd):
+        # ⭐ run in main process
+        repeated = False
+        if cmd == self.last_user_input:
+            repeated = True
+            cmd = ""
+        else:
+            self.last_user_input = cmd
+        self.parent_conn.send(PipeCom("user_input", cmd))
+        return repeated, cmd
+
+    def immediate_showoff_when_possible(self, fp):
+        # ⭐ 主进程
+        # 获取fp的拓展名
+        file_type = fp.split('.')[-1]
+        # 如果是文本文件, 则直接显示文本内容
+        if file_type.lower() in ['png', 'jpg']:
+            image_path = os.path.abspath(fp)
+            self.chatbot.append([
+                '检测到新生图像:',
+                f'本地文件预览: <br/><div align="center"><img src="file={image_path}"></div>'
+            ])
+            yield from update_ui(chatbot=self.chatbot, history=self.history)
+
+    def overwatch_workdir_file_change(self):
+        # ⭐ 主进程 Docker 外挂文件夹监控
+        path_to_overwatch = self.autogen_work_dir
+        change_list = []
+        # 扫描路径下的所有文件, 并与self.previous_work_dir_files中所记录的文件进行对比，
+        # 如果有新文件出现，或者文件的修改时间发生变化，则更新self.previous_work_dir_files中
+        # 把新文件和发生变化的文件的路径记录到 change_list 中
+        for root, dirs, files in os.walk(path_to_overwatch):
+            for file in files:
+                file_path = os.path.join(root, file)
+                if file_path not in self.previous_work_dir_files.keys():
+                    last_modified_time = os.stat(file_path).st_mtime
+                    self.previous_work_dir_files.update({file_path: last_modified_time})
+                    change_list.append(file_path)
+                else:
+                    last_modified_time = os.stat(file_path).st_mtime
+                    if last_modified_time != self.previous_work_dir_files[file_path]:
+                        self.previous_work_dir_files[file_path] = last_modified_time
+                        change_list.append(file_path)
+        if len(change_list) > 0:
+            file_links = ""
+            for f in change_list:
+                res = promote_file_to_downloadzone(f)
+                file_links += f'<br/><a href="file={res}" target="_blank">{res}</a>'
+                yield from self.immediate_showoff_when_possible(f)
+
+            self.chatbot.append(['检测到新生文档.', f'文档清单如下: {file_links}'])
+            yield from update_ui(chatbot=self.chatbot, history=self.history)
+        return change_list
+
+
+    def main_process_ui_control(self, txt, create_or_resume) -> str:
+        # ⭐ 主进程
+        if create_or_resume == 'create':
+            self.cnt = 1
+            self.parent_conn = self.launch_subprocess_with_pipe() # ⭐⭐⭐
+        repeated, cmd_to_autogen = self.send_command(txt)
+        if txt == 'exit':
+            self.chatbot.append([f"结束", "结束信号已明确，终止AutoGen程序。"])
+            yield from update_ui(chatbot=self.chatbot, history=self.history)
+            self.terminate()
+            return "terminate"
+
+        # patience = 10
+
+        while True:
+            time.sleep(0.5)
+            if not self.alive:
+                # the heartbeat watchdog might have it killed
+                self.terminate()
+                return "terminate"
+            if self.parent_conn.poll():
+                self.feed_heartbeat_watchdog()
+                if "[GPT-Academic] 等待中" in self.chatbot[-1][-1]:
+                    self.chatbot.pop(-1)  # remove the last line
+                if "等待您的进一步指令" in self.chatbot[-1][-1]:
+                    self.chatbot.pop(-1)  # remove the last line
+                if '[GPT-Academic] 等待中' in self.chatbot[-1][-1]:
+                    self.chatbot.pop(-1)    # remove the last line
+                msg = self.parent_conn.recv() # PipeCom
+                if msg.cmd == "done":
+                    self.chatbot.append([f"结束", msg.content])
+                    self.cnt += 1
+                    yield from update_ui(chatbot=self.chatbot, history=self.history)
+                    self.terminate()
+                    break
+                if msg.cmd == "show":
+                    yield from self.overwatch_workdir_file_change()
+                    notice = ""
+                    if repeated: notice = "（自动忽略重复的输入）"
+                    self.chatbot.append([f"运行阶段-{self.cnt}（上次用户反馈输入为: 「{cmd_to_autogen}」{notice}", msg.content])
+                    self.cnt += 1
+                    yield from update_ui(chatbot=self.chatbot, history=self.history)
+                if msg.cmd == "interact":
+                    yield from self.overwatch_workdir_file_change()
+                    self.chatbot.append([f"程序抵达用户反馈节点.", msg.content +
+                                         "\n\n等待您的进一步指令." +
+                                         "\n\n(1) 一般情况下您不需要说什么, 清空输入区, 然后直接点击“提交”以继续. " +
+                                         "\n\n(2) 如果您需要补充些什么, 输入要反馈的内容, 直接点击“提交”以继续. " +
+                                         "\n\n(3) 如果您想终止程序, 输入exit, 直接点击“提交”以终止AutoGen并解锁. "
+                    ])
+                    yield from update_ui(chatbot=self.chatbot, history=self.history)
+                    # do not terminate here, leave the subprocess_worker instance alive
+                    return "wait_feedback"
+            else:
+                self.feed_heartbeat_watchdog()
+                if '[GPT-Academic] 等待中' not in self.chatbot[-1][-1]:
+                    # begin_waiting_time = time.time()
+                    self.chatbot.append(["[GPT-Academic] 等待AutoGen执行结果 ...", "[GPT-Academic] 等待中"])
+                self.chatbot[-1] = [self.chatbot[-1][0], self.chatbot[-1][1].replace("[GPT-Academic] 等待中", "[GPT-Academic] 等待中.")]
+                yield from update_ui(chatbot=self.chatbot, history=self.history)
+                # if time.time() - begin_waiting_time > patience:
+                #     self.chatbot.append([f"结束", "等待超时, 终止AutoGen程序。"])
+                #     yield from update_ui(chatbot=self.chatbot, history=self.history)
+                #     self.terminate()
+                #     return "terminate"
+
+        self.terminate()
+        return "terminate"
+
+    def subprocess_worker_wait_user_feedback(self, wait_msg="wait user feedback"):
+        # ⭐⭐ run in subprocess
+        patience = 5 * 60
+        begin_waiting_time = time.time()
+        self.child_conn.send(PipeCom("interact", wait_msg))
+        while True:
+            time.sleep(0.5)
+            if self.child_conn.poll():
+                wait_success = True
+                break
+            if time.time() - begin_waiting_time > patience:
+                self.child_conn.send(PipeCom("done", ""))
+                wait_success = False
+                break
+        return wait_success

crazy_functions/agent_fns/python_comment_agent.py

@@ -0,0 +1,393 @@
+import datetime
+import re
+import os
+from loguru import logger
+from textwrap import dedent
+from toolbox import CatchException, update_ui
+from request_llms.bridge_all import predict_no_ui_long_connection
+from crazy_functions.crazy_utils import request_gpt_model_in_new_thread_with_ui_alive
+
+# TODO: 解决缩进问题
+
+find_function_end_prompt = '''
+Below is a page of code that you need to read. This page may not yet complete, you job is to split this page to sperate functions, class functions etc.
+- Provide the line number where the first visible function ends.
+- Provide the line number where the next visible function begins.
+- If there are no other functions in this page, you should simply return the line number of the last line.
+- Only focus on functions declared by `def` keyword. Ignore inline functions. Ignore function calls.
+
+------------------ Example ------------------
+INPUT:
+
+    ```
+    L0000 |import sys
+    L0001 |import re
+    L0002 |
+    L0003 |def trimmed_format_exc():
+    L0004 |    import os
+    L0005 |    import traceback
+    L0006 |    str = traceback.format_exc()
+    L0007 |    current_path = os.getcwd()
+    L0008 |    replace_path = "."
+    L0009 |    return str.replace(current_path, replace_path)
+    L0010 |
+    L0011 |
+    L0012 |def trimmed_format_exc_markdown():
+    L0013 |    ...
+    L0014 |    ...
+    ```
+
+OUTPUT:
+
+    ```
+    <first_function_end_at>L0009</first_function_end_at>
+    <next_function_begin_from>L0012</next_function_begin_from>
+    ```
+
+------------------ End of Example ------------------
+
+
+------------------ the real INPUT you need to process NOW ------------------
+```
+{THE_TAGGED_CODE}
+```
+'''
+
+
+
+
+
+
+
+revise_funtion_prompt = '''
+You need to read the following code, and revise the source code ({FILE_BASENAME}) according to following instructions:
+1. You should analyze the purpose of the functions (if there are any).
+2. You need to add docstring for the provided functions (if there are any).
+
+Be aware:
+1. You must NOT modify the indent of code.
+2. You are NOT authorized to change or translate non-comment code, and you are NOT authorized to add empty lines either, toggle qu.
+3. Use {LANG} to add comments and docstrings. Do NOT translate Chinese that is already in the code.
+
+------------------ Example ------------------
+INPUT:
+```
+L0000 |
+L0001 |def zip_result(folder):
+L0002 |    t = gen_time_str()
+L0003 |    zip_folder(folder, get_log_folder(), f"result.zip")
+L0004 |    return os.path.join(get_log_folder(), f"result.zip")
+L0005 |
+L0006 |
+```
+
+OUTPUT:
+
+<instruction_1_purpose>
+This function compresses a given folder, and return the path of the resulting `zip` file.
+</instruction_1_purpose>
+<instruction_2_revised_code>
+```
+def zip_result(folder):
+    """
+    Compresses the specified folder into a zip file and stores it in the log folder.
+
+    Args:
+        folder (str): The path to the folder that needs to be compressed.
+
+    Returns:
+        str: The path to the created zip file in the log folder.
+    """
+    t = gen_time_str()
+    zip_folder(folder, get_log_folder(), f"result.zip")  # ⭐ Execute the zipping of folder
+    return os.path.join(get_log_folder(), f"result.zip")
+```
+</instruction_2_revised_code>
+------------------ End of Example ------------------
+
+
+------------------ the real INPUT you need to process NOW ({FILE_BASENAME}) ------------------
+```
+{THE_CODE}
+```
+{INDENT_REMINDER}
+{BRIEF_REMINDER}
+{HINT_REMINDER}
+'''
+
+
+
+class PythonCodeComment():
+
+    def __init__(self, llm_kwargs, language) -> None:
+        self.original_content = ""
+        self.full_context = []
+        self.full_context_with_line_no = []
+        self.current_page_start = 0
+        self.page_limit = 100 # 100 lines of code each page
+        self.ignore_limit = 20
+        self.llm_kwargs = llm_kwargs
+        self.language = language
+        self.path = None
+        self.file_basename = None
+        self.file_brief = ""
+
+    def generate_tagged_code_from_full_context(self):
+        for i, code in enumerate(self.full_context):
+            number = i
+            padded_number = f"{number:04}"
+            result = f"L{padded_number}"
+            self.full_context_with_line_no.append(f"{result} | {code}")
+        return self.full_context_with_line_no
+
+    def read_file(self, path, brief):
+        with open(path, 'r', encoding='utf8') as f:
+            self.full_context = f.readlines()
+        self.original_content = ''.join(self.full_context)
+        self.file_basename = os.path.basename(path)
+        self.file_brief = brief
+        self.full_context_with_line_no = self.generate_tagged_code_from_full_context()
+        self.path = path
+
+    def find_next_function_begin(self, tagged_code:list, begin_and_end):
+        begin, end = begin_and_end
+        THE_TAGGED_CODE = ''.join(tagged_code)
+        self.llm_kwargs['temperature'] = 0
+        result = predict_no_ui_long_connection(
+            inputs=find_function_end_prompt.format(THE_TAGGED_CODE=THE_TAGGED_CODE),
+            llm_kwargs=self.llm_kwargs,
+            history=[],
+            sys_prompt="",
+            observe_window=[],
+            console_slience=True
+        )
+
+        def extract_number(text):
+            # 使用正则表达式匹配模式
+            match = re.search(r'<next_function_begin_from>L(\d+)</next_function_begin_from>', text)
+            if match:
+                # 提取匹配的数字部分并转换为整数
+                return int(match.group(1))
+            return None
+
+        line_no = extract_number(result)
+        if line_no is not None:
+            return line_no
+        else:
+            return end
+
+    def _get_next_window(self):
+        #
+        current_page_start = self.current_page_start
+
+        if self.current_page_start == len(self.full_context) + 1:
+            raise StopIteration
+
+        # 如果剩余的行数非常少，一鼓作气处理掉
+        if len(self.full_context) - self.current_page_start < self.ignore_limit:
+            future_page_start = len(self.full_context) + 1
+            self.current_page_start = future_page_start
+            return current_page_start, future_page_start
+
+
+        tagged_code = self.full_context_with_line_no[ self.current_page_start: self.current_page_start + self.page_limit]
+        line_no = self.find_next_function_begin(tagged_code, [self.current_page_start, self.current_page_start + self.page_limit])
+
+        if line_no > len(self.full_context) - 5:
+            line_no = len(self.full_context) + 1
+
+        future_page_start = line_no
+        self.current_page_start = future_page_start
+
+        # ! consider eof
+        return current_page_start, future_page_start
+
+    def dedent(self, text):
+        """Remove any common leading whitespace from every line in `text`.
+        """
+        # Look for the longest leading string of spaces and tabs common to
+        # all lines.
+        margin = None
+        _whitespace_only_re = re.compile('^[ \t]+$', re.MULTILINE)
+        _leading_whitespace_re = re.compile('(^[ \t]*)(?:[^ \t\n])', re.MULTILINE)
+        text = _whitespace_only_re.sub('', text)
+        indents = _leading_whitespace_re.findall(text)
+        for indent in indents:
+            if margin is None:
+                margin = indent
+
+            # Current line more deeply indented than previous winner:
+            # no change (previous winner is still on top).
+            elif indent.startswith(margin):
+                pass
+
+            # Current line consistent with and no deeper than previous winner:
+            # it's the new winner.
+            elif margin.startswith(indent):
+                margin = indent
+
+            # Find the largest common whitespace between current line and previous
+            # winner.
+            else:
+                for i, (x, y) in enumerate(zip(margin, indent)):
+                    if x != y:
+                        margin = margin[:i]
+                        break
+
+        # sanity check (testing/debugging only)
+        if 0 and margin:
+            for line in text.split("\n"):
+                assert not line or line.startswith(margin), \
+                    "line = %r, margin = %r" % (line, margin)
+
+        if margin:
+            text = re.sub(r'(?m)^' + margin, '', text)
+            return text, len(margin)
+        else:
+            return text, 0
+
+    def get_next_batch(self):
+        current_page_start, future_page_start = self._get_next_window()
+        return ''.join(self.full_context[current_page_start: future_page_start]), current_page_start, future_page_start
+
+    def tag_code(self, fn, hint):
+        code = fn
+        _, n_indent = self.dedent(code)
+        indent_reminder = "" if n_indent == 0 else "(Reminder: as you can see, this piece of code has indent made up with {n_indent} whitespace, please preseve them in the OUTPUT.)"
+        brief_reminder = "" if self.file_brief == "" else f"({self.file_basename} abstract: {self.file_brief})"
+        hint_reminder = "" if hint is None else f"(Reminder: do not ignore or modify code such as `{hint}`, provide complete code in the OUTPUT.)"
+        self.llm_kwargs['temperature'] = 0
+        result = predict_no_ui_long_connection(
+            inputs=revise_funtion_prompt.format(
+                LANG=self.language, 
+                FILE_BASENAME=self.file_basename, 
+                THE_CODE=code, 
+                INDENT_REMINDER=indent_reminder, 
+                BRIEF_REMINDER=brief_reminder,
+                HINT_REMINDER=hint_reminder
+            ),
+            llm_kwargs=self.llm_kwargs,
+            history=[],
+            sys_prompt="",
+            observe_window=[],
+            console_slience=True
+        )
+
+        def get_code_block(reply):
+            import re
+            pattern = r"```([\s\S]*?)```" # regex pattern to match code blocks
+            matches = re.findall(pattern, reply) # find all code blocks in text
+            if len(matches) == 1:
+                return matches[0].strip('python') #  code block
+            return None
+
+        code_block = get_code_block(result)
+        if code_block is not None:
+            code_block = self.sync_and_patch(original=code, revised=code_block)
+            return code_block
+        else:
+            return code
+        
+    def get_markdown_block_in_html(self, html):
+        from bs4 import BeautifulSoup
+        soup = BeautifulSoup(html, 'lxml')
+        found_list = soup.find_all("div", class_="markdown-body")
+        if found_list:
+            res = found_list[0]
+            return res.prettify()
+        else:
+            return None
+
+
+    def sync_and_patch(self, original, revised):
+        """Ensure the number of pre-string empty lines in revised matches those in original."""
+
+        def count_leading_empty_lines(s, reverse=False):
+            """Count the number of leading empty lines in a string."""
+            lines = s.split('\n')
+            if reverse: lines = list(reversed(lines))
+            count = 0
+            for line in lines:
+                if line.strip() == '':
+                    count += 1
+                else:
+                    break
+            return count
+
+        original_empty_lines = count_leading_empty_lines(original)
+        revised_empty_lines = count_leading_empty_lines(revised)
+
+        if original_empty_lines > revised_empty_lines:
+            additional_lines = '\n' * (original_empty_lines - revised_empty_lines)
+            revised = additional_lines + revised
+        elif original_empty_lines < revised_empty_lines:
+            lines = revised.split('\n')
+            revised = '\n'.join(lines[revised_empty_lines - original_empty_lines:])
+
+        original_empty_lines = count_leading_empty_lines(original, reverse=True)
+        revised_empty_lines = count_leading_empty_lines(revised, reverse=True)
+
+        if original_empty_lines > revised_empty_lines:
+            additional_lines = '\n' * (original_empty_lines - revised_empty_lines)
+            revised =  revised + additional_lines
+        elif original_empty_lines < revised_empty_lines:
+            lines = revised.split('\n')
+            revised = '\n'.join(lines[:-(revised_empty_lines - original_empty_lines)])
+
+        return revised
+
+    def begin_comment_source_code(self, chatbot=None, history=None):
+        # from toolbox import update_ui_lastest_msg
+        assert self.path is not None
+        assert '.py' in self.path   # must be python source code
+        # write_target = self.path + '.revised.py'
+
+        write_content = ""
+        # with open(self.path + '.revised.py', 'w+', encoding='utf8') as f:
+        while True:
+            try:
+                # yield from update_ui_lastest_msg(f"({self.file_basename}) 正在读取下一段代码片段:\n", chatbot=chatbot, history=history, delay=0)
+                next_batch, line_no_start, line_no_end = self.get_next_batch()
+                # yield from update_ui_lastest_msg(f"({self.file_basename}) 处理代码片段:\n\n{next_batch}", chatbot=chatbot, history=history, delay=0)
+                
+                hint = None
+                MAX_ATTEMPT = 2
+                for attempt in range(MAX_ATTEMPT):
+                    result = self.tag_code(next_batch, hint)
+                    try:
+                        successful, hint = self.verify_successful(next_batch, result)
+                    except Exception as e:
+                        logger.error('ignored exception:\n' + str(e))
+                        break
+                    if successful:
+                        break
+                    if attempt == MAX_ATTEMPT - 1:
+                        # cannot deal with this, give up
+                        result = next_batch
+                        break
+
+                # f.write(result)
+                write_content += result
+            except StopIteration:
+                next_batch, line_no_start, line_no_end = [], -1, -1
+                return None, write_content
+
+    def verify_successful(self, original, revised):
+        """ Determine whether the revised code contains every line that already exists
+        """
+        from crazy_functions.ast_fns.comment_remove import remove_python_comments
+        original = remove_python_comments(original)
+        original_lines = original.split('\n')
+        revised_lines = revised.split('\n')
+
+        for l in original_lines:
+            l = l.strip()
+            if '\'' in l or '\"' in l: continue  # ast sometimes toggle " to '
+            found = False
+            for lt in revised_lines:
+                if l in lt:
+                    found = True
+                    break
+            if not found:
+                return False, l
+        return True, None

crazy_functions/agent_fns/python_comment_compare.html

@@ -0,0 +1,45 @@
+<!DOCTYPE html>
+<html lang="zh-CN">
+<head>
+    <style>ADVANCED_CSS</style>
+    <meta charset="UTF-8">
+    <title>源文件对比</title>
+    <style>
+        body {
+            font-family: Arial, sans-serif;
+            display: flex;
+            justify-content: center;
+            align-items: center;
+            height: 100vh;
+            margin: 0;
+        }
+        .container {
+            display: flex;
+            width: 95%;
+            height: -webkit-fill-available;
+        }
+        .code-container {
+            flex: 1;
+            margin: 0px;
+            padding: 0px;
+            border: 1px solid #ccc;
+            background-color: #f9f9f9;
+            overflow: auto;
+        }
+        pre {
+            white-space: pre-wrap;
+            word-wrap: break-word;
+        }
+    </style>
+</head>
+<body>
+<div class="container">
+<div class="code-container">
+REPLACE_CODE_FILE_LEFT
+</div>
+<div class="code-container">
+REPLACE_CODE_FILE_RIGHT
+</div>
+</div>
+</body>
+</html>

crazy_functions/agent_fns/watchdog.py

@@ -0,0 +1,29 @@
+import threading, time
+from loguru import logger
+
+class WatchDog():
+    def __init__(self, timeout, bark_fn, interval=3, msg="") -> None:
+        self.last_feed = None
+        self.timeout = timeout
+        self.bark_fn = bark_fn
+        self.interval = interval
+        self.msg = msg
+        self.kill_dog = False
+
+    def watch(self):
+        while True:
+            if self.kill_dog: break
+            if time.time() - self.last_feed > self.timeout:
+                if len(self.msg) > 0: logger.info(self.msg)
+                self.bark_fn()
+                break
+            time.sleep(self.interval)
+
+    def begin_watch(self):
+        self.last_feed = time.time()
+        th = threading.Thread(target=self.watch)
+        th.daemon = True
+        th.start()
+
+    def feed(self):
+        self.last_feed = time.time()

crazy_functions/ast_fns/comment_remove.py

@@ -0,0 +1,46 @@
+import ast
+
+class CommentRemover(ast.NodeTransformer):
+    def visit_FunctionDef(self, node):
+        # 移除函数的文档字符串
+        if (node.body and isinstance(node.body[0], ast.Expr) and
+                isinstance(node.body[0].value, ast.Str)):
+            node.body = node.body[1:]
+        self.generic_visit(node)
+        return node
+
+    def visit_ClassDef(self, node):
+        # 移除类的文档字符串
+        if (node.body and isinstance(node.body[0], ast.Expr) and
+                isinstance(node.body[0].value, ast.Str)):
+            node.body = node.body[1:]
+        self.generic_visit(node)
+        return node
+
+    def visit_Module(self, node):
+        # 移除模块的文档字符串
+        if (node.body and isinstance(node.body[0], ast.Expr) and
+                isinstance(node.body[0].value, ast.Str)):
+            node.body = node.body[1:]
+        self.generic_visit(node)
+        return node
+    
+
+def remove_python_comments(source_code):
+    # 解析源代码为 AST
+    tree = ast.parse(source_code)
+    # 移除注释
+    transformer = CommentRemover()
+    tree = transformer.visit(tree)
+    # 将处理后的 AST 转换回源代码
+    return ast.unparse(tree)
+
+# 示例使用
+if __name__ == "__main__":
+    with open("source.py", "r", encoding="utf-8") as f:
+        source_code = f.read()
+
+    cleaned_code = remove_python_comments(source_code)
+
+    with open("cleaned_source.py", "w", encoding="utf-8") as f:
+        f.write(cleaned_code)

crazy_functions/chatglm微调工具.py

@@ -1,5 +1,5 @@
 from toolbox import CatchException, update_ui, promote_file_to_downloadzone
-from .crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
+from crazy_functions.crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
 import datetime, json
 
 def fetch_items(list_of_items, batch_size):
@@ -32,7 +32,7 @@ def string_to_options(arguments):
     return args
 
 @CatchException
-def 微调数据集生成(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 微调数据集生成(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     """
     txt             输入栏用户输入的文本，例如需要翻译的一段话，再例如一个包含了待处理文件的路径
     llm_kwargs      gpt模型参数，如温度和top_p等，一般原样传递下去就行
@@ -40,7 +40,7 @@ def 微调数据集生成(txt, llm_kwargs, plugin_kwargs, chatbot, history, syst
     chatbot         聊天显示框的句柄，用于显示给用户
     history         聊天历史，前情提要
     system_prompt   给gpt的静默提醒
-    web_port        当前软件运行的端口号
+    user_request    当前用户的请求信息（IP地址等）
     """
     history = []    # 清空历史，以免输入溢出
     chatbot.append(("这是什么功能？", "[Local Message] 微调数据集生成"))
@@ -80,7 +80,7 @@ def 微调数据集生成(txt, llm_kwargs, plugin_kwargs, chatbot, history, syst
 
 
 @CatchException
-def 启动微调(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, web_port):
+def 启动微调(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
     """
     txt             输入栏用户输入的文本，例如需要翻译的一段话，再例如一个包含了待处理文件的路径
     llm_kwargs      gpt模型参数，如温度和top_p等，一般原样传递下去就行
@@ -88,7 +88,7 @@ def 启动微调(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt
     chatbot         聊天显示框的句柄，用于显示给用户
     history         聊天历史，前情提要
     system_prompt   给gpt的静默提醒
-    web_port        当前软件运行的端口号
+    user_request    当前用户的请求信息（IP地址等）
     """
     import subprocess
     history = []    # 清空历史，以免输入溢出

crazy_functions/crazy_utils.py

@@ -1,23 +1,39 @@
-from toolbox import update_ui, get_conf, trimmed_format_exc
+import os
 import threading
+from loguru import logger
+from shared_utils.char_visual_effect import scolling_visual_effect
+from toolbox import update_ui, get_conf, trimmed_format_exc, get_max_token, Singleton
 
-def input_clipping(inputs, history, max_token_limit):
+def input_clipping(inputs, history, max_token_limit, return_clip_flags=False):
+    """
+    当输入文本 + 历史文本超出最大限制时，采取措施丢弃一部分文本。
+    输入：
+        - inputs 本次请求
+        - history 历史上下文
+        - max_token_limit 最大token限制
+    输出:
+        - inputs 本次请求（经过clip）
+        - history 历史上下文（经过clip）
+    """
     import numpy as np
-    from request_llm.bridge_all import model_info
+    from request_llms.bridge_all import model_info
     enc = model_info["gpt-3.5-turbo"]['tokenizer']
     def get_token_num(txt): return len(enc.encode(txt, disallowed_special=()))
 
+
     mode = 'input-and-history'
     # 当 输入部分的token占比 小于 全文的一半时，只裁剪历史
     input_token_num = get_token_num(inputs)
+    original_input_len = len(inputs)
     if input_token_num < max_token_limit//2:
         mode = 'only-history'
         max_token_limit = max_token_limit - input_token_num
 
     everything = [inputs] if mode == 'input-and-history' else ['']
     everything.extend(history)
-    n_token = get_token_num('\n'.join(everything))
+    full_token_num = n_token = get_token_num('\n'.join(everything))
     everything_token = [get_token_num(e) for e in everything]
+    everything_token_num = sum(everything_token)
     delta = max(everything_token) // 16 # 截断时的颗粒度
 
     while n_token > max_token_limit:
@@ -30,10 +46,24 @@ def input_clipping(inputs, history, max_token_limit):
 
     if mode == 'input-and-history':
         inputs = everything[0]
+        full_token_num = everything_token_num
     else:
-        pass
+        full_token_num = everything_token_num + input_token_num
+
     history = everything[1:]
+
+    flags = {
+        "mode": mode,
+        "original_input_token_num": input_token_num,
+        "original_full_token_num": full_token_num,
+        "original_input_len": original_input_len,
+        "clipped_input_len": len(inputs),
+    }
+
+    if not return_clip_flags:
         return inputs, history
+    else:
+        return inputs, history, flags
 
 def request_gpt_model_in_new_thread_with_ui_alive(
         inputs, inputs_show_user, llm_kwargs,
@@ -61,18 +91,21 @@ def request_gpt_model_in_new_thread_with_ui_alive(
     """
     import time
     from concurrent.futures import ThreadPoolExecutor
-    from request_llm.bridge_all import predict_no_ui_long_connection
+    from request_llms.bridge_all import predict_no_ui_long_connection
     # 用户反馈
     chatbot.append([inputs_show_user, ""])
     yield from update_ui(chatbot=chatbot, history=[]) # 刷新界面
     executor = ThreadPoolExecutor(max_workers=16)
     mutable = ["", time.time(), ""]
+    # 看门狗耐心
+    watch_dog_patience = 5
+    # 请求任务
     def _req_gpt(inputs, history, sys_prompt):
         retry_op = retry_times_at_unknown_error
         exceeded_cnt = 0
         while True:
             # watchdog error
-            if len(mutable) >= 2 and (time.time()-mutable[1]) > 5: 
+            if len(mutable) >= 2 and (time.time()-mutable[1]) > watch_dog_patience:
                 raise RuntimeError("检测到程序终止。")
             try:
                 # 【第一种情况】：顺利完成
@@ -87,7 +120,7 @@ def request_gpt_model_in_new_thread_with_ui_alive(
                     # 【选择处理】 尝试计算比例，尽可能多地保留文本
                     from toolbox import get_reduce_token_percent
                     p_ratio, n_exceed = get_reduce_token_percent(str(token_exceeded_error))
-                    MAX_TOKEN = 4096
+                    MAX_TOKEN = get_max_token(llm_kwargs)
                     EXCEED_ALLO = 512 + 512 * exceeded_cnt
                     inputs, history = input_clipping(inputs, history, max_token_limit=MAX_TOKEN-EXCEED_ALLO)
                     mutable[0] += f'[Local Message] 警告，文本过长将进行截断，Token溢出数：{n_exceed}。\n\n'
@@ -100,7 +133,7 @@ def request_gpt_model_in_new_thread_with_ui_alive(
             except:
                 # 【第三种情况】：其他错误：重试几次
                 tb_str = '```\n' + trimmed_format_exc() + '```'
-                print(tb_str)
+                logger.error(tb_str)
                 mutable[0] += f"[Local Message] 警告，在执行过程中遭遇问题, Traceback：\n\n{tb_str}\n\n"
                 if retry_op > 0:
                     retry_op -= 1
@@ -130,16 +163,30 @@ def request_gpt_model_in_new_thread_with_ui_alive(
     yield from update_ui(chatbot=chatbot, history=[]) # 如果最后成功了，则删除报错信息
     return final_result
 
-def can_multi_process(llm):
+def can_multi_process(llm) -> bool:
+    from request_llms.bridge_all import model_info
+
+    def default_condition(llm) -> bool:
+        # legacy condition
         if llm.startswith('gpt-'): return True
         if llm.startswith('api2d-'): return True
         if llm.startswith('azure-'): return True
+        if llm.startswith('spark'): return True
+        if llm.startswith('zhipuai') or llm.startswith('glm-'): return True
         return False
 
+    if llm in model_info:
+        if 'can_multi_thread' in model_info[llm]:
+            return model_info[llm]['can_multi_thread']
+        else:
+            return default_condition(llm)
+    else:
+        return default_condition(llm)
+
 def request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
         inputs_array, inputs_show_user_array, llm_kwargs,
         chatbot, history_array, sys_prompt_array,
-        refresh_interval=0.2, max_workers=-1, scroller_max_len=30,
+        refresh_interval=0.2, max_workers=-1, scroller_max_len=75,
         handle_token_exceed=True, show_user_at_complete=False,
         retry_times_at_unknown_error=2,
         ):
@@ -172,11 +219,11 @@ def request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
     """
     import time, random
     from concurrent.futures import ThreadPoolExecutor
-    from request_llm.bridge_all import predict_no_ui_long_connection
+    from request_llms.bridge_all import predict_no_ui_long_connection
     assert len(inputs_array) == len(history_array)
     assert len(inputs_array) == len(sys_prompt_array)
     if max_workers == -1: # 读取配置文件
-        try: max_workers, = get_conf('DEFAULT_WORKER_NUM')
+        try: max_workers = get_conf('DEFAULT_WORKER_NUM')
         except: max_workers = 8
         if max_workers <= 0: max_workers = 3
     # 屏蔽掉 chatglm的多线程，可能会导致严重卡顿
@@ -191,19 +238,21 @@ def request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
     # 跨线程传递
     mutable = [["", time.time(), "等待中"] for _ in range(n_frag)]
 
+    # 看门狗耐心
+    watch_dog_patience = 5
+
     # 子线程任务
     def _req_gpt(index, inputs, history, sys_prompt):
         gpt_say = ""
         retry_op = retry_times_at_unknown_error
         exceeded_cnt = 0
         mutable[index][2] = "执行中"
+        detect_timeout = lambda: len(mutable[index]) >= 2 and (time.time()-mutable[index][1]) > watch_dog_patience
         while True:
             # watchdog error
-            if len(mutable[index]) >= 2 and (time.time()-mutable[index][1]) > 5: 
-                raise RuntimeError("检测到程序终止。")
+            if detect_timeout(): raise RuntimeError("检测到程序终止。")
             try:
                 # 【第一种情况】：顺利完成
-                # time.sleep(10); raise RuntimeError("测试")
                 gpt_say = predict_no_ui_long_connection(
                     inputs=inputs, llm_kwargs=llm_kwargs, history=history,
                     sys_prompt=sys_prompt, observe_window=mutable[index], console_slience=True
@@ -211,13 +260,13 @@ def request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
                 mutable[index][2] = "已成功"
                 return gpt_say
             except ConnectionAbortedError as token_exceeded_error:
-                # 【第二种情况】：Token溢出，
+                # 【第二种情况】：Token溢出
                 if handle_token_exceed:
                     exceeded_cnt += 1
                     # 【选择处理】 尝试计算比例，尽可能多地保留文本
                     from toolbox import get_reduce_token_percent
                     p_ratio, n_exceed = get_reduce_token_percent(str(token_exceeded_error))
-                    MAX_TOKEN = 4096
+                    MAX_TOKEN = get_max_token(llm_kwargs)
                     EXCEED_ALLO = 512 + 512 * exceeded_cnt
                     inputs, history = input_clipping(inputs, history, max_token_limit=MAX_TOKEN-EXCEED_ALLO)
                     gpt_say += f'[Local Message] 警告，文本过长将进行截断，Token溢出数：{n_exceed}。\n\n'
@@ -232,8 +281,9 @@ def request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
                     return gpt_say # 放弃
             except:
                 # 【第三种情况】：其他错误
+                if detect_timeout(): raise RuntimeError("检测到程序终止。")
                 tb_str = '```\n' + trimmed_format_exc() + '```'
-                print(tb_str)
+                logger.error(tb_str)
                 gpt_say += f"[Local Message] 警告，线程{index}在执行过程中遭遇问题, Traceback：\n\n{tb_str}\n\n"
                 if len(mutable[index][0]) > 0: gpt_say += "此线程失败前收到的回答：\n\n" + mutable[index][0]
                 if retry_op > 0:
@@ -248,6 +298,7 @@ def request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
                     for i in range(wait):
                         mutable[index][2] = f"{fail_info}等待重试 {wait-i}"; time.sleep(1)
                     # 开始重试
+                    if detect_timeout(): raise RuntimeError("检测到程序终止。")
                     mutable[index][2] = f"重试中 {retry_times_at_unknown_error-retry_op}/{retry_times_at_unknown_error}"
                     continue # 返回重试
                 else:
@@ -260,6 +311,8 @@ def request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
     futures = [executor.submit(_req_gpt, index, inputs, history, sys_prompt) for index, inputs, history, sys_prompt in zip(
         range(len(inputs_array)), inputs_array, history_array, sys_prompt_array)]
     cnt = 0
+
+
     while True:
         # yield一次以刷新前端页面
         time.sleep(refresh_interval)
@@ -272,9 +325,7 @@ def request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
             mutable[thread_index][1] = time.time()
         # 在前端打印些好玩的东西
         for thread_index, _ in enumerate(worker_done):
-            print_something_really_funny = "[ ...`"+mutable[thread_index][0][-scroller_max_len:].\
-                replace('\n', '').replace('```', '...').replace(
-                    ' ', '.').replace('<br/>', '.....').replace('$', '.')+"`... ]"
+            print_something_really_funny = f"[ ...`{scolling_visual_effect(mutable[thread_index][0], scroller_max_len)}`... ]"
             observe_win.append(print_something_really_funny)
         # 在前端打印些好玩的东西
         stat_str = ''.join([f'`{mutable[thread_index][2]}`: {obs}\n\n'
@@ -299,99 +350,10 @@ def request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
             gpt_res = f.result()
             chatbot.append([inputs_show_user, gpt_res])
             yield from update_ui(chatbot=chatbot, history=[]) # 刷新界面
-            time.sleep(0.3)
+            time.sleep(0.5)
     return gpt_response_collection
 
 
-def breakdown_txt_to_satisfy_token_limit(txt, get_token_fn, limit):
-    def cut(txt_tocut, must_break_at_empty_line):  # 递归
-        if get_token_fn(txt_tocut) <= limit:
-            return [txt_tocut]
-        else:
-            lines = txt_tocut.split('\n')
-            estimated_line_cut = limit / get_token_fn(txt_tocut) * len(lines)
-            estimated_line_cut = int(estimated_line_cut)
-            for cnt in reversed(range(estimated_line_cut)):
-                if must_break_at_empty_line:
-                    if lines[cnt] != "":
-                        continue
-                print(cnt)
-                prev = "\n".join(lines[:cnt])
-                post = "\n".join(lines[cnt:])
-                if get_token_fn(prev) < limit:
-                    break
-            if cnt == 0:
-                raise RuntimeError("存在一行极长的文本！")
-            # print(len(post))
-            # 列表递归接龙
-            result = [prev]
-            result.extend(cut(post, must_break_at_empty_line))
-            return result
-    try:
-        return cut(txt, must_break_at_empty_line=True)
-    except RuntimeError:
-        return cut(txt, must_break_at_empty_line=False)
-
-
-def force_breakdown(txt, limit, get_token_fn):
-    """
-    当无法用标点、空行分割时，我们用最暴力的方法切割
-    """
-    for i in reversed(range(len(txt))):
-        if get_token_fn(txt[:i]) < limit:
-            return txt[:i], txt[i:]
-    return "Tiktoken未知错误", "Tiktoken未知错误"
-
-def breakdown_txt_to_satisfy_token_limit_for_pdf(txt, get_token_fn, limit):
-    # 递归
-    def cut(txt_tocut, must_break_at_empty_line, break_anyway=False):  
-        if get_token_fn(txt_tocut) <= limit:
-            return [txt_tocut]
-        else:
-            lines = txt_tocut.split('\n')
-            estimated_line_cut = limit / get_token_fn(txt_tocut) * len(lines)
-            estimated_line_cut = int(estimated_line_cut)
-            cnt = 0
-            for cnt in reversed(range(estimated_line_cut)):
-                if must_break_at_empty_line:
-                    if lines[cnt] != "":
-                        continue
-                prev = "\n".join(lines[:cnt])
-                post = "\n".join(lines[cnt:])
-                if get_token_fn(prev) < limit:
-                    break
-            if cnt == 0:
-                if break_anyway:
-                    prev, post = force_breakdown(txt_tocut, limit, get_token_fn)
-                else:
-                    raise RuntimeError(f"存在一行极长的文本！{txt_tocut}")
-            # print(len(post))
-            # 列表递归接龙
-            result = [prev]
-            result.extend(cut(post, must_break_at_empty_line, break_anyway=break_anyway))
-            return result
-    try:
-        # 第1次尝试，将双空行（\n\n）作为切分点
-        return cut(txt, must_break_at_empty_line=True)
-    except RuntimeError:
-        try:
-            # 第2次尝试，将单空行（\n）作为切分点
-            return cut(txt, must_break_at_empty_line=False)
-        except RuntimeError:
-            try:
-                # 第3次尝试，将英文句号（.）作为切分点
-                res = cut(txt.replace('.', '。\n'), must_break_at_empty_line=False) # 这个中文的句号是故意的，作为一个标识而存在
-                return [r.replace('。\n', '.') for r in res]
-            except RuntimeError as e:
-                try:
-                    # 第4次尝试，将中文句号（。）作为切分点
-                    res = cut(txt.replace('。', '。。\n'), must_break_at_empty_line=False)
-                    return [r.replace('。。\n', '。') for r in res]
-                except RuntimeError as e:
-                    # 第5次尝试，没办法了，随便切一下敷衍吧
-                    return cut(txt, must_break_at_empty_line=False, break_anyway=True)
-
-
 
 def read_and_clean_pdf_text(fp):
     """
@@ -416,7 +378,7 @@ def read_and_clean_pdf_text(fp):
     import fitz, copy
     import re
     import numpy as np
-    from colorful import print亮黄, print亮绿
+    # from shared_utils.colorful import print亮黄, print亮绿
     fc = 0  # Index 0 文本
     fs = 1  # Index 1 字体
     fb = 2  # Index 2 框框
@@ -469,6 +431,7 @@ def read_and_clean_pdf_text(fp):
                     '- ', '') for t in text_areas['blocks'] if 'lines' in t]
 
         ############################## <第 2 步，获取正文主字体> ##################################
+        try:
             fsize_statiscs = {}
             for span in meta_span:
                 if span[1] not in fsize_statiscs: fsize_statiscs[span[1]] = 0
@@ -476,7 +439,8 @@ def read_and_clean_pdf_text(fp):
             main_fsize = max(fsize_statiscs, key=fsize_statiscs.get)
             if REMOVE_FOOT_NOTE:
                 give_up_fize_threshold = main_fsize * REMOVE_FOOT_FFSIZE_PERCENT
-
+        except:
+            raise RuntimeError(f'抱歉, 我们暂时无法解析此PDF文档: {fp}。')
         ############################## <第 3 步，切分和重新整合> ##################################
         mega_sec = []
         sec = []
@@ -542,6 +506,9 @@ def read_and_clean_pdf_text(fp):
                     return True
                 else:
                     return False
+            # 对于某些PDF会有第一个段落就以小写字母开头,为了避免索引错误将其更改为大写
+            if starts_with_lowercase_word(meta_txt[0]):
+                meta_txt[0] = meta_txt[0].capitalize()
             for _ in range(100):
                 for index, block_txt in enumerate(meta_txt):
                     if starts_with_lowercase_word(block_txt):
@@ -591,11 +558,16 @@ def get_files_from_everything(txt, type): # type='.md'
         # 网络的远程文件
         import requests
         from toolbox import get_conf
-        proxies, = get_conf('proxies')
+        from toolbox import get_log_folder, gen_time_str
+        proxies = get_conf('proxies')
+        try:
             r = requests.get(txt, proxies=proxies)
-        with open('./gpt_log/temp'+type, 'wb+') as f: f.write(r.content)
-        project_folder = './gpt_log/'
-        file_manifest = ['./gpt_log/temp'+type]
+        except:
+            raise ConnectionRefusedError(f"无法下载资源{txt}，请检查。")
+        path = os.path.join(get_log_folder(plugin_name='web_download'), gen_time_str()+type)
+        with open(path, 'wb+') as f: f.write(r.content)
+        project_folder = get_log_folder(plugin_name='web_download')
+        file_manifest = [path]
     elif txt.endswith(type):
         # 直接给定文件
         file_manifest = [txt]
@@ -615,139 +587,64 @@ def get_files_from_everything(txt, type): # type='.md'
 
 
 
-
-def Singleton(cls):
-    _instance = {}
- 
-    def _singleton(*args, **kargs):
-        if cls not in _instance:
-            _instance[cls] = cls(*args, **kargs)
-        return _instance[cls]
- 
-    return _singleton
-
-
 @Singleton
-class knowledge_archive_interface():
-    def __init__(self) -> None:
+class nougat_interface():
+    def __init__(self):
         self.threadLock = threading.Lock()
-        self.current_id = ""
-        self.kai_path = None
-        self.qa_handle = None
-        self.text2vec_large_chinese = None
 
-    def get_chinese_text2vec(self):
-        if self.text2vec_large_chinese is None:
-            # < -------------------预热文本向量化模组--------------- >
+    def nougat_with_timeout(self, command, cwd, timeout=3600):
+        import subprocess
         from toolbox import ProxyNetworkActivate
-            print('Checking Text2vec ...')
-            from langchain.embeddings.huggingface import HuggingFaceEmbeddings
-            with ProxyNetworkActivate():    # 临时地激活代理网络
-                self.text2vec_large_chinese = HuggingFaceEmbeddings(model_name="GanymedeNil/text2vec-large-chinese")
-
-        return self.text2vec_large_chinese
+        logger.info(f'正在执行命令 {command}')
+        with ProxyNetworkActivate("Nougat_Download"):
+            process = subprocess.Popen(command, shell=False, cwd=cwd, env=os.environ)
+        try:
+            stdout, stderr = process.communicate(timeout=timeout)
+        except subprocess.TimeoutExpired:
+            process.kill()
+            stdout, stderr = process.communicate()
+            logger.error("Process timed out!")
+            return False
+        return True
 
 
-    def feed_archive(self, file_manifest, id="default"):
+    def NOUGAT_parse_pdf(self, fp, chatbot, history):
+        from toolbox import update_ui_lastest_msg
+
+        yield from update_ui_lastest_msg("正在解析论文, 请稍候。进度：正在排队, 等待线程锁...",
+                                         chatbot=chatbot, history=history, delay=0)
         self.threadLock.acquire()
-        # import uuid
-        self.current_id = id
-        from zh_langchain import construct_vector_store
-        self.qa_handle, self.kai_path = construct_vector_store(   
-            vs_id=self.current_id, 
-            files=file_manifest, 
-            sentence_size=100,
-            history=[],
-            one_conent="",
-            one_content_segmentation="",
-            text2vec = self.get_chinese_text2vec(),
-        )
+        import glob, threading, os
+        from toolbox import get_log_folder, gen_time_str
+        dst = os.path.join(get_log_folder(plugin_name='nougat'), gen_time_str())
+        os.makedirs(dst)
+
+        yield from update_ui_lastest_msg("正在解析论文, 请稍候。进度：正在加载NOUGAT... （提示：首次运行需要花费较长时间下载NOUGAT参数）",
+                                         chatbot=chatbot, history=history, delay=0)
+        command = ['nougat', '--out', os.path.abspath(dst), os.path.abspath(fp)]
+        self.nougat_with_timeout(command, cwd=os.getcwd(), timeout=3600)
+        res = glob.glob(os.path.join(dst,'*.mmd'))
+        if len(res) == 0:
             self.threadLock.release()
-
-    def get_current_archive_id(self):
-        return self.current_id
-    
-    def get_loaded_file(self):
-        return self.qa_handle.get_loaded_file()
-
-    def answer_with_archive_by_id(self, txt, id):
-        self.threadLock.acquire()
-        if not self.current_id == id:
-            self.current_id = id
-            from zh_langchain import construct_vector_store
-            self.qa_handle, self.kai_path = construct_vector_store(   
-                vs_id=self.current_id, 
-                files=[], 
-                sentence_size=100,
-                history=[],
-                one_conent="",
-                one_content_segmentation="",
-                text2vec = self.get_chinese_text2vec(),
-            )
-        VECTOR_SEARCH_SCORE_THRESHOLD = 0
-        VECTOR_SEARCH_TOP_K = 4
-        CHUNK_SIZE = 512
-        resp, prompt = self.qa_handle.get_knowledge_based_conent_test(
-            query = txt,
-            vs_path = self.kai_path,
-            score_threshold=VECTOR_SEARCH_SCORE_THRESHOLD,
-            vector_search_top_k=VECTOR_SEARCH_TOP_K, 
-            chunk_conent=True,
-            chunk_size=CHUNK_SIZE,
-            text2vec = self.get_chinese_text2vec(),
-        )
+            raise RuntimeError("Nougat解析论文失败。")
         self.threadLock.release()
-        return resp, prompt
+        return res[0]
 
-def try_install_deps(deps):
+
+
+
+def try_install_deps(deps, reload_m=[]):
+    import subprocess, sys, importlib
     for dep in deps:
-        import subprocess, sys
         subprocess.check_call([sys.executable, '-m', 'pip', 'install', '--user', dep])
+    import site
+    importlib.reload(site)
+    for m in reload_m:
+        importlib.reload(__import__(m))
 
 
-class construct_html():
-    def __init__(self) -> None:
-        self.css = """
-.row {
-  display: flex;
-  flex-wrap: wrap;
-}
-
-.column {
-  flex: 1;
-  padding: 10px;
-}
-
-.table-header {
-  font-weight: bold;
-  border-bottom: 1px solid black;
-}
-
-.table-row {
-  border-bottom: 1px solid lightgray;
-}
-
-.table-cell {
-  padding: 5px;
-}
-        """
-        self.html_string = f'<!DOCTYPE html><head><meta charset="utf-8"><title>翻译结果</title><style>{self.css}</style></head>'
-
-
-    def add_row(self, a, b):
-        tmp = """
-<div class="row table-row">
-    <div class="column table-cell">REPLACE_A</div>
-    <div class="column table-cell">REPLACE_B</div>
-</div>
-        """
-        from toolbox import markdown_convertion
-        tmp = tmp.replace('REPLACE_A', markdown_convertion(a))
-        tmp = tmp.replace('REPLACE_B', markdown_convertion(b))
-        self.html_string += tmp
-
-
-    def save_file(self, file_name):
-        with open(f'./gpt_log/{file_name}', 'w', encoding='utf8') as f:
-            f.write(self.html_string.encode('utf-8', 'ignore').decode())
-
+def get_plugin_arg(plugin_kwargs, key, default):
+    # 如果参数是空的
+    if (key in plugin_kwargs) and (plugin_kwargs[key] == ""): plugin_kwargs.pop(key)
+    # 正常情况
+    return plugin_kwargs.get(key, default)

crazy_functions/diagram_fns/file_tree.py

@@ -0,0 +1,127 @@
+import os
+from textwrap import indent
+from loguru import logger
+
+class FileNode:
+    def __init__(self, name, build_manifest=False):
+        self.name = name
+        self.children = []
+        self.is_leaf = False
+        self.level = 0
+        self.parenting_ship = []
+        self.comment = ""
+        self.comment_maxlen_show = 50
+        self.build_manifest = build_manifest
+        self.manifest = {}
+
+    @staticmethod
+    def add_linebreaks_at_spaces(string, interval=10):
+        return '\n'.join(string[i:i+interval] for i in range(0, len(string), interval))
+
+    def sanitize_comment(self, comment):
+        if len(comment) > self.comment_maxlen_show: suf = '...'
+        else: suf = ''
+        comment = comment[:self.comment_maxlen_show]
+        comment = comment.replace('\"', '').replace('`', '').replace('\n', '').replace('`', '').replace('$', '')
+        comment = self.add_linebreaks_at_spaces(comment, 10)
+        return '`' + comment + suf + '`'
+
+    def add_file(self, file_path, file_comment):
+        directory_names, file_name = os.path.split(file_path)
+        current_node = self
+        level = 1
+        if directory_names == "":
+            new_node = FileNode(file_name)
+            self.manifest[file_path] = new_node
+            current_node.children.append(new_node)
+            new_node.is_leaf = True
+            new_node.comment = self.sanitize_comment(file_comment)
+            new_node.level = level
+            current_node = new_node
+        else:
+            dnamesplit = directory_names.split(os.sep)
+            for i, directory_name in enumerate(dnamesplit):
+                found_child = False
+                level += 1
+                for child in current_node.children:
+                    if child.name == directory_name:
+                        current_node = child
+                        found_child = True
+                        break
+                if not found_child:
+                    new_node = FileNode(directory_name)
+                    current_node.children.append(new_node)
+                    new_node.level = level - 1
+                    current_node = new_node
+            term = FileNode(file_name)
+            self.manifest[file_path] = term
+            term.level = level
+            term.comment = self.sanitize_comment(file_comment)
+            term.is_leaf = True
+            current_node.children.append(term)
+
+    def print_files_recursively(self, level=0, code="R0"):
+        logger.info('    '*level + self.name + ' ' + str(self.is_leaf) + ' ' + str(self.level))
+        for j, child in enumerate(self.children):
+            child.print_files_recursively(level=level+1, code=code+str(j))
+            self.parenting_ship.extend(child.parenting_ship)
+            p1 = f"""{code}[\"🗎{self.name}\"]""" if self.is_leaf else f"""{code}[[\"📁{self.name}\"]]"""
+            p2 = """ --> """
+            p3 = f"""{code+str(j)}[\"🗎{child.name}\"]""" if child.is_leaf else f"""{code+str(j)}[[\"📁{child.name}\"]]"""
+            edge_code = p1 + p2 + p3
+            if edge_code in self.parenting_ship:
+                continue
+            self.parenting_ship.append(edge_code)
+        if self.comment != "":
+            pc1 = f"""{code}[\"🗎{self.name}\"]""" if self.is_leaf else f"""{code}[[\"📁{self.name}\"]]"""
+            pc2 = f""" -.-x """
+            pc3 = f"""C{code}[\"{self.comment}\"]:::Comment"""
+            edge_code = pc1 + pc2 + pc3
+            self.parenting_ship.append(edge_code)
+
+
+MERMAID_TEMPLATE = r"""
+```mermaid
+flowchart LR
+    %% <gpt_academic_hide_mermaid_code> 一个特殊标记，用于在生成mermaid图表时隐藏代码块
+    classDef Comment stroke-dasharray: 5 5
+    subgraph {graph_name}
+{relationship}
+    end
+```
+"""
+
+def build_file_tree_mermaid_diagram(file_manifest, file_comments, graph_name):
+    # Create the root node
+    file_tree_struct = FileNode("root")
+    # Build the tree structure
+    for file_path, file_comment in zip(file_manifest, file_comments):
+        file_tree_struct.add_file(file_path, file_comment)
+    file_tree_struct.print_files_recursively()
+    cc = "\n".join(file_tree_struct.parenting_ship)
+    ccc = indent(cc, prefix=" "*8)
+    return MERMAID_TEMPLATE.format(graph_name=graph_name, relationship=ccc)
+
+if __name__ == "__main__":
+    # File manifest
+    file_manifest = [
+        "cradle_void_terminal.ipynb",
+        "tests/test_utils.py",
+        "tests/test_plugins.py",
+        "tests/test_llms.py",
+        "config.py",
+        "build/ChatGLM-6b-onnx-u8s8/chatglm-6b-int8-onnx-merged/model_weights_0.bin",
+        "crazy_functions/latex_fns/latex_actions.py",
+        "crazy_functions/latex_fns/latex_toolbox.py"
+    ]
+    file_comments = [
+        "根据位置和名称，可能是一个模块的初始化文件根据位置和名称，可能是一个模块的初始化文件根据位置和名称，可能是一个模块的初始化文件",
+        "包含一些用于文本处理和模型微调的函数和装饰器包含一些用于文本处理和模型微调的函数和装饰器包含一些用于文本处理和模型微调的函数和装饰器",
+        "用于构建HTML报告的类和方法用于构建HTML报告的类和方法用于构建HTML报告的类和方法",
+        "包含了用于文本切分的函数，以及处理PDF文件的示例代码包含了用于文本切分的函数，以及处理PDF文件的示例代码包含了用于文本切分的函数，以及处理PDF文件的示例代码",
+        "用于解析和翻译PDF文件的功能和相关辅助函数用于解析和翻译PDF文件的功能和相关辅助函数用于解析和翻译PDF文件的功能和相关辅助函数",
+        "是一个包的初始化文件，用于初始化包的属性和导入模块是一个包的初始化文件，用于初始化包的属性和导入模块是一个包的初始化文件，用于初始化包的属性和导入模块",
+        "用于加载和分割文件中的文本的通用文件加载器用于加载和分割文件中的文本的通用文件加载器用于加载和分割文件中的文本的通用文件加载器",
+        "包含了用于构建和管理向量数据库的函数和类包含了用于构建和管理向量数据库的函数和类包含了用于构建和管理向量数据库的函数和类",
+    ]
+    logger.info(build_file_tree_mermaid_diagram(file_manifest, file_comments, "项目文件树"))

crazy_functions/game_fns/game_ascii_art.py

@@ -0,0 +1,42 @@
+from toolbox import CatchException, update_ui, update_ui_lastest_msg
+from crazy_functions.multi_stage.multi_stage_utils import GptAcademicGameBaseState
+from crazy_functions.crazy_utils import request_gpt_model_in_new_thread_with_ui_alive
+from request_llms.bridge_all import predict_no_ui_long_connection
+from crazy_functions.game_fns.game_utils import get_code_block, is_same_thing
+import random
+
+
+class MiniGame_ASCII_Art(GptAcademicGameBaseState):
+    def step(self, prompt, chatbot, history):
+        if self.step_cnt == 0:
+            chatbot.append(["我画你猜（动物）", "请稍等..."])
+        else:
+            if prompt.strip() == 'exit':
+                self.delete_game = True
+                yield from update_ui_lastest_msg(lastmsg=f"谜底是{self.obj}，游戏结束。", chatbot=chatbot, history=history, delay=0.)
+                return
+            chatbot.append([prompt, ""])
+        yield from update_ui(chatbot=chatbot, history=history)
+
+        if self.step_cnt == 0:
+            self.lock_plugin(chatbot)
+            self.cur_task = 'draw'
+
+        if self.cur_task == 'draw':
+            avail_obj = ["狗","猫","鸟","鱼","老鼠","蛇"]
+            self.obj = random.choice(avail_obj)
+            inputs = "I want to play a game called Guess the ASCII art. You can draw the ASCII art and I will try to guess it. " + \
+                f"This time you draw a {self.obj}. Note that you must not indicate what you have draw in the text, and you should only produce the ASCII art wrapped by ```. "
+            raw_res = predict_no_ui_long_connection(inputs=inputs, llm_kwargs=self.llm_kwargs, history=[], sys_prompt="")
+            self.cur_task = 'identify user guess'
+            res = get_code_block(raw_res)
+            history += ['', f'the answer is {self.obj}', inputs, res]
+            yield from update_ui_lastest_msg(lastmsg=res, chatbot=chatbot, history=history, delay=0.)
+
+        elif self.cur_task == 'identify user guess':
+            if is_same_thing(self.obj, prompt, self.llm_kwargs):
+                self.delete_game = True
+                yield from update_ui_lastest_msg(lastmsg="你猜对了！", chatbot=chatbot, history=history, delay=0.)
+            else:
+                self.cur_task = 'identify user guess'
+                yield from update_ui_lastest_msg(lastmsg="猜错了，再试试，输入“exit”获取答案。", chatbot=chatbot, history=history, delay=0.)

crazy_functions/game_fns/game_interactive_story.py

@@ -0,0 +1,212 @@
+prompts_hs = """ 请以“{headstart}”为开头，编写一个小说的第一幕。
+
+- 尽量短，不要包含太多情节，因为你接下来将会与用户互动续写下面的情节，要留出足够的互动空间。
+- 出现人物时，给出人物的名字。
+- 积极地运用环境描写、人物描写等手法，让读者能够感受到你的故事世界。
+- 积极地运用修辞手法，比如比喻、拟人、排比、对偶、夸张等等。
+- 字数要求：第一幕的字数少于300字，且少于2个段落。
+"""
+
+prompts_interact = """ 小说的前文回顾：
+「
+{previously_on_story}
+」
+
+你是一个作家，根据以上的情节，给出4种不同的后续剧情发展方向，每个发展方向都精明扼要地用一句话说明。稍后，我将在这4个选择中，挑选一种剧情发展。
+
+输出格式例如：
+1. 后续剧情发展1
+2. 后续剧情发展2
+3. 后续剧情发展3
+4. 后续剧情发展4
+"""
+
+
+prompts_resume = """小说的前文回顾：
+「
+{previously_on_story}
+」
+
+你是一个作家，我们正在互相讨论，确定后续剧情的发展。
+在以下的剧情发展中，
+「
+{choice}
+」
+我认为更合理的是：{user_choice}。
+请在前文的基础上（不要重复前文），围绕我选定的剧情情节，编写小说的下一幕。
+
+- 禁止杜撰不符合我选择的剧情。
+- 尽量短，不要包含太多情节，因为你接下来将会与用户互动续写下面的情节，要留出足够的互动空间。
+- 不要重复前文。
+- 出现人物时，给出人物的名字。
+- 积极地运用环境描写、人物描写等手法，让读者能够感受到你的故事世界。
+- 积极地运用修辞手法，比如比喻、拟人、排比、对偶、夸张等等。
+- 小说的下一幕字数少于300字，且少于2个段落。
+"""
+
+
+prompts_terminate = """小说的前文回顾：
+「
+{previously_on_story}
+」
+
+你是一个作家，我们正在互相讨论，确定后续剧情的发展。
+现在，故事该结束了，我认为最合理的故事结局是：{user_choice}。
+
+请在前文的基础上（不要重复前文），编写小说的最后一幕。
+
+- 不要重复前文。
+- 出现人物时，给出人物的名字。
+- 积极地运用环境描写、人物描写等手法，让读者能够感受到你的故事世界。
+- 积极地运用修辞手法，比如比喻、拟人、排比、对偶、夸张等等。
+- 字数要求：最后一幕的字数少于1000字。
+"""
+
+
+from toolbox import CatchException, update_ui, update_ui_lastest_msg
+from crazy_functions.multi_stage.multi_stage_utils import GptAcademicGameBaseState
+from crazy_functions.crazy_utils import request_gpt_model_in_new_thread_with_ui_alive
+from request_llms.bridge_all import predict_no_ui_long_connection
+from crazy_functions.game_fns.game_utils import get_code_block, is_same_thing
+import random
+
+
+class MiniGame_ResumeStory(GptAcademicGameBaseState):
+    story_headstart = [
+        '先行者知道，他现在是全宇宙中唯一的一个人了。',
+        '深夜，一个年轻人穿过天安门广场向纪念堂走去。在二十二世纪编年史中，计算机把他的代号定为M102。',
+        '他知道，这最后一课要提前讲了。又一阵剧痛从肝部袭来，几乎使他晕厥过去。',
+        '在距地球五万光年的远方，在银河系的中心，一场延续了两万年的星际战争已接近尾声。那里的太空中渐渐隐现出一个方形区域，仿佛灿烂的群星的背景被剪出一个方口。',
+        '伊依一行三人乘坐一艘游艇在南太平洋上做吟诗航行，他们的目的地是南极，如果几天后能顺利到达那里，他们将钻出地壳去看诗云。',
+        '很多人生来就会莫名其妙地迷上一样东西，仿佛他的出生就是要和这东西约会似的，正是这样，圆圆迷上了肥皂泡。'
+    ]
+
+
+    def begin_game_step_0(self, prompt, chatbot, history):
+        # init game at step 0
+        self.headstart = random.choice(self.story_headstart)
+        self.story = []
+        chatbot.append(["互动写故事", f"这次的故事开头是：{self.headstart}"])
+        self.sys_prompt_ = '你是一个想象力丰富的杰出作家。正在与你的朋友互动，一起写故事，因此你每次写的故事段落应少于300字（结局除外）。'
+
+
+    def generate_story_image(self, story_paragraph):
+        try:
+            from crazy_functions.Image_Generate import gen_image
+            prompt_ = predict_no_ui_long_connection(inputs=story_paragraph, llm_kwargs=self.llm_kwargs, history=[], sys_prompt='你需要根据用户给出的小说段落，进行简短的环境描写。要求：80字以内。')
+            image_url, image_path = gen_image(self.llm_kwargs, prompt_, '512x512', model="dall-e-2", quality='standard', style='natural')
+            return f'<br/><div align="center"><img src="file={image_path}"></div>'
+        except:
+            return ''
+
+    def step(self, prompt, chatbot, history):
+
+        """
+        首先，处理游戏初始化等特殊情况
+        """
+        if self.step_cnt == 0:
+            self.begin_game_step_0(prompt, chatbot, history)
+            self.lock_plugin(chatbot)
+            self.cur_task = 'head_start'
+        else:
+            if prompt.strip() == 'exit' or prompt.strip() == '结束剧情':
+                # should we terminate game here?
+                self.delete_game = True
+                yield from update_ui_lastest_msg(lastmsg=f"游戏结束。", chatbot=chatbot, history=history, delay=0.)
+                return
+            if '剧情收尾' in prompt:
+                self.cur_task = 'story_terminate'
+            # # well, game resumes
+            # chatbot.append([prompt, ""])
+        # update ui, don't keep the user waiting
+        yield from update_ui(chatbot=chatbot, history=history)
+
+
+        """
+        处理游戏的主体逻辑
+        """
+        if self.cur_task == 'head_start':
+            """
+            这是游戏的第一步
+            """
+            inputs_ = prompts_hs.format(headstart=self.headstart)
+            history_ = []
+            story_paragraph = yield from request_gpt_model_in_new_thread_with_ui_alive(
+                inputs_, '故事开头', self.llm_kwargs,
+                chatbot, history_, self.sys_prompt_
+            )
+            self.story.append(story_paragraph)
+            # # 配图
+            yield from update_ui_lastest_msg(lastmsg=story_paragraph + '<br/>正在生成插图中 ...', chatbot=chatbot, history=history, delay=0.)
+            yield from update_ui_lastest_msg(lastmsg=story_paragraph + '<br/>'+ self.generate_story_image(story_paragraph), chatbot=chatbot, history=history, delay=0.)
+
+            # # 构建后续剧情引导
+            previously_on_story = ""
+            for s in self.story:
+                previously_on_story += s + '\n'
+            inputs_ = prompts_interact.format(previously_on_story=previously_on_story)
+            history_ = []
+            self.next_choices = yield from request_gpt_model_in_new_thread_with_ui_alive(
+                inputs_, '请在以下几种故事走向中，选择一种（当然，您也可以选择给出其他故事走向）：', self.llm_kwargs,
+                chatbot,
+                history_,
+                self.sys_prompt_
+            )
+            self.cur_task = 'user_choice'
+
+
+        elif self.cur_task == 'user_choice':
+            """
+            根据用户的提示，确定故事的下一步
+            """
+            if '请在以下几种故事走向中，选择一种' in chatbot[-1][0]: chatbot.pop(-1)
+            previously_on_story = ""
+            for s in self.story:
+                previously_on_story += s + '\n'
+            inputs_ = prompts_resume.format(previously_on_story=previously_on_story, choice=self.next_choices, user_choice=prompt)
+            history_ = []
+            story_paragraph = yield from request_gpt_model_in_new_thread_with_ui_alive(
+                inputs_, f'下一段故事（您的选择是：{prompt}）。', self.llm_kwargs,
+                chatbot, history_, self.sys_prompt_
+            )
+            self.story.append(story_paragraph)
+            # # 配图
+            yield from update_ui_lastest_msg(lastmsg=story_paragraph + '<br/>正在生成插图中 ...', chatbot=chatbot, history=history, delay=0.)
+            yield from update_ui_lastest_msg(lastmsg=story_paragraph + '<br/>'+ self.generate_story_image(story_paragraph), chatbot=chatbot, history=history, delay=0.)
+
+            # # 构建后续剧情引导
+            previously_on_story = ""
+            for s in self.story:
+                previously_on_story += s + '\n'
+            inputs_ = prompts_interact.format(previously_on_story=previously_on_story)
+            history_ = []
+            self.next_choices = yield from request_gpt_model_in_new_thread_with_ui_alive(
+                inputs_,
+                '请在以下几种故事走向中，选择一种。当然，您也可以给出您心中的其他故事走向。另外，如果您希望剧情立即收尾，请输入剧情走向，并以“剧情收尾”四个字提示程序。', self.llm_kwargs,
+                chatbot,
+                history_,
+                self.sys_prompt_
+            )
+            self.cur_task = 'user_choice'
+
+
+        elif self.cur_task == 'story_terminate':
+            """
+            根据用户的提示，确定故事的结局
+            """
+            previously_on_story = ""
+            for s in self.story:
+                previously_on_story += s + '\n'
+            inputs_ = prompts_terminate.format(previously_on_story=previously_on_story, user_choice=prompt)
+            history_ = []
+            story_paragraph = yield from request_gpt_model_in_new_thread_with_ui_alive(
+                inputs_, f'故事收尾（您的选择是：{prompt}）。', self.llm_kwargs,
+                chatbot, history_, self.sys_prompt_
+            )
+            # # 配图
+            yield from update_ui_lastest_msg(lastmsg=story_paragraph + '<br/>正在生成插图中 ...', chatbot=chatbot, history=history, delay=0.)
+            yield from update_ui_lastest_msg(lastmsg=story_paragraph + '<br/>'+ self.generate_story_image(story_paragraph), chatbot=chatbot, history=history, delay=0.)
+
+            # terminate game
+            self.delete_game = True
+            return

crazy_functions/game_fns/game_utils.py

@@ -0,0 +1,35 @@
+
+from crazy_functions.json_fns.pydantic_io import GptJsonIO, JsonStringError
+from request_llms.bridge_all import predict_no_ui_long_connection
+def get_code_block(reply):
+    import re
+    pattern = r"```([\s\S]*?)```" # regex pattern to match code blocks
+    matches = re.findall(pattern, reply) # find all code blocks in text
+    if len(matches) == 1:
+        return "```" + matches[0] + "```" #  code block
+    raise RuntimeError("GPT is not generating proper code.")
+
+def is_same_thing(a, b, llm_kwargs):
+    from pydantic import BaseModel, Field
+    class IsSameThing(BaseModel):
+        is_same_thing: bool = Field(description="determine whether two objects are same thing.", default=False)
+
+    def run_gpt_fn(inputs, sys_prompt, history=[]):
+        return predict_no_ui_long_connection(
+            inputs=inputs, llm_kwargs=llm_kwargs,
+            history=history, sys_prompt=sys_prompt, observe_window=[]
+        )
+
+    gpt_json_io = GptJsonIO(IsSameThing)
+    inputs_01 = "Identity whether the user input and the target is the same thing: \n target object: {a} \n user input object: {b} \n\n\n".format(a=a, b=b)
+    inputs_01 += "\n\n\n Note that the user may describe the target object with a different language, e.g. cat and 猫 are the same thing."
+    analyze_res_cot_01 = run_gpt_fn(inputs_01, "", [])
+
+    inputs_02 = inputs_01 + gpt_json_io.format_instructions
+    analyze_res = run_gpt_fn(inputs_02, "", [inputs_01, analyze_res_cot_01])
+
+    try:
+        res = gpt_json_io.generate_output_auto_repair(analyze_res, run_gpt_fn)
+        return res.is_same_thing
+    except JsonStringError as e:
+        return False

crazy_functions/gen_fns/gen_fns_shared.py

@@ -0,0 +1,70 @@
+import time
+import importlib
+from toolbox import trimmed_format_exc, gen_time_str, get_log_folder
+from toolbox import CatchException, update_ui, gen_time_str, trimmed_format_exc, is_the_upload_folder
+from toolbox import promote_file_to_downloadzone, get_log_folder, update_ui_lastest_msg
+import multiprocessing
+
+def get_class_name(class_string):
+    import re
+    # Use regex to extract the class name
+    class_name = re.search(r'class (\w+)\(', class_string).group(1)
+    return class_name
+
+def try_make_module(code, chatbot):
+    module_file = 'gpt_fn_' + gen_time_str().replace('-','_')
+    fn_path = f'{get_log_folder(plugin_name="gen_plugin_verify")}/{module_file}.py'
+    with open(fn_path, 'w', encoding='utf8') as f: f.write(code)
+    promote_file_to_downloadzone(fn_path, chatbot=chatbot)
+    class_name = get_class_name(code)
+    manager = multiprocessing.Manager()
+    return_dict = manager.dict()
+    p = multiprocessing.Process(target=is_function_successfully_generated, args=(fn_path, class_name, return_dict))
+    # only has 10 seconds to run
+    p.start(); p.join(timeout=10)
+    if p.is_alive(): p.terminate(); p.join()
+    p.close()
+    return return_dict["success"], return_dict['traceback']
+
+# check is_function_successfully_generated
+def is_function_successfully_generated(fn_path, class_name, return_dict):
+    return_dict['success'] = False
+    return_dict['traceback'] = ""
+    try:
+        # Create a spec for the module
+        module_spec = importlib.util.spec_from_file_location('example_module', fn_path)
+        # Load the module
+        example_module = importlib.util.module_from_spec(module_spec)
+        module_spec.loader.exec_module(example_module)
+        # Now you can use the module
+        some_class = getattr(example_module, class_name)
+        # Now you can create an instance of the class
+        instance = some_class()
+        return_dict['success'] = True
+        return
+    except:
+        return_dict['traceback'] = trimmed_format_exc()
+        return
+
+def subprocess_worker(code, file_path, return_dict):
+    return_dict['result'] = None
+    return_dict['success'] = False
+    return_dict['traceback'] = ""
+    try:
+        module_file = 'gpt_fn_' + gen_time_str().replace('-','_')
+        fn_path = f'{get_log_folder(plugin_name="gen_plugin_run")}/{module_file}.py'
+        with open(fn_path, 'w', encoding='utf8') as f: f.write(code)
+        class_name = get_class_name(code)
+        # Create a spec for the module
+        module_spec = importlib.util.spec_from_file_location('example_module', fn_path)
+        # Load the module
+        example_module = importlib.util.module_from_spec(module_spec)
+        module_spec.loader.exec_module(example_module)
+        # Now you can use the module
+        some_class = getattr(example_module, class_name)
+        # Now you can create an instance of the class
+        instance = some_class()
+        return_dict['result'] = instance.run(file_path)
+        return_dict['success'] = True
+    except:
+        return_dict['traceback'] = trimmed_format_exc()

crazy_functions/ipc_fns/mp.py

@@ -0,0 +1,37 @@
+import platform
+import pickle
+import multiprocessing
+
+def run_in_subprocess_wrapper_func(v_args):
+    func, args, kwargs, return_dict, exception_dict = pickle.loads(v_args)
+    import sys
+    try:
+        result = func(*args, **kwargs)
+        return_dict['result'] = result
+    except Exception as e:
+        exc_info = sys.exc_info()
+        exception_dict['exception'] = exc_info
+
+def run_in_subprocess_with_timeout(func, timeout=60):
+    if platform.system() == 'Linux':
+        def wrapper(*args, **kwargs):
+            return_dict = multiprocessing.Manager().dict()
+            exception_dict = multiprocessing.Manager().dict()
+            v_args = pickle.dumps((func, args, kwargs, return_dict, exception_dict))
+            process = multiprocessing.Process(target=run_in_subprocess_wrapper_func, args=(v_args,))
+            process.start()
+            process.join(timeout)
+            if process.is_alive():
+                process.terminate()
+                raise TimeoutError(f'功能单元{str(func)}未能在规定时间内完成任务')
+            process.close()
+            if 'exception' in exception_dict:
+                # ooops, the subprocess ran into an exception
+                exc_info = exception_dict['exception']
+                raise exc_info[1].with_traceback(exc_info[2])
+            if 'result' in return_dict.keys():
+                # If the subprocess ran successfully, return the result
+                return return_dict['result']
+        return wrapper
+    else:
+        return func

crazy_functions/json_fns/pydantic_io.py

@@ -0,0 +1,111 @@
+"""
+https://github.com/langchain-ai/langchain/blob/master/docs/extras/modules/model_io/output_parsers/pydantic.ipynb
+
+Example 1.
+
+# Define your desired data structure.
+class Joke(BaseModel):
+    setup: str = Field(description="question to set up a joke")
+    punchline: str = Field(description="answer to resolve the joke")
+
+    # You can add custom validation logic easily with Pydantic.
+    @validator("setup")
+    def question_ends_with_question_mark(cls, field):
+        if field[-1] != "?":
+            raise ValueError("Badly formed question!")
+        return field
+
+
+Example 2.
+
+# Here's another example, but with a compound typed field.
+class Actor(BaseModel):
+    name: str = Field(description="name of an actor")
+    film_names: List[str] = Field(description="list of names of films they starred in")
+"""
+
+import json, re
+from loguru import logger as logging
+
+PYDANTIC_FORMAT_INSTRUCTIONS = """The output should be formatted as a JSON instance that conforms to the JSON schema below.
+
+As an example, for the schema {{"properties": {{"foo": {{"title": "Foo", "description": "a list of strings", "type": "array", "items": {{"type": "string"}}}}}}, "required": ["foo"]}}
+the object {{"foo": ["bar", "baz"]}} is a well-formatted instance of the schema. The object {{"properties": {{"foo": ["bar", "baz"]}}}} is not well-formatted.
+
+Here is the output schema:
+```
+{schema}
+```"""
+
+
+PYDANTIC_FORMAT_INSTRUCTIONS_SIMPLE = """The output should be formatted as a JSON instance that conforms to the JSON schema below.
+```
+{schema}
+```"""
+
+class JsonStringError(Exception): ...
+
+class GptJsonIO():
+
+    def __init__(self, schema, example_instruction=True):
+        self.pydantic_object = schema
+        self.example_instruction = example_instruction
+        self.format_instructions = self.generate_format_instructions()
+
+    def generate_format_instructions(self):
+        schema = self.pydantic_object.schema()
+
+        # Remove extraneous fields.
+        reduced_schema = schema
+        if "title" in reduced_schema:
+            del reduced_schema["title"]
+        if "type" in reduced_schema:
+            del reduced_schema["type"]
+        # Ensure json in context is well-formed with double quotes.
+        schema_str = json.dumps(reduced_schema)
+        if self.example_instruction:
+            return PYDANTIC_FORMAT_INSTRUCTIONS.format(schema=schema_str)
+        else:
+            return PYDANTIC_FORMAT_INSTRUCTIONS_SIMPLE.format(schema=schema_str)
+
+    def generate_output(self, text):
+        # Greedy search for 1st json candidate.
+        match = re.search(
+            r"\{.*\}", text.strip(), re.MULTILINE | re.IGNORECASE | re.DOTALL
+        )
+        json_str = ""
+        if match: json_str = match.group()
+        json_object = json.loads(json_str, strict=False)
+        final_object = self.pydantic_object.parse_obj(json_object)
+        return final_object
+
+    def generate_repair_prompt(self, broken_json, error):
+        prompt = "Fix a broken json string.\n\n" + \
+                 "(1) The broken json string need to fix is: \n\n" + \
+                 "```" + "\n" + \
+                 broken_json + "\n" + \
+                 "```" + "\n\n" + \
+                 "(2) The error message is: \n\n" + \
+                 error + "\n\n" + \
+                "Now, fix this json string. \n\n"
+        return prompt
+
+    def generate_output_auto_repair(self, response, gpt_gen_fn):
+        """
+        response: string containing canidate json
+        gpt_gen_fn: gpt_gen_fn(inputs, sys_prompt)
+        """
+        try:
+            result = self.generate_output(response)
+        except Exception as e:
+            try:
+                logging.info(f'Repairing json：{response}')
+                repair_prompt = self.generate_repair_prompt(broken_json = response, error=repr(e))
+                result = self.generate_output(gpt_gen_fn(repair_prompt, self.format_instructions))
+                logging.info('Repaire json success.')
+            except Exception as e:
+                # 没辙了，放弃治疗
+                logging.info('Repaire json fail.')
+                raise JsonStringError('Cannot repair json.', str(e))
+        return result
+

crazy_functions/json_fns/select_tool.py

@@ -0,0 +1,26 @@
+from crazy_functions.json_fns.pydantic_io import GptJsonIO, JsonStringError
+
+def structure_output(txt, prompt, err_msg, run_gpt_fn, pydantic_cls):
+    gpt_json_io = GptJsonIO(pydantic_cls)
+    analyze_res = run_gpt_fn(
+        txt, 
+        sys_prompt=prompt + gpt_json_io.format_instructions
+    )
+    try:
+        friend = gpt_json_io.generate_output_auto_repair(analyze_res, run_gpt_fn)
+    except JsonStringError as e:
+        return None, err_msg
+
+    err_msg = ""
+    return friend, err_msg
+
+
+def select_tool(prompt, run_gpt_fn, pydantic_cls):
+    pydantic_cls_instance, err_msg = structure_output(
+        txt=prompt,
+        prompt="根据提示, 分析应该调用哪个工具函数\n\n",
+        err_msg=f"不能理解该联系人",
+        run_gpt_fn=run_gpt_fn,
+        pydantic_cls=pydantic_cls
+    )
+    return pydantic_cls_instance, err_msg

crazy_functions/latex_fns/latex_actions.py

@@ -1,13 +1,17 @@
-from toolbox import update_ui, update_ui_lastest_msg    # 刷新Gradio前端界面
-from toolbox import zip_folder, objdump, objload, promote_file_to_downloadzone
-from .latex_toolbox import PRESERVE, TRANSFORM
-from .latex_toolbox import set_forbidden_text, set_forbidden_text_begin_end, set_forbidden_text_careful_brace
-from .latex_toolbox import reverse_forbidden_text_careful_brace, reverse_forbidden_text, convert_to_linklist, post_process
-from .latex_toolbox import fix_content, find_main_tex_file, merge_tex_files, compile_latex_with_timeout
-
-import os, shutil
+import os
 import re
+import shutil
 import numpy as np
+from loguru import logger
+from toolbox import update_ui, update_ui_lastest_msg, get_log_folder
+from toolbox import get_conf, promote_file_to_downloadzone
+from crazy_functions.latex_fns.latex_toolbox import PRESERVE, TRANSFORM
+from crazy_functions.latex_fns.latex_toolbox import set_forbidden_text, set_forbidden_text_begin_end, set_forbidden_text_careful_brace
+from crazy_functions.latex_fns.latex_toolbox import reverse_forbidden_text_careful_brace, reverse_forbidden_text, convert_to_linklist, post_process
+from crazy_functions.latex_fns.latex_toolbox import fix_content, find_main_tex_file, merge_tex_files, compile_latex_with_timeout
+from crazy_functions.latex_fns.latex_toolbox import find_title_and_abs
+from crazy_functions.latex_fns.latex_pickle_io import objdump, objload
+
 
 pj = os.path.join
 
@@ -90,7 +94,18 @@ class LatexPaperSplit():
             "项目Github地址 \\url{https://github.com/binary-husky/gpt_academic/}。"
         # 请您不要删除或修改这行警告，除非您是论文的原作者（如果您是论文原作者，欢迎加REAME中的QQ联系开发者）
         self.msg_declare = "为了防止大语言模型的意外谬误产生扩散影响，禁止移除或修改此警告。}}\\\\"
+        self.title = "unknown"
+        self.abstract = "unknown"
 
+    def read_title_and_abstract(self, txt):
+        try:
+            title, abstract = find_title_and_abs(txt)
+            if title is not None:
+                self.title = title.replace('\n', ' ').replace('\\\\', ' ').replace('  ', '').replace('  ', '')
+            if abstract is not None:
+                self.abstract = abstract.replace('\n', ' ').replace('\\\\', ' ').replace('  ', '').replace('  ', '')
+        except:
+            pass
 
     def merge_result(self, arr, mode, msg, buggy_lines=[], buggy_line_surgery_n_lines=10):
         """
@@ -163,9 +178,8 @@ class LatexPaperFileGroup():
         self.sp_file_contents = []
         self.sp_file_index = []
         self.sp_file_tag = []
-
         # count_token
-        from request_llm.bridge_all import model_info
+        from request_llms.bridge_all import model_info
         enc = model_info["gpt-3.5-turbo"]['tokenizer']
         def get_token_num(txt): return len(enc.encode(txt, disallowed_special=()))
         self.get_token_num = get_token_num
@@ -180,13 +194,12 @@ class LatexPaperFileGroup():
                 self.sp_file_index.append(index)
                 self.sp_file_tag.append(self.file_paths[index])
             else:
-                from ..crazy_utils import breakdown_txt_to_satisfy_token_limit_for_pdf
-                segments = breakdown_txt_to_satisfy_token_limit_for_pdf(file_content, self.get_token_num, max_token_limit)
+                from crazy_functions.pdf_fns.breakdown_txt import breakdown_text_to_satisfy_token_limit
+                segments = breakdown_text_to_satisfy_token_limit(file_content, max_token_limit)
                 for j, segment in enumerate(segments):
                     self.sp_file_contents.append(segment)
                     self.sp_file_index.append(index)
                     self.sp_file_tag.append(self.file_paths[index] + f".part-{j}.tex")
-        print('Segmentation: done')
 
     def merge_result(self):
         self.file_result = ["" for _ in range(len(self.file_paths))]
@@ -234,8 +247,8 @@ def Latex精细分解与转化(file_manifest, project_folder, llm_kwargs, plugin
     chatbot.append((f"Latex文件融合完成", f'[Local Message] 正在精细切分latex文件，这需要一段时间计算，文档越长耗时越长，请耐心等待。'))
     yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
     lps = LatexPaperSplit()
+    lps.read_title_and_abstract(merged_content)
     res = lps.split(merged_content, project_folder, opts) # 消耗时间的函数
-
     #  <-------- 拆分过长的latex片段 ---------->
     pfg = LatexPaperFileGroup()
     for index, r in enumerate(res):
@@ -256,12 +269,19 @@ def Latex精细分解与转化(file_manifest, project_folder, llm_kwargs, plugin
 
     else:
         #  <-------- gpt 多线程请求 ---------->
+        history_array = [[""] for _ in range(n_split)]
+        # LATEX_EXPERIMENTAL, = get_conf('LATEX_EXPERIMENTAL')
+        # if LATEX_EXPERIMENTAL:
+        #     paper_meta = f"The paper you processing is `{lps.title}`, a part of the abstraction is `{lps.abstract}`"
+        #     paper_meta_max_len = 888
+        #     history_array = [[ paper_meta[:paper_meta_max_len] + '...',  "Understand, what should I do?"] for _ in range(n_split)]
+
         gpt_response_collection = yield from request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
             inputs_array=inputs_array,
             inputs_show_user_array=inputs_show_user_array,
             llm_kwargs=llm_kwargs,
             chatbot=chatbot,
-            history_array=[[""] for _ in range(n_split)],
+            history_array=history_array,
             sys_prompt_array=sys_prompt_array,
             # max_workers=5,  # 并行任务数量限制, 最多同时执行5个, 其他的排队等待
             scroller_max_len = 40
@@ -305,7 +325,7 @@ def remove_buggy_lines(file_path, log_path, tex_name, tex_name_pure, n_fix, work
         buggy_lines = [int(l) for l in buggy_lines]
         buggy_lines = sorted(buggy_lines)
         buggy_line = buggy_lines[0]-1
-        print("reversing tex line that has errors", buggy_line)
+        logger.warning("reversing tex line that has errors", buggy_line)
 
         # 重组，逆转出错的段落
         if buggy_line not in fixed_line:
@@ -319,7 +339,7 @@ def remove_buggy_lines(file_path, log_path, tex_name, tex_name_pure, n_fix, work
 
         return True, f"{tex_name_pure}_fix_{n_fix}", buggy_lines
     except:
-        print("Fatal error occurred, but we cannot identify error, please download zip, read latex log, and compile manually.")
+        logger.error("Fatal error occurred, but we cannot identify error, please download zip, read latex log, and compile manually.")
         return False, -1, [-1]
 
 
@@ -362,8 +382,8 @@ def 编译Latex(chatbot, history, main_file_original, main_file_modified, work_f
 
             if mode!='translate_zh':
                 yield from update_ui_lastest_msg(f'尝试第 {n_fix}/{max_try} 次编译, 使用latexdiff生成论文转化前后对比 ...', chatbot, history) # 刷新Gradio前端界面
-                print(    f'latexdiff --encoding=utf8 --append-safecmd=subfile {work_folder_original}/{main_file_original}.tex  {work_folder_modified}/{main_file_modified}.tex --flatten > {work_folder}/merge_diff.tex')
-                ok = compile_latex_with_timeout(f'latexdiff --encoding=utf8 --append-safecmd=subfile {work_folder_original}/{main_file_original}.tex  {work_folder_modified}/{main_file_modified}.tex --flatten > {work_folder}/merge_diff.tex')
+                logger.info(    f'latexdiff --encoding=utf8 --append-safecmd=subfile {work_folder_original}/{main_file_original}.tex  {work_folder_modified}/{main_file_modified}.tex --flatten > {work_folder}/merge_diff.tex')
+                ok = compile_latex_with_timeout(f'latexdiff --encoding=utf8 --append-safecmd=subfile {work_folder_original}/{main_file_original}.tex  {work_folder_modified}/{main_file_modified}.tex --flatten > {work_folder}/merge_diff.tex', os.getcwd())
 
                 yield from update_ui_lastest_msg(f'尝试第 {n_fix}/{max_try} 次编译, 正在编译对比PDF ...', chatbot, history)   # 刷新Gradio前端界面
                 ok = compile_latex_with_timeout(f'pdflatex  -interaction=batchmode -file-line-error merge_diff.tex', work_folder)
@@ -385,7 +405,7 @@ def 编译Latex(chatbot, history, main_file_original, main_file_modified, work_f
             result_pdf = pj(work_folder_modified, f'merge_diff.pdf')    # get pdf path
             promote_file_to_downloadzone(result_pdf, rename_file=None, chatbot=chatbot)  # promote file to web UI
         if modified_pdf_success:
-            yield from update_ui_lastest_msg(f'转化PDF编译已经成功, 即将退出 ...', chatbot, history)    # 刷新Gradio前端界面
+            yield from update_ui_lastest_msg(f'转化PDF编译已经成功, 正在尝试生成对比PDF, 请稍候 ...', chatbot, history)    # 刷新Gradio前端界面
             result_pdf = pj(work_folder_modified, f'{main_file_modified}.pdf') # get pdf path
             origin_pdf = pj(work_folder_original, f'{main_file_original}.pdf') # get pdf path
             if os.path.exists(pj(work_folder, '..', 'translation')):
@@ -397,8 +417,11 @@ def 编译Latex(chatbot, history, main_file_original, main_file_modified, work_f
                     from .latex_toolbox import merge_pdfs
                     concat_pdf = pj(work_folder_modified, f'comparison.pdf')
                     merge_pdfs(origin_pdf, result_pdf, concat_pdf)
+                    if os.path.exists(pj(work_folder, '..', 'translation')):
+                        shutil.copyfile(concat_pdf, pj(work_folder, '..', 'translation', 'comparison.pdf'))
                     promote_file_to_downloadzone(concat_pdf, rename_file=None, chatbot=chatbot)  # promote file to web UI
                 except Exception as e:
+                    logger.error(e)
                     pass
             return True # 成功啦
         else:
@@ -423,7 +446,7 @@ def write_html(sp_file_contents, sp_file_result, chatbot, project_folder):
     # write html
     try:
         import shutil
-        from ..crazy_utils import construct_html
+        from crazy_functions.pdf_fns.report_gen_html import construct_html
         from toolbox import gen_time_str
         ch = construct_html()
         orig = ""
@@ -439,9 +462,9 @@ def write_html(sp_file_contents, sp_file_result, chatbot, project_folder):
                 trans = k
                 ch.add_row(a=orig, b=trans)
         create_report_file_name = f"{gen_time_str()}.trans.html"
-        ch.save_file(create_report_file_name)
-        shutil.copyfile(pj('./gpt_log/', create_report_file_name), pj(project_folder, create_report_file_name))
-        promote_file_to_downloadzone(file=f'./gpt_log/{create_report_file_name}', chatbot=chatbot)
+        res = ch.save_file(create_report_file_name)
+        shutil.copyfile(res, pj(project_folder, create_report_file_name))
+        promote_file_to_downloadzone(file=res, chatbot=chatbot)
     except:
         from toolbox import trimmed_format_exc
-        print('writing html result failed:', trimmed_format_exc())
+        logger.error('writing html result failed:', trimmed_format_exc())

crazy_functions/latex_fns/latex_pickle_io.py

@@ -0,0 +1,46 @@
+import pickle
+
+
+class SafeUnpickler(pickle.Unpickler):
+
+    def get_safe_classes(self):
+        from crazy_functions.latex_fns.latex_actions import LatexPaperFileGroup, LatexPaperSplit
+        from crazy_functions.latex_fns.latex_toolbox import LinkedListNode
+        # 定义允许的安全类
+        safe_classes = {
+            # 在这里添加其他安全的类
+            'LatexPaperFileGroup': LatexPaperFileGroup,
+            'LatexPaperSplit': LatexPaperSplit,
+            'LinkedListNode': LinkedListNode,
+        }
+        return safe_classes
+
+    def find_class(self, module, name):
+        # 只允许特定的类进行反序列化
+        self.safe_classes = self.get_safe_classes()
+        match_class_name = None
+        for class_name in self.safe_classes.keys():
+            if (class_name in f'{module}.{name}'):
+                match_class_name = class_name
+        if module == 'numpy' or module.startswith('numpy.'):
+            return super().find_class(module, name)
+        if match_class_name is not None:
+            return self.safe_classes[match_class_name]
+        # 如果尝试加载未授权的类，则抛出异常
+        raise pickle.UnpicklingError(f"Attempted to deserialize unauthorized class '{name}' from module '{module}'")
+
+def objdump(obj, file="objdump.tmp"):
+
+    with open(file, "wb+") as f:
+        pickle.dump(obj, f)
+    return
+
+
+def objload(file="objdump.tmp"):
+    import os
+
+    if not os.path.exists(file):
+        return
+    with open(file, "rb") as f:
+        unpickler = SafeUnpickler(f)
+        return unpickler.load()

crazy_functions/latex_fns/latex_toolbox.py

@@ -1,15 +1,20 @@
-import os, shutil
+import os
 import re
+import shutil
 import numpy as np
+from loguru import logger
+
 PRESERVE = 0
 TRANSFORM = 1
 
 pj = os.path.join
 
-class LinkedListNode():
+
+class LinkedListNode:
     """
     Linked List Node
     """
+
     def __init__(self, string, preserve=True) -> None:
         self.string = string
         self.preserve = preserve
@@ -18,19 +23,22 @@ class LinkedListNode():
         # self.begin_line = 0
         # self.begin_char = 0
 
+
 def convert_to_linklist(text, mask):
     root = LinkedListNode("", preserve=True)
     current_node = root
     for c, m, i in zip(text, mask, range(len(text))):
-        if (m==PRESERVE and current_node.preserve) \
-            or (m==TRANSFORM and not current_node.preserve):
+        if (m == PRESERVE and current_node.preserve) or (
+            m == TRANSFORM and not current_node.preserve
+        ):
             # add
             current_node.string += c
         else:
-            current_node.next = LinkedListNode(c, preserve=(m==PRESERVE))
+            current_node.next = LinkedListNode(c, preserve=(m == PRESERVE))
             current_node = current_node.next
     return root
 
+
 def post_process(root):
     # 修复括号
     node = root
@@ -38,21 +46,24 @@ def post_process(root):
         string = node.string
         if node.preserve:
             node = node.next
-            if node is None: break
+            if node is None:
+                break
             continue
+
         def break_check(string):
             str_stack = [""]  # (lv, index)
             for i, c in enumerate(string):
-                if c == '{':
-                    str_stack.append('{')
-                elif c == '}':
+                if c == "{":
+                    str_stack.append("{")
+                elif c == "}":
                     if len(str_stack) == 1:
-                        print('stack fix')
+                        logger.warning("fixing brace error")
                         return i
                     str_stack.pop(-1)
                 else:
                     str_stack[-1] += c
             return -1
+
         bp = break_check(string)
 
         if bp == -1:
@@ -69,51 +80,66 @@ def post_process(root):
             node.next = q
 
         node = node.next
-        if node is None: break
+        if node is None:
+            break
 
     # 屏蔽空行和太短的句子
     node = root
     while True:
-        if len(node.string.strip('\n').strip(''))==0: node.preserve = True
-        if len(node.string.strip('\n').strip(''))<42: node.preserve = True
+        if len(node.string.strip("\n").strip("")) == 0:
+            node.preserve = True
+        if len(node.string.strip("\n").strip("")) < 42:
+            node.preserve = True
         node = node.next
-        if node is None: break
+        if node is None:
+            break
     node = root
     while True:
         if node.next and node.preserve and node.next.preserve:
             node.string += node.next.string
             node.next = node.next.next
         node = node.next
-        if node is None: break
+        if node is None:
+            break
 
     # 将前后断行符脱离
     node = root
     prev_node = None
     while True:
         if not node.preserve:
-            lstriped_ = node.string.lstrip().lstrip('\n')
-            if (prev_node is not None) and (prev_node.preserve) and (len(lstriped_)!=len(node.string)):
-                prev_node.string += node.string[:-len(lstriped_)]
+            lstriped_ = node.string.lstrip().lstrip("\n")
+            if (
+                (prev_node is not None)
+                and (prev_node.preserve)
+                and (len(lstriped_) != len(node.string))
+            ):
+                prev_node.string += node.string[: -len(lstriped_)]
                 node.string = lstriped_
-            rstriped_ = node.string.rstrip().rstrip('\n')
-            if (node.next is not None) and (node.next.preserve) and (len(rstriped_)!=len(node.string)):
-                node.next.string = node.string[len(rstriped_):] + node.next.string
+            rstriped_ = node.string.rstrip().rstrip("\n")
+            if (
+                (node.next is not None)
+                and (node.next.preserve)
+                and (len(rstriped_) != len(node.string))
+            ):
+                node.next.string = node.string[len(rstriped_) :] + node.next.string
                 node.string = rstriped_
-        # =====
+        # =-=-=
         prev_node = node
         node = node.next
-        if node is None: break
+        if node is None:
+            break
 
     # 标注节点的行数范围
     node = root
     n_line = 0
     expansion = 2
     while True:
-        n_l = node.string.count('\n')
-        node.range = [n_line-expansion, n_line+n_l+expansion]   # 失败时，扭转的范围
-        n_line = n_line+n_l
+        n_l = node.string.count("\n")
+        node.range = [n_line - expansion, n_line + n_l + expansion]  # 失败时，扭转的范围
+        n_line = n_line + n_l
         node = node.next
-        if node is None: break
+        if node is None:
+            break
     return root
 
 
@@ -131,12 +157,14 @@ def set_forbidden_text(text, mask, pattern, flags=0):
     you can mask out (mask = PRESERVE so that text become untouchable for GPT)
     everything between "\begin{equation}" and "\end{equation}"
     """
-    if isinstance(pattern, list): pattern = '|'.join(pattern)
+    if isinstance(pattern, list):
+        pattern = "|".join(pattern)
     pattern_compile = re.compile(pattern, flags)
     for res in pattern_compile.finditer(text):
-        mask[res.span()[0]:res.span()[1]] = PRESERVE
+        mask[res.span()[0] : res.span()[1]] = PRESERVE
     return text, mask
 
+
 def reverse_forbidden_text(text, mask, pattern, flags=0, forbid_wrapper=True):
     """
     Move area out of preserve area (make text editable for GPT)
@@ -144,17 +172,19 @@ def reverse_forbidden_text(text, mask, pattern, flags=0, forbid_wrapper=True):
     e.g.
     \begin{abstract} blablablablablabla. \end{abstract}
     """
-    if isinstance(pattern, list): pattern = '|'.join(pattern)
+    if isinstance(pattern, list):
+        pattern = "|".join(pattern)
     pattern_compile = re.compile(pattern, flags)
     for res in pattern_compile.finditer(text):
         if not forbid_wrapper:
-            mask[res.span()[0]:res.span()[1]] = TRANSFORM
+            mask[res.span()[0] : res.span()[1]] = TRANSFORM
         else:
-            mask[res.regs[0][0]: res.regs[1][0]] = PRESERVE   # '\\begin{abstract}'
-            mask[res.regs[1][0]: res.regs[1][1]] = TRANSFORM   # abstract
-            mask[res.regs[1][1]: res.regs[0][1]] = PRESERVE   # abstract
+            mask[res.regs[0][0] : res.regs[1][0]] = PRESERVE  # '\\begin{abstract}'
+            mask[res.regs[1][0] : res.regs[1][1]] = TRANSFORM  # abstract
+            mask[res.regs[1][1] : res.regs[0][1]] = PRESERVE  # abstract
     return text, mask
 
+
 def set_forbidden_text_careful_brace(text, mask, pattern, flags=0):
     """
     Add a preserve text area in this paper (text become untouchable for GPT).
@@ -166,16 +196,22 @@ def set_forbidden_text_careful_brace(text, mask, pattern, flags=0):
     for res in pattern_compile.finditer(text):
         brace_level = -1
         p = begin = end = res.regs[0][0]
-        for _ in range(1024*16):
-            if text[p] == '}' and brace_level == 0: break
-            elif text[p] == '}':  brace_level -= 1
-            elif text[p] == '{':  brace_level += 1
+        for _ in range(1024 * 16):
+            if text[p] == "}" and brace_level == 0:
+                break
+            elif text[p] == "}":
+                brace_level -= 1
+            elif text[p] == "{":
+                brace_level += 1
             p += 1
-        end = p+1
+        end = p + 1
         mask[begin:end] = PRESERVE
     return text, mask
 
-def reverse_forbidden_text_careful_brace(text, mask, pattern, flags=0, forbid_wrapper=True):
+
+def reverse_forbidden_text_careful_brace(
+    text, mask, pattern, flags=0, forbid_wrapper=True
+):
     """
     Move area out of preserve area (make text editable for GPT)
     count the number of the braces so as to catch compelete text area.
@@ -186,39 +222,57 @@ def reverse_forbidden_text_careful_brace(text, mask, pattern, flags=0, forbid_wr
     for res in pattern_compile.finditer(text):
         brace_level = 0
         p = begin = end = res.regs[1][0]
-        for _ in range(1024*16):
-            if text[p] == '}' and brace_level == 0: break
-            elif text[p] == '}':  brace_level -= 1
-            elif text[p] == '{':  brace_level += 1
+        for _ in range(1024 * 16):
+            if text[p] == "}" and brace_level == 0:
+                break
+            elif text[p] == "}":
+                brace_level -= 1
+            elif text[p] == "{":
+                brace_level += 1
             p += 1
         end = p
         mask[begin:end] = TRANSFORM
         if forbid_wrapper:
-            mask[res.regs[0][0]:begin] = PRESERVE
-            mask[end:res.regs[0][1]] = PRESERVE
+            mask[res.regs[0][0] : begin] = PRESERVE
+            mask[end : res.regs[0][1]] = PRESERVE
     return text, mask
 
+
 def set_forbidden_text_begin_end(text, mask, pattern, flags=0, limit_n_lines=42):
     """
     Find all \begin{} ... \end{} text block that with less than limit_n_lines lines.
     Add it to preserve area
     """
     pattern_compile = re.compile(pattern, flags)
+
     def search_with_line_limit(text, mask):
         for res in pattern_compile.finditer(text):
             cmd = res.group(1)  # begin{what}
             this = res.group(2)  # content between begin and end
-            this_mask = mask[res.regs[2][0]:res.regs[2][1]]
-            white_list = ['document', 'abstract', 'lemma', 'definition', 'sproof', 
-                          'em', 'emph', 'textit', 'textbf', 'itemize', 'enumerate']
-            if (cmd in white_list) or this.count('\n') >= limit_n_lines: # use a magical number 42
+            this_mask = mask[res.regs[2][0] : res.regs[2][1]]
+            white_list = [
+                "document",
+                "abstract",
+                "lemma",
+                "definition",
+                "sproof",
+                "em",
+                "emph",
+                "textit",
+                "textbf",
+                "itemize",
+                "enumerate",
+            ]
+            if (cmd in white_list) or this.count(
+                "\n"
+            ) >= limit_n_lines:  # use a magical number 42
                 this, this_mask = search_with_line_limit(this, this_mask)
-                mask[res.regs[2][0]:res.regs[2][1]] = this_mask
+                mask[res.regs[2][0] : res.regs[2][1]] = this_mask
             else:
-                mask[res.regs[0][0]:res.regs[0][1]] = PRESERVE
+                mask[res.regs[0][0] : res.regs[0][1]] = PRESERVE
         return text, mask
-    return search_with_line_limit(text, mask) 
 
+    return search_with_line_limit(text, mask)
 
 
 """
@@ -227,6 +281,7 @@ Latex Merge File
 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
 """
 
+
 def find_main_tex_file(file_manifest, mode):
     """
     在多Tex文档中，寻找主文件，必须包含documentclass，返回找到的第一个。
@@ -234,28 +289,38 @@ def find_main_tex_file(file_manifest, mode):
     """
     canidates = []
     for texf in file_manifest:
-        if os.path.basename(texf).startswith('merge'):
+        if os.path.basename(texf).startswith("merge"):
             continue
-        with open(texf, 'r', encoding='utf8', errors='ignore') as f:
+        with open(texf, "r", encoding="utf8", errors="ignore") as f:
             file_content = f.read()
-        if r'\documentclass' in file_content:
+        if r"\documentclass" in file_content:
             canidates.append(texf)
         else:
             continue
 
     if len(canidates) == 0:
-        raise RuntimeError('无法找到一个主Tex文件（包含documentclass关键字）')
+        raise RuntimeError("无法找到一个主Tex文件（包含documentclass关键字）")
     elif len(canidates) == 1:
         return canidates[0]
     else:  # if len(canidates) >= 2 通过一些Latex模板中常见（但通常不会出现在正文）的单词，对不同latex源文件扣分，取评分最高者返回
         canidates_score = []
         # 给出一些判定模板文档的词作为扣分项
-        unexpected_words = ['\LaTeX', 'manuscript', 'Guidelines', 'font', 'citations', 'rejected', 'blind review', 'reviewers']
-        expected_words = ['\input', '\ref', '\cite']
+        unexpected_words = [
+            "\\LaTeX",
+            "manuscript",
+            "Guidelines",
+            "font",
+            "citations",
+            "rejected",
+            "blind review",
+            "reviewers",
+        ]
+        expected_words = ["\\input", "\\ref", "\\cite"]
         for texf in canidates:
             canidates_score.append(0)
-            with open(texf, 'r', encoding='utf8', errors='ignore') as f:
+            with open(texf, "r", encoding="utf8", errors="ignore") as f:
                 file_content = f.read()
+                file_content = rm_comments(file_content)
             for uw in unexpected_words:
                 if uw in file_content:
                     canidates_score[-1] -= 1
@@ -265,6 +330,7 @@ def find_main_tex_file(file_manifest, mode):
         select = np.argmax(canidates_score)  # 取评分最高者返回
         return canidates[select]
 
+
 def rm_comments(main_file):
     new_file_remove_comment_lines = []
     for l in main_file.splitlines():
@@ -273,23 +339,39 @@ def rm_comments(main_file):
             pass
         else:
             new_file_remove_comment_lines.append(l)
-    main_file = '\n'.join(new_file_remove_comment_lines)
+    main_file = "\n".join(new_file_remove_comment_lines)
     # main_file = re.sub(r"\\include{(.*?)}", r"\\input{\1}", main_file)  # 将 \include 命令转换为 \input 命令
-    main_file = re.sub(r'(?<!\\)%.*', '', main_file)  # 使用正则表达式查找半行注释, 并替换为空字符串
+    main_file = re.sub(r"(?<!\\)%.*", "", main_file)  # 使用正则表达式查找半行注释, 并替换为空字符串
     return main_file
 
+
 def find_tex_file_ignore_case(fp):
     dir_name = os.path.dirname(fp)
     base_name = os.path.basename(fp)
-    if not base_name.endswith('.tex'): base_name+='.tex'
-    if os.path.exists(pj(dir_name, base_name)): return pj(dir_name, base_name)
-    # go case in-sensitive
+    # 如果输入的文件路径是正确的
+    if os.path.isfile(pj(dir_name, base_name)):
+        return pj(dir_name, base_name)
+    # 如果不正确，试着加上.tex后缀试试
+    if not base_name.endswith(".tex"):
+        base_name += ".tex"
+    if os.path.isfile(pj(dir_name, base_name)):
+        return pj(dir_name, base_name)
+    # 如果还找不到，解除大小写限制，再试一次
     import glob
-    for f in glob.glob(dir_name+'/*.tex'):
+
+    for f in glob.glob(dir_name + "/*.tex"):
         base_name_s = os.path.basename(fp)
-        if base_name_s.lower() == base_name.lower(): return f
+        base_name_f = os.path.basename(f)
+        if base_name_s.lower() == base_name_f.lower():
+            return f
+        # 试着加上.tex后缀试试
+        if not base_name_s.endswith(".tex"):
+            base_name_s += ".tex"
+        if base_name_s.lower() == base_name_f.lower():
+            return f
     return None
 
+
 def merge_tex_files_(project_foler, main_file, mode):
     """
     Merge Tex project recrusively
@@ -298,15 +380,53 @@ def merge_tex_files_(project_foler, main_file, mode):
     for s in reversed([q for q in re.finditer(r"\\input\{(.*?)\}", main_file, re.M)]):
         f = s.group(1)
         fp = os.path.join(project_foler, f)
-        fp = find_tex_file_ignore_case(fp)
-        if fp:
-            with open(fp, 'r', encoding='utf-8', errors='replace') as fx: c = fx.read()
+        fp_ = find_tex_file_ignore_case(fp)
+        if fp_:
+            try:
+                with open(fp_, "r", encoding="utf-8", errors="replace") as fx:
+                    c = fx.read()
+            except:
+                c = f"\n\nWarning from GPT-Academic: LaTex source file is missing!\n\n"
         else:
-            raise RuntimeError(f'找不到{fp}，Tex源文件缺失！')
+            raise RuntimeError(f"找不到{fp}，Tex源文件缺失！")
         c = merge_tex_files_(project_foler, c, mode)
-        main_file = main_file[:s.span()[0]] + c + main_file[s.span()[1]:]
+        main_file = main_file[: s.span()[0]] + c + main_file[s.span()[1] :]
     return main_file
 
+
+def find_title_and_abs(main_file):
+    def extract_abstract_1(text):
+        pattern = r"\\abstract\{(.*?)\}"
+        match = re.search(pattern, text, re.DOTALL)
+        if match:
+            return match.group(1)
+        else:
+            return None
+
+    def extract_abstract_2(text):
+        pattern = r"\\begin\{abstract\}(.*?)\\end\{abstract\}"
+        match = re.search(pattern, text, re.DOTALL)
+        if match:
+            return match.group(1)
+        else:
+            return None
+
+    def extract_title(string):
+        pattern = r"\\title\{(.*?)\}"
+        match = re.search(pattern, string, re.DOTALL)
+
+        if match:
+            return match.group(1)
+        else:
+            return None
+
+    abstract = extract_abstract_1(main_file)
+    if abstract is None:
+        abstract = extract_abstract_2(main_file)
+    title = extract_title(main_file)
+    return title, abstract
+
+
 def merge_tex_files(project_foler, main_file, mode):
     """
     Merge Tex project recrusively
@@ -316,33 +436,91 @@ def merge_tex_files(project_foler, main_file, mode):
     main_file = merge_tex_files_(project_foler, main_file, mode)
     main_file = rm_comments(main_file)
 
-    if mode == 'translate_zh':
+    if mode == "translate_zh":
         # find paper documentclass
-        pattern = re.compile(r'\\documentclass.*\n')
+        pattern = re.compile(r"\\documentclass.*\n")
         match = pattern.search(main_file)
         assert match is not None, "Cannot find documentclass statement!"
         position = match.end()
-        add_ctex = '\\usepackage{ctex}\n'
-        add_url = '\\usepackage{url}\n' if '{url}' not in main_file else ''
+        add_ctex = "\\usepackage{ctex}\n"
+        add_url = "\\usepackage{url}\n" if "{url}" not in main_file else ""
         main_file = main_file[:position] + add_ctex + add_url + main_file[position:]
         # fontset=windows
         import platform
-        main_file = re.sub(r"\\documentclass\[(.*?)\]{(.*?)}", r"\\documentclass[\1,fontset=windows,UTF8]{\2}",main_file)
-        main_file = re.sub(r"\\documentclass{(.*?)}", r"\\documentclass[fontset=windows,UTF8]{\1}",main_file)
+
+        main_file = re.sub(
+            r"\\documentclass\[(.*?)\]{(.*?)}",
+            r"\\documentclass[\1,fontset=windows,UTF8]{\2}",
+            main_file,
+        )
+        main_file = re.sub(
+            r"\\documentclass{(.*?)}",
+            r"\\documentclass[fontset=windows,UTF8]{\1}",
+            main_file,
+        )
         # find paper abstract
-        pattern_opt1 = re.compile(r'\\begin\{abstract\}.*\n')
+        pattern_opt1 = re.compile(r"\\begin\{abstract\}.*\n")
         pattern_opt2 = re.compile(r"\\abstract\{(.*?)\}", flags=re.DOTALL)
         match_opt1 = pattern_opt1.search(main_file)
         match_opt2 = pattern_opt2.search(main_file)
-        assert (match_opt1 is not None) or (match_opt2 is not None), "Cannot find paper abstract section!"
+        if (match_opt1 is None) and (match_opt2 is None):
+            # "Cannot find paper abstract section!"
+            main_file = insert_abstract(main_file)
+        match_opt1 = pattern_opt1.search(main_file)
+        match_opt2 = pattern_opt2.search(main_file)
+        assert (match_opt1 is not None) or (
+            match_opt2 is not None
+        ), "Cannot find paper abstract section!"
     return main_file
 
 
+insert_missing_abs_str = r"""
+\begin{abstract}
+The GPT-Academic program cannot find abstract section in this paper.
+\end{abstract}
+"""
+
+
+def insert_abstract(tex_content):
+    if "\\maketitle" in tex_content:
+        # find the position of "\maketitle"
+        find_index = tex_content.index("\\maketitle")
+        # find the nearest ending line
+        end_line_index = tex_content.find("\n", find_index)
+        # insert "abs_str" on the next line
+        modified_tex = (
+            tex_content[: end_line_index + 1]
+            + "\n\n"
+            + insert_missing_abs_str
+            + "\n\n"
+            + tex_content[end_line_index + 1 :]
+        )
+        return modified_tex
+    elif r"\begin{document}" in tex_content:
+        # find the position of "\maketitle"
+        find_index = tex_content.index(r"\begin{document}")
+        # find the nearest ending line
+        end_line_index = tex_content.find("\n", find_index)
+        # insert "abs_str" on the next line
+        modified_tex = (
+            tex_content[: end_line_index + 1]
+            + "\n\n"
+            + insert_missing_abs_str
+            + "\n\n"
+            + tex_content[end_line_index + 1 :]
+        )
+        return modified_tex
+    else:
+        return tex_content
+
+
 """
 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
 Post process
 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
 """
+
+
 def mod_inbraket(match):
     """
     为啥chatgpt会把cite里面的逗号换成中文逗号呀
@@ -351,11 +529,12 @@ def mod_inbraket(match):
     cmd = match.group(1)
     str_to_modify = match.group(2)
     # modify the matched string
-    str_to_modify = str_to_modify.replace('：', ':')    # 前面是中文冒号，后面是英文冒号
-    str_to_modify = str_to_modify.replace('，', ',')    # 前面是中文逗号，后面是英文逗号
+    str_to_modify = str_to_modify.replace("：", ":")  # 前面是中文冒号，后面是英文冒号
+    str_to_modify = str_to_modify.replace("，", ",")  # 前面是中文逗号，后面是英文逗号
     # str_to_modify = 'BOOM'
     return "\\" + cmd + "{" + str_to_modify + "}"
 
+
 def fix_content(final_tex, node_string):
     """
     Fix common GPT errors to increase success rate
@@ -367,9 +546,9 @@ def fix_content(final_tex, node_string):
 
     if "Traceback" in final_tex and "[Local Message]" in final_tex:
         final_tex = node_string  # 出问题了，还原原文
-    if node_string.count('\\begin') != final_tex.count('\\begin'):
+    if node_string.count("\\begin") != final_tex.count("\\begin"):
         final_tex = node_string  # 出问题了，还原原文
-    if node_string.count('\_') > 0 and node_string.count('\_') > final_tex.count('\_'):
+    if node_string.count("\_") > 0 and node_string.count("\_") > final_tex.count("\_"):
         # walk and replace any _ without \
         final_tex = re.sub(r"(?<!\\)_", "\\_", final_tex)
 
@@ -377,24 +556,32 @@ def fix_content(final_tex, node_string):
         # this function count the number of { and }
         brace_level = 0
         for c in string:
-            if c == "{": brace_level += 1
-            elif c == "}": brace_level -= 1
+            if c == "{":
+                brace_level += 1
+            elif c == "}":
+                brace_level -= 1
         return brace_level
+
     def join_most(tex_t, tex_o):
         # this function join translated string and original string when something goes wrong
         p_t = 0
         p_o = 0
+
         def find_next(string, chars, begin):
             p = begin
             while p < len(string):
-                if string[p] in chars: return p, string[p]
+                if string[p] in chars:
+                    return p, string[p]
                 p += 1
             return None, None
+
         while True:
-            res1, char = find_next(tex_o, ['{','}'], p_o)
-            if res1 is None: break
+            res1, char = find_next(tex_o, ["{", "}"], p_o)
+            if res1 is None:
+                break
             res2, char = find_next(tex_t, [char], p_t)
-            if res2 is None: break
+            if res2 is None:
+                break
             p_o = res1 + 1
             p_t = res2 + 1
         return tex_t[:p_t] + tex_o[p_o:]
@@ -404,28 +591,69 @@ def fix_content(final_tex, node_string):
         final_tex = join_most(final_tex, node_string)
     return final_tex
 
+
 def compile_latex_with_timeout(command, cwd, timeout=60):
     import subprocess
-    process = subprocess.Popen(command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, cwd=cwd)
+
+    process = subprocess.Popen(
+        command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, cwd=cwd
+    )
     try:
         stdout, stderr = process.communicate(timeout=timeout)
     except subprocess.TimeoutExpired:
         process.kill()
         stdout, stderr = process.communicate()
-        print("Process timed out!")
+        logger.error("Process timed out (compile_latex_with_timeout)!")
         return False
     return True
 
 
+def run_in_subprocess_wrapper_func(func, args, kwargs, return_dict, exception_dict):
+    import sys
 
-def merge_pdfs(pdf1_path, pdf2_path, output_path):
-    import PyPDF2
-    Percent = 0.8
+    try:
+        result = func(*args, **kwargs)
+        return_dict["result"] = result
+    except Exception as e:
+        exc_info = sys.exc_info()
+        exception_dict["exception"] = exc_info
+
+
+def run_in_subprocess(func):
+    import multiprocessing
+
+    def wrapper(*args, **kwargs):
+        return_dict = multiprocessing.Manager().dict()
+        exception_dict = multiprocessing.Manager().dict()
+        process = multiprocessing.Process(
+            target=run_in_subprocess_wrapper_func,
+            args=(func, args, kwargs, return_dict, exception_dict),
+        )
+        process.start()
+        process.join()
+        process.close()
+        if "exception" in exception_dict:
+            # ooops, the subprocess ran into an exception
+            exc_info = exception_dict["exception"]
+            raise exc_info[1].with_traceback(exc_info[2])
+        if "result" in return_dict.keys():
+            # If the subprocess ran successfully, return the result
+            return return_dict["result"]
+
+    return wrapper
+
+
+def _merge_pdfs(pdf1_path, pdf2_path, output_path):
+    import PyPDF2  # PyPDF2这个库有严重的内存泄露问题，把它放到子进程中运行，从而方便内存的释放
+    from PyPDF2.generic import NameObject, TextStringObject,ArrayObject,FloatObject,NumberObject
+
+    Percent = 1
+    # raise RuntimeError('PyPDF2 has a serious memory leak problem, please use other tools to merge PDF files.')
     # Open the first PDF file
-    with open(pdf1_path, 'rb') as pdf1_file:
+    with open(pdf1_path, "rb") as pdf1_file:
         pdf1_reader = PyPDF2.PdfFileReader(pdf1_file)
         # Open the second PDF file
-        with open(pdf2_path, 'rb') as pdf2_file:
+        with open(pdf2_path, "rb") as pdf2_file:
             pdf2_reader = PyPDF2.PdfFileReader(pdf2_file)
             # Create a new PDF file to store the merged pages
             output_writer = PyPDF2.PdfFileWriter()
@@ -445,12 +673,84 @@ def merge_pdfs(pdf1_path, pdf2_path, output_path):
                     page2 = PyPDF2.PageObject.createBlankPage(pdf1_reader)
                 # Create a new empty page with double width
                 new_page = PyPDF2.PageObject.createBlankPage(
-                    width = int(int(page1.mediaBox.getWidth()) + int(page2.mediaBox.getWidth()) * Percent),
-                    height = max(page1.mediaBox.getHeight(), page2.mediaBox.getHeight())
+                    width=int(
+                        int(page1.mediaBox.getWidth())
+                        + int(page2.mediaBox.getWidth()) * Percent
+                    ),
+                    height=max(page1.mediaBox.getHeight(), page2.mediaBox.getHeight()),
                 )
                 new_page.mergeTranslatedPage(page1, 0, 0)
-                new_page.mergeTranslatedPage(page2, int(int(page1.mediaBox.getWidth())-int(page2.mediaBox.getWidth())* (1-Percent)), 0)
+                new_page.mergeTranslatedPage(
+                    page2,
+                    int(
+                        int(page1.mediaBox.getWidth())
+                        - int(page2.mediaBox.getWidth()) * (1 - Percent)
+                    ),
+                    0,
+                )
+                if '/Annots' in page1:
+                    page1_annot_id = [annot.idnum for annot in page1['/Annots']]
+                else:
+                    page1_annot_id = []
+
+                if '/Annots' in page2:
+                    page2_annot_id = [annot.idnum for annot in page2['/Annots']]
+                else:
+                    page2_annot_id = []
+                if '/Annots' in new_page:
+                    annotations = new_page['/Annots']
+                    for i, annot in enumerate(annotations):
+                        annot_obj = annot.get_object()
+
+                        # 检查注释类型是否是链接（/Link）
+                        if annot_obj.get('/Subtype') == '/Link':
+                            # 检查是否为内部链接跳转（/GoTo）或外部URI链接（/URI）
+                            action = annot_obj.get('/A')
+                            if action:
+
+                                if '/S' in action and action['/S'] == '/GoTo':
+                                    # 内部链接：跳转到文档中的某个页面
+                                    dest = action.get('/D')  # 目标页或目标位置
+                                    if dest and annot.idnum in page2_annot_id:
+                                        # 获取原始文件中跳转信息，包括跳转页面
+                                        destination = pdf2_reader.named_destinations[dest]
+                                        page_number = pdf2_reader.get_destination_page_number(destination)
+                                        #更新跳转信息，跳转到对应的页面和，指定坐标 (100, 150)，缩放比例为 100%
+                                        #“/D”:[10,'/XYZ',100,100,0]
+                                        annot_obj['/A'].update({
+                                            NameObject("/D"): ArrayObject([NumberObject(page_number),destination.dest_array[1], FloatObject(destination.dest_array[2] + int(page1.mediaBox.getWidth())) ,destination.dest_array[3],destination.dest_array[4]])  # 确保键和值是 PdfObject
+                                        })
+                                        rect = annot_obj.get('/Rect')
+                                        # 更新点击坐标
+                                        rect = ArrayObject([FloatObject(rect[0]+ int(page1.mediaBox.getWidth())),rect[1],
+                                                            FloatObject(rect[2]+int(page1.mediaBox.getWidth())),rect[3] ])
+                                        annot_obj.update({
+                                            NameObject("/Rect"): rect  # 确保键和值是 PdfObject
+                                        })
+                                    if dest and annot.idnum in page1_annot_id:
+                                        # 获取原始文件中跳转信息，包括跳转页面
+                                        destination = pdf1_reader.named_destinations[dest]
+                                        page_number = pdf1_reader.get_destination_page_number(destination)
+                                        #更新跳转信息，跳转到对应的页面和，指定坐标 (100, 150)，缩放比例为 100%
+                                        #“/D”:[10,'/XYZ',100,100,0]
+                                        annot_obj['/A'].update({
+                                            NameObject("/D"): ArrayObject([NumberObject(page_number),destination.dest_array[1], FloatObject(destination.dest_array[2]) ,destination.dest_array[3],destination.dest_array[4]])  # 确保键和值是 PdfObject
+                                        })
+                                        rect = annot_obj.get('/Rect')
+                                        rect = ArrayObject([FloatObject(rect[0]),rect[1],
+                                                            FloatObject(rect[2]),rect[3] ])
+                                        annot_obj.update({
+                                            NameObject("/Rect"): rect  # 确保键和值是 PdfObject
+                                        })
+
+                                elif '/S' in action and action['/S'] == '/URI':
+                                    # 外部链接：跳转到某个URI
+                                    uri = action.get('/URI')
+                output_writer.addPage(new_page)
                 output_writer.addPage(new_page)
             # Save the merged PDF file
-            with open(output_path, 'wb') as output_file:
+            with open(output_path, "wb") as output_file:
                 output_writer.write(output_file)
+
+
+merge_pdfs = run_in_subprocess(_merge_pdfs)  # PyPDF2这个库有严重的内存泄露问题，把它放到子进程中运行，从而方便内存的释放

crazy_functions/live_audio/aliyunASR.py

@@ -1,25 +1,124 @@
-import time, threading, json
+import time, json, sys, struct
+import numpy as np
+from loguru import logger as logging
+from scipy.io.wavfile import WAVE_FORMAT
+
+def write_numpy_to_wave(filename, rate, data, add_header=False):
+    """
+    Write a NumPy array as a WAV file.
+    """
+    def _array_tofile(fid, data):
+        # ravel gives a c-contiguous buffer
+        fid.write(data.ravel().view('b').data)
+
+    if hasattr(filename, 'write'):
+        fid = filename
+    else:
+        fid = open(filename, 'wb')
+
+    fs = rate
+
+    try:
+        dkind = data.dtype.kind
+        if not (dkind == 'i' or dkind == 'f' or (dkind == 'u' and
+                                                 data.dtype.itemsize == 1)):
+            raise ValueError("Unsupported data type '%s'" % data.dtype)
+
+        header_data = b''
+
+        header_data += b'RIFF'
+        header_data += b'\x00\x00\x00\x00'
+        header_data += b'WAVE'
+
+        # fmt chunk
+        header_data += b'fmt '
+        if dkind == 'f':
+            format_tag = WAVE_FORMAT.IEEE_FLOAT
+        else:
+            format_tag = WAVE_FORMAT.PCM
+        if data.ndim == 1:
+            channels = 1
+        else:
+            channels = data.shape[1]
+        bit_depth = data.dtype.itemsize * 8
+        bytes_per_second = fs*(bit_depth // 8)*channels
+        block_align = channels * (bit_depth // 8)
+
+        fmt_chunk_data = struct.pack('<HHIIHH', format_tag, channels, fs,
+                                     bytes_per_second, block_align, bit_depth)
+        if not (dkind == 'i' or dkind == 'u'):
+            # add cbSize field for non-PCM files
+            fmt_chunk_data += b'\x00\x00'
+
+        header_data += struct.pack('<I', len(fmt_chunk_data))
+        header_data += fmt_chunk_data
+
+        # fact chunk (non-PCM files)
+        if not (dkind == 'i' or dkind == 'u'):
+            header_data += b'fact'
+            header_data += struct.pack('<II', 4, data.shape[0])
+
+        # check data size (needs to be immediately before the data chunk)
+        if ((len(header_data)-4-4) + (4+4+data.nbytes)) > 0xFFFFFFFF:
+            raise ValueError("Data exceeds wave file size limit")
+        if add_header:
+            fid.write(header_data)
+            # data chunk
+            fid.write(b'data')
+            fid.write(struct.pack('<I', data.nbytes))
+            if data.dtype.byteorder == '>' or (data.dtype.byteorder == '=' and
+                                            sys.byteorder == 'big'):
+                data = data.byteswap()
+        _array_tofile(fid, data)
+
+        if add_header:
+            # Determine file size and place it in correct
+            #  position at start of the file.
+            size = fid.tell()
+            fid.seek(4)
+            fid.write(struct.pack('<I', size-8))
+
+    finally:
+        if not hasattr(filename, 'write'):
+            fid.close()
+        else:
+            fid.seek(0)
+
+def is_speaker_speaking(vad, data, sample_rate):
+    # Function to detect if the speaker is speaking
+    # The WebRTC VAD only accepts 16-bit mono PCM audio,
+    # sampled at 8000, 16000, 32000 or 48000 Hz.
+    # A frame must be either 10, 20, or 30 ms in duration:
+    frame_duration = 30
+    n_bit_each = int(sample_rate * frame_duration / 1000)*2 # x2 because audio is 16 bit (2 bytes)
+    res_list = []
+    for t in range(len(data)):
+        if t!=0 and t % n_bit_each == 0:
+            res_list.append(vad.is_speech(data[t-n_bit_each:t], sample_rate))
+
+    info = ''.join(['^' if r else '.' for r in res_list])
+    info = info[:10]
+    if any(res_list):
+        return True, info
+    else:
+        return False, info
 
 
 class AliyunASR():
 
     def test_on_sentence_begin(self, message, *args):
-        # print("test_on_sentence_begin:{}".format(message))
         pass
 
     def test_on_sentence_end(self, message, *args):
-        # print("test_on_sentence_end:{}".format(message))
         message = json.loads(message)
         self.parsed_sentence = message['payload']['result']
         self.event_on_entence_end.set()
-        print(self.parsed_sentence)
 
     def test_on_start(self, message, *args):
-        # print("test_on_start:{}".format(message))
         pass
 
     def test_on_error(self, message, *args):
-        print("on_error args=>{}".format(args))
+        logging.error("on_error args=>{}".format(args))
         pass
 
     def test_on_close(self, *args):
@@ -27,16 +126,13 @@ class AliyunASR():
         pass
 
     def test_on_result_chg(self, message, *args):
-        # print("test_on_chg:{}".format(message))
         message = json.loads(message)
         self.parsed_text = message['payload']['result']
         self.event_on_result_chg.set()
 
     def test_on_completed(self, message, *args):
-        # print("on_completed:args=>{} message=>{}".format(args, message))
         pass
 
-
     def audio_convertion_thread(self, uuid):
         # 在一个异步线程中采集音频
         import nls  # pip install git+https://github.com/aliyun/alibabacloud-nls-python-sdk.git
@@ -67,12 +163,22 @@ class AliyunASR():
                     on_close=self.test_on_close,
                     callback_args=[uuid.hex]
                 )
-
+        timeout_limit_second = 20
         r = sr.start(aformat="pcm",
+                timeout=timeout_limit_second,
                 enable_intermediate_result=True,
                 enable_punctuation_prediction=True,
                 enable_inverse_text_normalization=True)
 
+        import webrtcvad
+        vad = webrtcvad.Vad()
+        vad.set_mode(1)
+
+        is_previous_frame_transmitted = False   # 上一帧是否有人说话
+        previous_frame_data = None
+        echo_cnt = 0        # 在没有声音之后，继续向服务器发送n次音频数据
+        echo_cnt_max = 4   # 在没有声音之后，继续向服务器发送n次音频数据
+        keep_alive_last_send_time = time.time()
         while not self.stop:
             # time.sleep(self.capture_interval)
             audio = rad.read(uuid.hex)
@@ -80,12 +186,32 @@ class AliyunASR():
                 # convert to pcm file
                 temp_file = f'{temp_folder}/{uuid.hex}.pcm' #
                 dsdata = change_sample_rate(audio, rad.rate, NEW_SAMPLERATE) # 48000 --> 16000
-                io.wavfile.write(temp_file, NEW_SAMPLERATE, dsdata)
+                write_numpy_to_wave(temp_file, NEW_SAMPLERATE, dsdata)
                 # read pcm binary
                 with open(temp_file, "rb") as f: data = f.read()
-                # print('audio len:', len(audio), '\t ds len:', len(dsdata), '\t need n send:', len(data)//640)
+                is_speaking, info = is_speaker_speaking(vad, data, NEW_SAMPLERATE)
+
+                if is_speaking or echo_cnt > 0:
+                    # 如果话筒激活 / 如果处于回声收尾阶段
+                    echo_cnt -= 1
+                    if not is_previous_frame_transmitted:   # 上一帧没有人声，但是我们把上一帧同样加上
+                        if previous_frame_data is not None: data = previous_frame_data + data
+                    if is_speaking:
+                        echo_cnt = echo_cnt_max
                     slices = zip(*(iter(data),) * 640)      # 640个字节为一组
                     for i in slices: sr.send_audio(bytes(i))
+                    keep_alive_last_send_time = time.time()
+                    is_previous_frame_transmitted = True
+                else:
+                    is_previous_frame_transmitted = False
+                    echo_cnt = 0
+                    # 保持链接激活，即使没有声音，也根据时间间隔，发送一些音频片段给服务器
+                    if time.time() - keep_alive_last_send_time > timeout_limit_second/2:
+                        slices = zip(*(iter(data),) * 640)    # 640个字节为一组
+                        for i in slices: sr.send_audio(bytes(i))
+                        keep_alive_last_send_time = time.time()
+                        is_previous_frame_transmitted = True
+                self.audio_shape = info
             else:
                 time.sleep(0.1)
 
@@ -117,14 +243,14 @@ class AliyunASR():
 
         try:
             response = client.do_action_with_exception(request)
-            print(response)
+            logging.info(response)
             jss = json.loads(response)
             if 'Token' in jss and 'Id' in jss['Token']:
                 token = jss['Token']['Id']
                 expireTime = jss['Token']['ExpireTime']
-                print("token = " + token)
-                print("expireTime = " + str(expireTime))
+                logging.info("token = " + token)
+                logging.info("expireTime = " + str(expireTime))
         except Exception as e:
-            print(e)
+            logging.error(e)
 
         return token

crazy_functions/live_audio/audio_io.py

@@ -35,7 +35,7 @@ class RealtimeAudioDistribution():
     def read(self, uuid):
         if uuid in self.data:
             res = self.data.pop(uuid)
-            print('\r read-', len(res), '-', max(res), end='', flush=True)
+            # print('\r read-', len(res), '-', max(res), end='', flush=True)
         else:
             res = None
         return res

crazy_functions/multi_stage/multi_stage_utils.py

@@ -0,0 +1,93 @@
+from pydantic import BaseModel, Field
+from typing import List
+from toolbox import update_ui_lastest_msg, disable_auto_promotion
+from toolbox import CatchException, update_ui, get_conf, select_api_key, get_log_folder
+from request_llms.bridge_all import predict_no_ui_long_connection
+from crazy_functions.json_fns.pydantic_io import GptJsonIO, JsonStringError
+import time
+import pickle
+
+def have_any_recent_upload_files(chatbot):
+    _5min = 5 * 60
+    if not chatbot: return False    # chatbot is None
+    most_recent_uploaded = chatbot._cookies.get("most_recent_uploaded", None)
+    if not most_recent_uploaded: return False   # most_recent_uploaded is None
+    if time.time() - most_recent_uploaded["time"] < _5min: return True # most_recent_uploaded is new
+    else: return False  # most_recent_uploaded is too old
+
+class GptAcademicState():
+    def __init__(self):
+        self.reset()
+
+    def reset(self):
+        pass
+
+    def dump_state(self, chatbot):
+        chatbot._cookies['plugin_state'] = pickle.dumps(self)
+
+    def set_state(self, chatbot, key, value):
+        setattr(self, key, value)
+        chatbot._cookies['plugin_state'] = pickle.dumps(self)
+
+    def get_state(chatbot, cls=None):
+        state = chatbot._cookies.get('plugin_state', None)
+        if state is not None:   state = pickle.loads(state)
+        elif cls is not None:   state = cls()
+        else:                   state = GptAcademicState()
+        state.chatbot = chatbot
+        return state
+
+
+class GptAcademicGameBaseState():
+    """
+    1. first init: __init__ ->
+    """
+    def init_game(self, chatbot, lock_plugin):
+        self.plugin_name = None
+        self.callback_fn = None
+        self.delete_game = False
+        self.step_cnt = 0
+
+    def lock_plugin(self, chatbot):
+        if self.callback_fn is None:
+            raise ValueError("callback_fn is None")
+        chatbot._cookies['lock_plugin'] = self.callback_fn
+        self.dump_state(chatbot)
+
+    def get_plugin_name(self):
+        if self.plugin_name is None:
+            raise ValueError("plugin_name is None")
+        return self.plugin_name
+
+    def dump_state(self, chatbot):
+        chatbot._cookies[f'plugin_state/{self.get_plugin_name()}'] = pickle.dumps(self)
+
+    def set_state(self, chatbot, key, value):
+        setattr(self, key, value)
+        chatbot._cookies[f'plugin_state/{self.get_plugin_name()}'] = pickle.dumps(self)
+
+    @staticmethod
+    def sync_state(chatbot, llm_kwargs, cls, plugin_name, callback_fn, lock_plugin=True):
+        state = chatbot._cookies.get(f'plugin_state/{plugin_name}', None)
+        if state is not None:
+            state = pickle.loads(state)
+        else:
+            state = cls()
+            state.init_game(chatbot, lock_plugin)
+        state.plugin_name = plugin_name
+        state.llm_kwargs = llm_kwargs
+        state.chatbot = chatbot
+        state.callback_fn = callback_fn
+        return state
+
+    def continue_game(self, prompt, chatbot, history):
+        # 游戏主体
+        yield from self.step(prompt, chatbot, history)
+        self.step_cnt += 1
+        # 保存状态，收尾
+        self.dump_state(chatbot)
+        # 如果游戏结束，清理
+        if self.delete_game:
+            chatbot._cookies['lock_plugin'] = None
+            chatbot._cookies[f'plugin_state/{self.get_plugin_name()}'] = None
+        yield from update_ui(chatbot=chatbot, history=history)

crazy_functions/pdf_fns/breakdown_txt.py

@@ -0,0 +1,126 @@
+from crazy_functions.ipc_fns.mp import run_in_subprocess_with_timeout
+from loguru import logger
+
+def force_breakdown(txt, limit, get_token_fn):
+    """ 当无法用标点、空行分割时，我们用最暴力的方法切割
+    """
+    for i in reversed(range(len(txt))):
+        if get_token_fn(txt[:i]) < limit:
+            return txt[:i], txt[i:]
+    return "Tiktoken未知错误", "Tiktoken未知错误"
+
+
+def maintain_storage(remain_txt_to_cut, remain_txt_to_cut_storage):
+    """ 为了加速计算，我们采样一个特殊的手段。当 remain_txt_to_cut > `_max` 时， 我们把 _max 后的文字转存至 remain_txt_to_cut_storage
+    当 remain_txt_to_cut < `_min` 时，我们再把 remain_txt_to_cut_storage 中的部分文字取出
+    """
+    _min = int(5e4)
+    _max = int(1e5)
+    # print(len(remain_txt_to_cut), len(remain_txt_to_cut_storage))
+    if len(remain_txt_to_cut) < _min and len(remain_txt_to_cut_storage) > 0:
+        remain_txt_to_cut = remain_txt_to_cut + remain_txt_to_cut_storage
+        remain_txt_to_cut_storage = ""
+    if len(remain_txt_to_cut) > _max:
+        remain_txt_to_cut_storage = remain_txt_to_cut[_max:] + remain_txt_to_cut_storage
+        remain_txt_to_cut = remain_txt_to_cut[:_max]
+    return remain_txt_to_cut, remain_txt_to_cut_storage
+
+
+def cut(limit, get_token_fn, txt_tocut, must_break_at_empty_line, break_anyway=False):
+    """ 文本切分
+    """
+    res = []
+    total_len = len(txt_tocut)
+    fin_len = 0
+    remain_txt_to_cut = txt_tocut
+    remain_txt_to_cut_storage = ""
+    # 为了加速计算，我们采样一个特殊的手段。当 remain_txt_to_cut > `_max` 时， 我们把 _max 后的文字转存至 remain_txt_to_cut_storage
+    remain_txt_to_cut, remain_txt_to_cut_storage = maintain_storage(remain_txt_to_cut, remain_txt_to_cut_storage)
+
+    while True:
+        if get_token_fn(remain_txt_to_cut) <= limit:
+            # 如果剩余文本的token数小于限制，那么就不用切了
+            res.append(remain_txt_to_cut); fin_len+=len(remain_txt_to_cut)
+            break
+        else:
+            # 如果剩余文本的token数大于限制，那么就切
+            lines = remain_txt_to_cut.split('\n')
+
+            # 估计一个切分点
+            estimated_line_cut = limit / get_token_fn(remain_txt_to_cut) * len(lines)
+            estimated_line_cut = int(estimated_line_cut)
+
+            # 开始查找合适切分点的偏移（cnt）
+            cnt = 0
+            for cnt in reversed(range(estimated_line_cut)):
+                if must_break_at_empty_line:
+                    # 首先尝试用双空行（\n\n）作为切分点
+                    if lines[cnt] != "":
+                        continue
+                prev = "\n".join(lines[:cnt])
+                post = "\n".join(lines[cnt:])
+                if get_token_fn(prev) < limit:
+                    break
+
+            if cnt == 0:
+                # 如果没有找到合适的切分点
+                if break_anyway:
+                    # 是否允许暴力切分
+                    prev, post = force_breakdown(remain_txt_to_cut, limit, get_token_fn)
+                else:
+                    # 不允许直接报错
+                    raise RuntimeError(f"存在一行极长的文本！{remain_txt_to_cut}")
+
+            # 追加列表
+            res.append(prev); fin_len+=len(prev)
+            # 准备下一次迭代
+            remain_txt_to_cut = post
+            remain_txt_to_cut, remain_txt_to_cut_storage = maintain_storage(remain_txt_to_cut, remain_txt_to_cut_storage)
+            process = fin_len/total_len
+            logger.info(f'正在文本切分 {int(process*100)}%')
+            if len(remain_txt_to_cut.strip()) == 0:
+                break
+    return res
+
+
+def breakdown_text_to_satisfy_token_limit_(txt, limit, llm_model="gpt-3.5-turbo"):
+    """ 使用多种方式尝试切分文本，以满足 token 限制
+    """
+    from request_llms.bridge_all import model_info
+    enc = model_info[llm_model]['tokenizer']
+    def get_token_fn(txt): return len(enc.encode(txt, disallowed_special=()))
+    try:
+        # 第1次尝试，将双空行（\n\n）作为切分点
+        return cut(limit, get_token_fn, txt, must_break_at_empty_line=True)
+    except RuntimeError:
+        try:
+            # 第2次尝试，将单空行（\n）作为切分点
+            return cut(limit, get_token_fn, txt, must_break_at_empty_line=False)
+        except RuntimeError:
+            try:
+                # 第3次尝试，将英文句号（.）作为切分点
+                res = cut(limit, get_token_fn, txt.replace('.', '。\n'), must_break_at_empty_line=False) # 这个中文的句号是故意的，作为一个标识而存在
+                return [r.replace('。\n', '.') for r in res]
+            except RuntimeError as e:
+                try:
+                    # 第4次尝试，将中文句号（。）作为切分点
+                    res = cut(limit, get_token_fn, txt.replace('。', '。。\n'), must_break_at_empty_line=False)
+                    return [r.replace('。。\n', '。') for r in res]
+                except RuntimeError as e:
+                    # 第5次尝试，没办法了，随便切一下吧
+                    return cut(limit, get_token_fn, txt, must_break_at_empty_line=False, break_anyway=True)
+
+breakdown_text_to_satisfy_token_limit = run_in_subprocess_with_timeout(breakdown_text_to_satisfy_token_limit_, timeout=60)
+
+if __name__ == '__main__':
+    from crazy_functions.crazy_utils import read_and_clean_pdf_text
+    file_content, page_one = read_and_clean_pdf_text("build/assets/at.pdf")
+
+    from request_llms.bridge_all import model_info
+    for i in range(5):
+        file_content += file_content
+
+    logger.info(len(file_content))
+    TOKEN_LIMIT_PER_FRAGMENT = 2500
+    res = breakdown_text_to_satisfy_token_limit(file_content, TOKEN_LIMIT_PER_FRAGMENT)
+

crazy_functions/pdf_fns/parse_pdf.py

@@ -0,0 +1,171 @@
+from functools import lru_cache
+from toolbox import gen_time_str
+from toolbox import promote_file_to_downloadzone
+from toolbox import write_history_to_file, promote_file_to_downloadzone
+from toolbox import get_conf
+from toolbox import ProxyNetworkActivate
+from shared_utils.colorful import *
+import requests
+import random
+import copy
+import os
+import math
+
+class GROBID_OFFLINE_EXCEPTION(Exception): pass
+
+def get_avail_grobid_url():
+    GROBID_URLS = get_conf('GROBID_URLS')
+    if len(GROBID_URLS) == 0: return None
+    try:
+        _grobid_url = random.choice(GROBID_URLS) # 随机负载均衡
+        if _grobid_url.endswith('/'): _grobid_url = _grobid_url.rstrip('/')
+        with ProxyNetworkActivate('Connect_Grobid'):
+            res = requests.get(_grobid_url+'/api/isalive')
+        if res.text=='true': return _grobid_url
+        else: return None
+    except:
+        return None
+
+@lru_cache(maxsize=32)
+def parse_pdf(pdf_path, grobid_url):
+    import scipdf   # pip install scipdf_parser
+    if grobid_url.endswith('/'): grobid_url = grobid_url.rstrip('/')
+    try:
+        with ProxyNetworkActivate('Connect_Grobid'):
+            article_dict = scipdf.parse_pdf_to_dict(pdf_path, grobid_url=grobid_url)
+    except GROBID_OFFLINE_EXCEPTION:
+        raise GROBID_OFFLINE_EXCEPTION("GROBID服务不可用，请修改config中的GROBID_URL，可修改成本地GROBID服务。")
+    except:
+        raise RuntimeError("解析PDF失败，请检查PDF是否损坏。")
+    return article_dict
+
+
+def produce_report_markdown(gpt_response_collection, meta, paper_meta_info, chatbot, fp, generated_conclusion_files):
+    # -=-=-=-=-=-=-=-= 写出第1个文件：翻译前后混合 -=-=-=-=-=-=-=-=
+    res_path = write_history_to_file(meta +  ["# Meta Translation" , paper_meta_info] + gpt_response_collection, file_basename=f"{gen_time_str()}translated_and_original.md", file_fullname=None)
+    promote_file_to_downloadzone(res_path, rename_file=os.path.basename(res_path)+'.md', chatbot=chatbot)
+    generated_conclusion_files.append(res_path)
+
+    # -=-=-=-=-=-=-=-= 写出第2个文件：仅翻译后的文本 -=-=-=-=-=-=-=-=
+    translated_res_array = []
+    # 记录当前的大章节标题：
+    last_section_name = ""
+    for index, value in enumerate(gpt_response_collection):
+        # 先挑选偶数序列号：
+        if index % 2 != 0:
+            # 先提取当前英文标题：
+            cur_section_name = gpt_response_collection[index-1].split('\n')[0].split(" Part")[0]
+            # 如果index是1的话，则直接使用first section name：
+            if cur_section_name != last_section_name:
+                cur_value = cur_section_name + '\n'
+                last_section_name = copy.deepcopy(cur_section_name)
+            else:
+                cur_value = ""
+            # 再做一个小修改：重新修改当前part的标题，默认用英文的
+            cur_value += value
+            translated_res_array.append(cur_value)
+    res_path = write_history_to_file(meta +  ["# Meta Translation" , paper_meta_info] + translated_res_array,
+                                     file_basename = f"{gen_time_str()}-translated_only.md",
+                                     file_fullname = None,
+                                     auto_caption = False)
+    promote_file_to_downloadzone(res_path, rename_file=os.path.basename(res_path)+'.md', chatbot=chatbot)
+    generated_conclusion_files.append(res_path)
+    return res_path
+
+def translate_pdf(article_dict, llm_kwargs, chatbot, fp, generated_conclusion_files, TOKEN_LIMIT_PER_FRAGMENT, DST_LANG, plugin_kwargs={}):
+    from crazy_functions.pdf_fns.report_gen_html import construct_html
+    from crazy_functions.pdf_fns.breakdown_txt import breakdown_text_to_satisfy_token_limit
+    from crazy_functions.crazy_utils import request_gpt_model_in_new_thread_with_ui_alive
+    from crazy_functions.crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
+
+    prompt = "以下是一篇学术论文的基本信息:\n"
+    # title
+    title = article_dict.get('title', '无法获取 title'); prompt += f'title:{title}\n\n'
+    # authors
+    authors = article_dict.get('authors', '无法获取 authors')[:100]; prompt += f'authors:{authors}\n\n'
+    # abstract
+    abstract = article_dict.get('abstract', '无法获取 abstract'); prompt += f'abstract:{abstract}\n\n'
+    # command
+    prompt += f"请将题目和摘要翻译为{DST_LANG}。"
+    meta = [f'# Title:\n\n', title, f'# Abstract:\n\n', abstract ]
+
+    # 单线，获取文章meta信息
+    paper_meta_info = yield from request_gpt_model_in_new_thread_with_ui_alive(
+        inputs=prompt,
+        inputs_show_user=prompt,
+        llm_kwargs=llm_kwargs,
+        chatbot=chatbot, history=[],
+        sys_prompt="You are an academic paper reader。",
+    )
+
+    # 多线，翻译
+    inputs_array = []
+    inputs_show_user_array = []
+
+    # get_token_num
+    from request_llms.bridge_all import model_info
+    enc = model_info[llm_kwargs['llm_model']]['tokenizer']
+    def get_token_num(txt): return len(enc.encode(txt, disallowed_special=()))
+
+    def break_down(txt):
+        raw_token_num = get_token_num(txt)
+        if raw_token_num <= TOKEN_LIMIT_PER_FRAGMENT:
+            return [txt]
+        else:
+            # raw_token_num > TOKEN_LIMIT_PER_FRAGMENT
+            # find a smooth token limit to achieve even seperation
+            count = int(math.ceil(raw_token_num / TOKEN_LIMIT_PER_FRAGMENT))
+            token_limit_smooth = raw_token_num // count + count
+            return breakdown_text_to_satisfy_token_limit(txt, limit=token_limit_smooth, llm_model=llm_kwargs['llm_model'])
+
+    for section in article_dict.get('sections'):
+        if len(section['text']) == 0: continue
+        section_frags = break_down(section['text'])
+        for i, fragment in enumerate(section_frags):
+            heading = section['heading']
+            if len(section_frags) > 1: heading += f' Part-{i+1}'
+            inputs_array.append(
+                f"你需要翻译{heading}章节，内容如下: \n\n{fragment}"
+            )
+            inputs_show_user_array.append(
+                f"# {heading}\n\n{fragment}"
+            )
+
+    gpt_response_collection = yield from request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
+        inputs_array=inputs_array,
+        inputs_show_user_array=inputs_show_user_array,
+        llm_kwargs=llm_kwargs,
+        chatbot=chatbot,
+        history_array=[meta for _ in inputs_array],
+        sys_prompt_array=[
+            "请你作为一个学术翻译，负责把学术论文准确翻译成中文。注意文章中的每一句话都要翻译。" + plugin_kwargs.get("additional_prompt", "") for _ in inputs_array],
+    )
+    # -=-=-=-=-=-=-=-= 写出Markdown文件 -=-=-=-=-=-=-=-=
+    produce_report_markdown(gpt_response_collection, meta, paper_meta_info, chatbot, fp, generated_conclusion_files)
+
+    # -=-=-=-=-=-=-=-= 写出HTML文件 -=-=-=-=-=-=-=-=
+    ch = construct_html()
+    orig = ""
+    trans = ""
+    gpt_response_collection_html = copy.deepcopy(gpt_response_collection)
+    for i,k in enumerate(gpt_response_collection_html):
+        if i%2==0:
+            gpt_response_collection_html[i] = inputs_show_user_array[i//2]
+        else:
+            # 先提取当前英文标题：
+            cur_section_name = gpt_response_collection[i-1].split('\n')[0].split(" Part")[0]
+            cur_value = cur_section_name + "\n" + gpt_response_collection_html[i]
+            gpt_response_collection_html[i] = cur_value
+
+    final = ["", "", "一、论文概况",  "", "Abstract", paper_meta_info,  "二、论文翻译",  ""]
+    final.extend(gpt_response_collection_html)
+    for i, k in enumerate(final):
+        if i%2==0:
+            orig = k
+        if i%2==1:
+            trans = k
+            ch.add_row(a=orig, b=trans)
+    create_report_file_name = f"{os.path.basename(fp)}.trans.html"
+    html_file = ch.save_file(create_report_file_name)
+    generated_conclusion_files.append(html_file)
+    promote_file_to_downloadzone(html_file, rename_file=os.path.basename(html_file), chatbot=chatbot)

crazy_functions/pdf_fns/parse_pdf_grobid.py

@@ -0,0 +1,26 @@
+import os
+from toolbox import CatchException, report_exception, get_log_folder, gen_time_str, check_packages
+from toolbox import update_ui, promote_file_to_downloadzone, update_ui_lastest_msg, disable_auto_promotion
+from toolbox import write_history_to_file, promote_file_to_downloadzone, get_conf, extract_archive
+from crazy_functions.pdf_fns.parse_pdf import parse_pdf, translate_pdf
+
+def 解析PDF_基于GROBID(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, grobid_url):
+    import copy, json
+    TOKEN_LIMIT_PER_FRAGMENT = 1024
+    generated_conclusion_files = []
+    generated_html_files = []
+    DST_LANG = "中文"
+    from crazy_functions.pdf_fns.report_gen_html import construct_html
+    for index, fp in enumerate(file_manifest):
+        chatbot.append(["当前进度：", f"正在连接GROBID服务，请稍候: {grobid_url}\n如果等待时间过长，请修改config中的GROBID_URL，可修改成本地GROBID服务。"]); yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        article_dict = parse_pdf(fp, grobid_url)
+        grobid_json_res = os.path.join(get_log_folder(), gen_time_str() + "grobid.json")
+        with open(grobid_json_res, 'w+', encoding='utf8') as f:
+            f.write(json.dumps(article_dict, indent=4, ensure_ascii=False))
+        promote_file_to_downloadzone(grobid_json_res, chatbot=chatbot)
+        if article_dict is None: raise RuntimeError("解析PDF失败，请检查PDF是否损坏。")
+        yield from translate_pdf(article_dict, llm_kwargs, chatbot, fp, generated_conclusion_files, TOKEN_LIMIT_PER_FRAGMENT, DST_LANG, plugin_kwargs=plugin_kwargs)
+    chatbot.append(("给出输出文件清单", str(generated_conclusion_files + generated_html_files)))
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+
+

crazy_functions/pdf_fns/parse_pdf_legacy.py

@@ -0,0 +1,111 @@
+from toolbox import get_log_folder
+from toolbox import update_ui, promote_file_to_downloadzone
+from toolbox import write_history_to_file, promote_file_to_downloadzone
+from crazy_functions.crazy_utils import request_gpt_model_in_new_thread_with_ui_alive
+from crazy_functions.crazy_utils import request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency
+from crazy_functions.crazy_utils import read_and_clean_pdf_text
+from shared_utils.colorful import *
+from loguru import logger
+import os
+
+def 解析PDF_简单拆解(file_manifest, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt):
+    """
+    注意：此函数已经弃用！！新函数位于：crazy_functions/pdf_fns/parse_pdf.py
+    """
+    import copy
+    TOKEN_LIMIT_PER_FRAGMENT = 1024
+    generated_conclusion_files = []
+    generated_html_files = []
+    from crazy_functions.pdf_fns.report_gen_html import construct_html
+    for index, fp in enumerate(file_manifest):
+        # 读取PDF文件
+        file_content, page_one = read_and_clean_pdf_text(fp)
+        file_content = file_content.encode('utf-8', 'ignore').decode()   # avoid reading non-utf8 chars
+        page_one = str(page_one).encode('utf-8', 'ignore').decode()      # avoid reading non-utf8 chars
+
+        # 递归地切割PDF文件
+        from crazy_functions.pdf_fns.breakdown_txt import breakdown_text_to_satisfy_token_limit
+        paper_fragments = breakdown_text_to_satisfy_token_limit(txt=file_content, limit=TOKEN_LIMIT_PER_FRAGMENT, llm_model=llm_kwargs['llm_model'])
+        page_one_fragments = breakdown_text_to_satisfy_token_limit(txt=page_one, limit=TOKEN_LIMIT_PER_FRAGMENT//4, llm_model=llm_kwargs['llm_model'])
+
+        # 为了更好的效果，我们剥离Introduction之后的部分（如果有）
+        paper_meta = page_one_fragments[0].split('introduction')[0].split('Introduction')[0].split('INTRODUCTION')[0]
+
+        # 单线，获取文章meta信息
+        paper_meta_info = yield from request_gpt_model_in_new_thread_with_ui_alive(
+            inputs=f"以下是一篇学术论文的基础信息，请从中提取出“标题”、“收录会议或期刊”、“作者”、“摘要”、“编号”、“作者邮箱”这六个部分。请用markdown格式输出，最后用中文翻译摘要部分。请提取：{paper_meta}",
+            inputs_show_user=f"请从{fp}中提取出“标题”、“收录会议或期刊”等基本信息。",
+            llm_kwargs=llm_kwargs,
+            chatbot=chatbot, history=[],
+            sys_prompt="Your job is to collect information from materials。",
+        )
+
+        # 多线，翻译
+        gpt_response_collection = yield from request_gpt_model_multi_threads_with_very_awesome_ui_and_high_efficiency(
+            inputs_array=[
+                f"你需要翻译以下内容：\n{frag}" for frag in paper_fragments],
+            inputs_show_user_array=[f"\n---\n 原文： \n\n {frag.replace('#', '')}  \n---\n 翻译：\n " for frag in paper_fragments],
+            llm_kwargs=llm_kwargs,
+            chatbot=chatbot,
+            history_array=[[paper_meta] for _ in paper_fragments],
+            sys_prompt_array=[
+                "请你作为一个学术翻译，负责把学术论文准确翻译成中文。注意文章中的每一句话都要翻译。" + plugin_kwargs.get("additional_prompt", "")
+                for _ in paper_fragments],
+            # max_workers=5  # OpenAI所允许的最大并行过载
+        )
+        gpt_response_collection_md = copy.deepcopy(gpt_response_collection)
+        # 整理报告的格式
+        for i,k in enumerate(gpt_response_collection_md):
+            if i%2==0:
+                gpt_response_collection_md[i] = f"\n\n---\n\n ## 原文[{i//2}/{len(gpt_response_collection_md)//2}]： \n\n {paper_fragments[i//2].replace('#', '')}  \n\n---\n\n ## 翻译[{i//2}/{len(gpt_response_collection_md)//2}]：\n "
+            else:
+                gpt_response_collection_md[i] = gpt_response_collection_md[i]
+        final = ["一、论文概况\n\n---\n\n", paper_meta_info.replace('# ', '### ') + '\n\n---\n\n', "二、论文翻译", ""]
+        final.extend(gpt_response_collection_md)
+        create_report_file_name = f"{os.path.basename(fp)}.trans.md"
+        res = write_history_to_file(final, create_report_file_name)
+        promote_file_to_downloadzone(res, chatbot=chatbot)
+
+        # 更新UI
+        generated_conclusion_files.append(f'{get_log_folder()}/{create_report_file_name}')
+        chatbot.append((f"{fp}完成了吗？", res))
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+
+        # write html
+        try:
+            ch = construct_html()
+            orig = ""
+            trans = ""
+            gpt_response_collection_html = copy.deepcopy(gpt_response_collection)
+            for i,k in enumerate(gpt_response_collection_html):
+                if i%2==0:
+                    gpt_response_collection_html[i] = paper_fragments[i//2].replace('#', '')
+                else:
+                    gpt_response_collection_html[i] = gpt_response_collection_html[i]
+            final = ["论文概况", paper_meta_info.replace('# ', '### '),  "二、论文翻译",  ""]
+            final.extend(gpt_response_collection_html)
+            for i, k in enumerate(final):
+                if i%2==0:
+                    orig = k
+                if i%2==1:
+                    trans = k
+                    ch.add_row(a=orig, b=trans)
+            create_report_file_name = f"{os.path.basename(fp)}.trans.html"
+            generated_html_files.append(ch.save_file(create_report_file_name))
+        except:
+            from toolbox import trimmed_format_exc
+            logger.error('writing html result failed:', trimmed_format_exc())
+
+    # 准备文件的下载
+    for pdf_path in generated_conclusion_files:
+        # 重命名文件
+        rename_file = f'翻译-{os.path.basename(pdf_path)}'
+        promote_file_to_downloadzone(pdf_path, rename_file=rename_file, chatbot=chatbot)
+    for html_path in generated_html_files:
+        # 重命名文件
+        rename_file = f'翻译-{os.path.basename(html_path)}'
+        promote_file_to_downloadzone(html_path, rename_file=rename_file, chatbot=chatbot)
+    chatbot.append(("给出输出文件清单", str(generated_conclusion_files + generated_html_files)))
+    yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+
+

crazy_functions/pdf_fns/parse_pdf_via_doc2x.py

@@ -0,0 +1,213 @@
+from toolbox import get_log_folder, gen_time_str, get_conf
+from toolbox import update_ui, promote_file_to_downloadzone
+from toolbox import promote_file_to_downloadzone, extract_archive
+from toolbox import generate_file_link, zip_folder
+from crazy_functions.crazy_utils import get_files_from_everything
+from shared_utils.colorful import *
+import os
+
+def refresh_key(doc2x_api_key):
+    import requests, json
+    url = "https://api.doc2x.noedgeai.com/api/token/refresh"
+    res = requests.post(
+        url,
+        headers={"Authorization": "Bearer " + doc2x_api_key}
+    )
+    res_json = []
+    if res.status_code == 200:
+        decoded = res.content.decode("utf-8")
+        res_json = json.loads(decoded)
+        doc2x_api_key = res_json['data']['token']
+    else:
+        raise RuntimeError(format("[ERROR] status code: %d, body: %s" % (res.status_code, res.text)))
+    return doc2x_api_key
+
+def 解析PDF_DOC2X_转Latex(pdf_file_path):
+    import requests, json, os
+    DOC2X_API_KEY = get_conf('DOC2X_API_KEY')
+    latex_dir = get_log_folder(plugin_name="pdf_ocr_latex")
+    doc2x_api_key = DOC2X_API_KEY
+    if doc2x_api_key.startswith('sk-'):
+        url = "https://api.doc2x.noedgeai.com/api/v1/pdf"
+    else:
+        doc2x_api_key = refresh_key(doc2x_api_key)
+        url = "https://api.doc2x.noedgeai.com/api/platform/pdf"
+
+    res = requests.post(
+        url,
+        files={"file": open(pdf_file_path, "rb")},
+        data={"ocr": "1"},
+        headers={"Authorization": "Bearer " + doc2x_api_key}
+    )
+    res_json = []
+    if res.status_code == 200:
+        decoded = res.content.decode("utf-8")
+        for z_decoded in decoded.split('\n'):
+            if len(z_decoded) == 0: continue
+            assert z_decoded.startswith("data: ")
+            z_decoded = z_decoded[len("data: "):]
+            decoded_json = json.loads(z_decoded)
+            res_json.append(decoded_json)
+    else:
+        raise RuntimeError(format("[ERROR] status code: %d, body: %s" % (res.status_code, res.text)))
+
+    uuid = res_json[0]['uuid']
+    to = "latex" # latex, md, docx
+    url = "https://api.doc2x.noedgeai.com/api/export"+"?request_id="+uuid+"&to="+to
+
+    res = requests.get(url, headers={"Authorization": "Bearer " + doc2x_api_key})
+    latex_zip_path = os.path.join(latex_dir, gen_time_str() + '.zip')
+    latex_unzip_path = os.path.join(latex_dir, gen_time_str())
+    if res.status_code == 200:
+        with open(latex_zip_path, "wb") as f: f.write(res.content)
+    else:
+        raise RuntimeError(format("[ERROR] status code: %d, body: %s" % (res.status_code, res.text)))
+
+    import zipfile
+    with zipfile.ZipFile(latex_zip_path, 'r') as zip_ref:
+        zip_ref.extractall(latex_unzip_path)
+
+
+    return latex_unzip_path
+
+
+
+
+def 解析PDF_DOC2X_单文件(fp, project_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, DOC2X_API_KEY, user_request):
+
+
+    def pdf2markdown(filepath):
+        import requests, json, os
+        markdown_dir = get_log_folder(plugin_name="pdf_ocr")
+        doc2x_api_key = DOC2X_API_KEY
+        if doc2x_api_key.startswith('sk-'):
+            url = "https://api.doc2x.noedgeai.com/api/v1/pdf"
+        else:
+            doc2x_api_key = refresh_key(doc2x_api_key)
+            url = "https://api.doc2x.noedgeai.com/api/platform/pdf"
+
+        chatbot.append((None, "加载PDF文件，发送至DOC2X解析..."))
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+
+        res = requests.post(
+            url,
+            files={"file": open(filepath, "rb")},
+            data={"ocr": "1"},
+            headers={"Authorization": "Bearer " + doc2x_api_key}
+        )
+        res_json = []
+        if res.status_code == 200:
+            decoded = res.content.decode("utf-8")
+            for z_decoded in decoded.split('\n'):
+                if len(z_decoded) == 0: continue
+                assert z_decoded.startswith("data: ")
+                z_decoded = z_decoded[len("data: "):]
+                decoded_json = json.loads(z_decoded)
+                res_json.append(decoded_json)
+            if 'limit exceeded' in decoded_json.get('status', ''):
+                raise RuntimeError("Doc2x API 页数受限，请联系 Doc2x 方面，并更换新的 API 秘钥。")
+        else:
+            raise RuntimeError(format("[ERROR] status code: %d, body: %s" % (res.status_code, res.text)))
+        uuid = res_json[0]['uuid']
+        to = "md" # latex, md, docx
+        url = "https://api.doc2x.noedgeai.com/api/export"+"?request_id="+uuid+"&to="+to
+
+        chatbot.append((None, f"读取解析: {url} ..."))
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+
+        res = requests.get(url, headers={"Authorization": "Bearer " + doc2x_api_key})
+        md_zip_path = os.path.join(markdown_dir, gen_time_str() + '.zip')
+        if res.status_code == 200:
+            with open(md_zip_path, "wb") as f: f.write(res.content)
+        else:
+            raise RuntimeError(format("[ERROR] status code: %d, body: %s" % (res.status_code, res.text)))
+        promote_file_to_downloadzone(md_zip_path, chatbot=chatbot)
+        chatbot.append((None, f"完成解析 {md_zip_path} ..."))
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        return md_zip_path
+
+    def deliver_to_markdown_plugin(md_zip_path, user_request):
+        from crazy_functions.Markdown_Translate import Markdown英译中
+        import shutil, re
+
+        time_tag = gen_time_str()
+        target_path_base = get_log_folder(chatbot.get_user())
+        file_origin_name = os.path.basename(md_zip_path)
+        this_file_path = os.path.join(target_path_base, file_origin_name)
+        os.makedirs(target_path_base, exist_ok=True)
+        shutil.copyfile(md_zip_path, this_file_path)
+        ex_folder = this_file_path + ".extract"
+        extract_archive(
+            file_path=this_file_path, dest_dir=ex_folder
+        )
+
+        # edit markdown files
+        success, file_manifest, project_folder = get_files_from_everything(ex_folder, type='.md')
+        for generated_fp in file_manifest:
+            # 修正一些公式问题
+            with open(generated_fp, 'r', encoding='utf8') as f:
+                content = f.read()
+            # 将公式中的\[ \]替换成$$
+            content = content.replace(r'\[', r'$$').replace(r'\]', r'$$')
+            # 将公式中的\( \)替换成$
+            content = content.replace(r'\(', r'$').replace(r'\)', r'$')
+            content = content.replace('```markdown', '\n').replace('```', '\n')
+            with open(generated_fp, 'w', encoding='utf8') as f:
+                f.write(content)
+            promote_file_to_downloadzone(generated_fp, chatbot=chatbot)
+            yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+
+            # 生成在线预览html
+            file_name = '在线预览翻译（原文）' + gen_time_str() + '.html'
+            preview_fp = os.path.join(ex_folder, file_name)
+            from shared_utils.advanced_markdown_format import markdown_convertion_for_file
+            with open(generated_fp, "r", encoding="utf-8") as f:
+                md = f.read()
+            #     # Markdown中使用不标准的表格，需要在表格前加上一个emoji，以便公式渲染
+            #     md = re.sub(r'^<table>', r'.<table>', md, flags=re.MULTILINE)
+            html = markdown_convertion_for_file(md)
+            with open(preview_fp, "w", encoding="utf-8") as f: f.write(html)
+            chatbot.append([None, f"生成在线预览：{generate_file_link([preview_fp])}"])
+            promote_file_to_downloadzone(preview_fp, chatbot=chatbot)
+
+
+
+        chatbot.append((None, f"调用Markdown插件 {ex_folder} ..."))
+        plugin_kwargs['markdown_expected_output_dir'] = ex_folder
+
+        translated_f_name = 'translated_markdown.md'
+        generated_fp = plugin_kwargs['markdown_expected_output_path'] = os.path.join(ex_folder, translated_f_name)
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        yield from Markdown英译中(ex_folder, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request)
+        if os.path.exists(generated_fp):
+            # 修正一些公式问题
+            with open(generated_fp, 'r', encoding='utf8') as f: content = f.read()
+            content = content.replace('```markdown', '\n').replace('```', '\n')
+            # Markdown中使用不标准的表格，需要在表格前加上一个emoji，以便公式渲染
+            # content = re.sub(r'^<table>', r'.<table>', content, flags=re.MULTILINE)
+            with open(generated_fp, 'w', encoding='utf8') as f: f.write(content)
+            # 生成在线预览html
+            file_name = '在线预览翻译' + gen_time_str() + '.html'
+            preview_fp = os.path.join(ex_folder, file_name)
+            from shared_utils.advanced_markdown_format import markdown_convertion_for_file
+            with open(generated_fp, "r", encoding="utf-8") as f:
+                md = f.read()
+            html = markdown_convertion_for_file(md)
+            with open(preview_fp, "w", encoding="utf-8") as f: f.write(html)
+            promote_file_to_downloadzone(preview_fp, chatbot=chatbot)
+            # 生成包含图片的压缩包
+            dest_folder = get_log_folder(chatbot.get_user())
+            zip_name = '翻译后的带图文档.zip'
+            zip_folder(source_folder=ex_folder, dest_folder=dest_folder, zip_name=zip_name)
+            zip_fp = os.path.join(dest_folder, zip_name)
+            promote_file_to_downloadzone(zip_fp, chatbot=chatbot)
+            yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+    md_zip_path = yield from pdf2markdown(fp)
+    yield from deliver_to_markdown_plugin(md_zip_path, user_request)
+
+def 解析PDF_基于DOC2X(file_manifest, *args):
+    for index, fp in enumerate(file_manifest):
+        yield from 解析PDF_DOC2X_单文件(fp, *args)
+    return
+
+

crazy_functions/pdf_fns/parse_word.py

@@ -0,0 +1,85 @@
+from crazy_functions.crazy_utils import read_and_clean_pdf_text, get_files_from_everything
+import os
+import re
+def extract_text_from_files(txt, chatbot, history):
+    """
+    查找pdf/md/word并获取文本内容并返回状态以及文本
+
+    输入参数 Args:
+        chatbot: chatbot inputs and outputs （用户界面对话窗口句柄，用于数据流可视化）
+        history (list): List of chat history （历史，对话历史列表）
+
+    输出 Returns:
+        文件是否存在(bool)
+        final_result(list):文本内容
+        page_one(list):第一页内容/摘要
+        file_manifest(list):文件路径
+        excption(string):需要用户手动处理的信息,如没出错则保持为空
+    """
+
+    final_result = []
+    page_one = []
+    file_manifest = []
+    excption = ""
+
+    if txt == "":
+        final_result.append(txt)
+        return False, final_result, page_one, file_manifest, excption   #如输入区内容不是文件则直接返回输入区内容
+
+    #查找输入区内容中的文件
+    file_pdf,pdf_manifest,folder_pdf = get_files_from_everything(txt, '.pdf')
+    file_md,md_manifest,folder_md = get_files_from_everything(txt, '.md')
+    file_word,word_manifest,folder_word = get_files_from_everything(txt, '.docx')
+    file_doc,doc_manifest,folder_doc = get_files_from_everything(txt, '.doc')
+
+    if file_doc:
+        excption = "word"
+        return False, final_result, page_one, file_manifest, excption
+
+    file_num = len(pdf_manifest) + len(md_manifest) + len(word_manifest)
+    if file_num == 0:
+        final_result.append(txt)
+        return False, final_result, page_one, file_manifest, excption   #如输入区内容不是文件则直接返回输入区内容
+
+    if file_pdf:
+        try:    # 尝试导入依赖，如果缺少依赖，则给出安装建议
+            import fitz
+        except:
+            excption = "pdf"
+            return False, final_result, page_one, file_manifest, excption
+        for index, fp in enumerate(pdf_manifest):
+            file_content, pdf_one = read_and_clean_pdf_text(fp) # （尝试）按照章节切割PDF
+            file_content = file_content.encode('utf-8', 'ignore').decode()   # avoid reading non-utf8 chars
+            pdf_one = str(pdf_one).encode('utf-8', 'ignore').decode()  # avoid reading non-utf8 chars
+            final_result.append(file_content)
+            page_one.append(pdf_one)
+            file_manifest.append(os.path.relpath(fp, folder_pdf))
+
+    if file_md:
+        for index, fp in enumerate(md_manifest):
+            with open(fp, 'r', encoding='utf-8', errors='replace') as f:
+                file_content = f.read()
+            file_content = file_content.encode('utf-8', 'ignore').decode()
+            headers = re.findall(r'^#\s(.*)$', file_content, re.MULTILINE)  #接下来提取md中的一级/二级标题作为摘要
+            if len(headers) > 0:
+                page_one.append("\n".join(headers)) #合并所有的标题,以换行符分割
+            else:
+                page_one.append("")
+            final_result.append(file_content)
+            file_manifest.append(os.path.relpath(fp, folder_md))
+
+    if file_word:
+        try:    # 尝试导入依赖，如果缺少依赖，则给出安装建议
+            from docx import Document
+        except:
+            excption = "word_pip"
+            return False, final_result, page_one, file_manifest, excption
+        for index, fp in enumerate(word_manifest):
+            doc = Document(fp)
+            file_content = '\n'.join([p.text for p in doc.paragraphs])
+            file_content = file_content.encode('utf-8', 'ignore').decode()
+            page_one.append(file_content[:200])
+            final_result.append(file_content)
+            file_manifest.append(os.path.relpath(fp, folder_word))
+
+    return True, final_result, page_one, file_manifest, excption

crazy_functions/pdf_fns/report_gen_html.py

@@ -0,0 +1,58 @@
+from toolbox import update_ui, get_conf, trimmed_format_exc, get_log_folder
+import os
+
+
+
+
+class construct_html():
+    def __init__(self) -> None:
+        self.html_string = ""
+
+    def add_row(self, a, b):
+        from toolbox import markdown_convertion
+        template = """
+            {
+                primary_col: {
+                    header: String.raw`__PRIMARY_HEADER__`,
+                    msg: String.raw`__PRIMARY_MSG__`,
+                },
+                secondary_rol: {
+                    header: String.raw`__SECONDARY_HEADER__`,
+                    msg: String.raw`__SECONDARY_MSG__`,
+                }
+            },
+        """
+        def std(str):
+            str = str.replace(r'`',r'`')
+            if str.endswith("\\"): str += ' '
+            if str.endswith("}"): str += ' '
+            if str.endswith("$"): str += ' '
+            return str
+
+        template_ = template
+        a_lines = a.split('\n')
+        b_lines = b.split('\n')
+
+        if len(a_lines) == 1 or len(a_lines[0]) > 50:
+            template_ = template_.replace("__PRIMARY_HEADER__", std(a[:20]))
+            template_ = template_.replace("__PRIMARY_MSG__", std(markdown_convertion(a)))
+        else:
+            template_ = template_.replace("__PRIMARY_HEADER__", std(a_lines[0]))
+            template_ = template_.replace("__PRIMARY_MSG__", std(markdown_convertion('\n'.join(a_lines[1:]))))
+
+        if len(b_lines) == 1 or len(b_lines[0]) > 50:
+            template_ = template_.replace("__SECONDARY_HEADER__", std(b[:20]))
+            template_ = template_.replace("__SECONDARY_MSG__", std(markdown_convertion(b)))
+        else:
+            template_ = template_.replace("__SECONDARY_HEADER__", std(b_lines[0]))
+            template_ = template_.replace("__SECONDARY_MSG__", std(markdown_convertion('\n'.join(b_lines[1:]))))
+        self.html_string += template_
+
+    def save_file(self, file_name):
+        from toolbox import get_log_folder
+        with open('crazy_functions/pdf_fns/report_template.html', 'r', encoding='utf8') as f:
+            html_template = f.read()
+        html_template = html_template.replace("__TF_ARR__", self.html_string)
+        with open(os.path.join(get_log_folder(), file_name), 'w', encoding='utf8') as f:
+            f.write(html_template.encode('utf-8', 'ignore').decode())
+        return os.path.join(get_log_folder(), file_name)

crazy_functions/pdf_fns/report_template_v2.html

@@ -0,0 +1,73 @@
+<!DOCTYPE html>
+<html xmlns="http://www.w3.org/1999/xhtml">
+
+<head>
+    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
+    <title>GPT-Academic 翻译报告书</title>
+    <style>
+        .centered-a {
+            color: red;
+            text-align: center;
+            margin-bottom: 2%;
+            font-size: 1.5em;
+        }
+        .centered-b {
+            color: red;
+            text-align: center;
+            margin-top: 10%;
+            margin-bottom: 20%;
+            font-size: 1.5em;
+        }
+        .centered-c {
+            color: rgba(255, 0, 0, 0);
+            text-align: center;
+            margin-top: 2%;
+            margin-bottom: 20%;
+            font-size: 7em;
+        }
+    </style>
+<script>
+        // Configure MathJax settings
+        MathJax = {
+            tex: {
+                inlineMath: [
+                    ['$', '$'],
+                    ['\(', '\)']
+                ]
+            }
+        }
+        addEventListener('zero-md-rendered', () => {MathJax.typeset(); console.log('MathJax typeset!');})
+    </script>
+    <!-- Load MathJax library -->
+    <script src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-chtml.js"></script>
+    <script
+        type="module"
+        src="https://cdn.jsdelivr.net/gh/zerodevx/zero-md@2/dist/zero-md.min.js"
+    ></script>
+
+</head>
+
+<body>
+    <div class="test_temp1" style="width:10%; height: 500px; float:left;">
+
+    </div>
+    <div class="test_temp2" style="width:80%; height: 500px; float:left;">
+        <!-- Simply set the `src` attribute to your MD file and win -->
+        <div class="centered-a">
+            请按Ctrl+S保存此页面，否则该页面可能在几分钟后失效。
+        </div>
+        <zero-md src="translated_markdown.md" no-shadow>
+        </zero-md>
+        <div class="centered-b">
+            本报告由GPT-Academic开源项目生成，地址：https://github.com/binary-husky/gpt_academic。
+        </div>
+        <div class="centered-c">
+            本报告由GPT-Academic开源项目生成，地址：https://github.com/binary-husky/gpt_academic。
+        </div>
+    </div>
+    <div class="test_temp3" style="width:10%; height: 500px; float:left;">
+    </div>
+
+    </body>
+
+</html>

crazy_functions/plugin_template/plugin_class_template.py

@@ -0,0 +1,52 @@
+import os, json, base64
+from pydantic import BaseModel, Field
+from textwrap import dedent
+from typing import List
+
+class ArgProperty(BaseModel): # PLUGIN_ARG_MENU
+    title: str = Field(description="The title", default="")
+    description: str = Field(description="The description", default="")
+    default_value: str = Field(description="The default value", default="")
+    type: str = Field(description="The type", default="")   # currently we support ['string', 'dropdown']
+    options: List[str] = Field(default=[], description="List of options available for the argument") # only used when type is 'dropdown'
+
+class GptAcademicPluginTemplate():
+    def __init__(self):
+        # please note that `execute` method may run in different threads,
+        # thus you should not store any state in the plugin instance,
+        # which may be accessed by multiple threads
+        pass
+
+
+    def define_arg_selection_menu(self):
+        """
+        An example as below:
+            ```
+            def define_arg_selection_menu(self):
+                gui_definition = {
+                    "main_input":
+                        ArgProperty(title="main input", description="description", default_value="default_value", type="string").model_dump_json(),
+                    "advanced_arg":
+                        ArgProperty(title="advanced arguments", description="description", default_value="default_value", type="string").model_dump_json(),
+                    "additional_arg_01":
+                        ArgProperty(title="additional", description="description", default_value="default_value", type="string").model_dump_json(),
+                }
+                return gui_definition
+            ```
+        """
+        raise NotImplementedError("You need to implement this method in your plugin class")
+
+
+    def get_js_code_for_generating_menu(self, btnName):
+        define_arg_selection = self.define_arg_selection_menu()
+
+        if len(define_arg_selection.keys()) > 8:
+            raise ValueError("You can only have up to 8 arguments in the define_arg_selection")
+        # if "main_input" not in define_arg_selection:
+        #     raise ValueError("You must have a 'main_input' in the define_arg_selection")
+
+        DEFINE_ARG_INPUT_INTERFACE = json.dumps(define_arg_selection)
+        return base64.b64encode(DEFINE_ARG_INPUT_INTERFACE.encode('utf-8')).decode('utf-8')
+
+    def execute(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_request):
+        raise NotImplementedError("You need to implement this method in your plugin class")

crazy_functions/prompts/internet.py

@@ -0,0 +1,87 @@
+SearchOptimizerPrompt="""作为一个网页搜索助手，你的任务是结合历史记录，从不同角度，为“原问题”生成个不同版本的“检索词”，从而提高网页检索的精度。生成的问题要求指向对象清晰明确，并与“原问题语言相同”。例如：
+历史记录: 
+"
+Q: 对话背景。
+A: 当前对话是关于 Nginx 的介绍和在Ubuntu上的使用等。
+"
+原问题: 怎么下载
+检索词: ["Nginx 下载","Ubuntu Nginx","Ubuntu安装Nginx"]
+----------------
+历史记录: 
+"
+Q: 对话背景。
+A: 当前对话是关于 Nginx 的介绍和使用等。
+Q: 报错 "no connection"
+A: 报错"no connection"可能是因为……
+"
+原问题: 怎么解决
+检索词: ["Nginx报错"no connection" 解决","Nginx'no connection'报错 原因","Nginx提示'no connection'"]
+----------------
+历史记录:
+"
+
+"
+原问题: 你知道 Python 么？
+检索词: ["Python","Python 使用教程。","Python 特点和优势"]
+----------------
+历史记录:
+"
+Q: 列出Java的三种特点？
+A: 1. Java 是一种编译型语言。
+   2. Java 是一种面向对象的编程语言。
+   3. Java 是一种跨平台的编程语言。
+"
+原问题: 介绍下第2点。
+检索词: ["Java 面向对象特点","Java 面向对象编程优势。","Java 面向对象编程"]
+----------------
+现在有历史记录:
+"
+{history}
+"
+有其原问题: {query}
+直接给出最多{num}个检索词，必须以json形式给出，不得有多余字符:
+"""
+
+SearchAcademicOptimizerPrompt="""作为一个学术论文搜索助手，你的任务是结合历史记录，从不同角度，为“原问题”生成个不同版本的“检索词”，从而提高学术论文检索的精度。生成的问题要求指向对象清晰明确，并与“原问题语言相同”。例如：
+历史记录: 
+"
+Q: 对话背景。
+A: 当前对话是关于深度学习的介绍和在图像识别中的应用等。
+"
+原问题: 怎么下载相关论文
+检索词: ["深度学习 图像识别 论文下载","图像识别 深度学习 研究论文","深度学习 图像识别 论文资源","Deep Learning Image Recognition Paper Download","Image Recognition Deep Learning Research Paper"]
+----------------
+历史记录: 
+"
+Q: 对话背景。
+A: 当前对话是关于深度学习的介绍和应用等。
+Q: 报错 "模型不收敛"
+A: 报错"模型不收敛"可能是因为……
+"
+原问题: 怎么解决
+检索词: ["深度学习 模型不收敛 解决方案 论文","深度学习 模型不收敛 原因 研究","深度学习 模型不收敛 论文","Deep Learning Model Convergence Issue Solution Paper","Deep Learning Model Convergence Problem Research"]
+----------------
+历史记录:
+"
+
+"
+原问题: 你知道 GAN 么？
+检索词: ["生成对抗网络 论文","GAN 使用教程 论文","GAN 特点和优势 研究","Generative Adversarial Network Paper","GAN Usage Tutorial Paper"]
+----------------
+历史记录:
+"
+Q: 列出机器学习的三种应用？
+A: 1. 机器学习在图像识别中的应用。
+   2. 机器学习在自然语言处理中的应用。
+   3. 机器学习在推荐系统中的应用。
+"
+原问题: 介绍下第2点。
+检索词: ["机器学习 自然语言处理 应用 论文","机器学习 自然语言处理 研究","机器学习 NLP 应用 论文","Machine Learning Natural Language Processing Application Paper","Machine Learning NLP Research"]
+----------------
+现在有历史记录:
+"
+{history}
+"
+有其原问题: {query}
+直接给出最多{num}个检索词，必须以json形式给出，不得有多余字符:
+"""

crazy_functions/rag_fns/llama_index_worker.py

@@ -0,0 +1,130 @@
+import llama_index
+import os
+import atexit
+from loguru import logger
+from typing import List
+from llama_index.core import Document
+from llama_index.core.schema import TextNode
+from request_llms.embed_models.openai_embed import OpenAiEmbeddingModel
+from shared_utils.connect_void_terminal import get_chat_default_kwargs
+from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
+from crazy_functions.rag_fns.vector_store_index import GptacVectorStoreIndex
+from llama_index.core.ingestion import run_transformations
+from llama_index.core import PromptTemplate
+from llama_index.core.response_synthesizers import TreeSummarize
+
+DEFAULT_QUERY_GENERATION_PROMPT = """\
+Now, you have context information as below:
+---------------------
+{context_str}
+---------------------
+Answer the user request below (use the context information if necessary, otherwise you can ignore them):
+---------------------
+{query_str}
+"""
+
+QUESTION_ANSWER_RECORD = """\
+{{
+    "type": "This is a previous conversation with the user",
+    "question": "{question}",
+    "answer": "{answer}",
+}}
+"""
+
+
+class SaveLoad():
+
+    def does_checkpoint_exist(self, checkpoint_dir=None):
+        import os, glob
+        if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        if not os.path.exists(checkpoint_dir): return False
+        if len(glob.glob(os.path.join(checkpoint_dir, "*.json"))) == 0: return False
+        return True
+
+    def save_to_checkpoint(self, checkpoint_dir=None):
+        logger.info(f'saving vector store to: {checkpoint_dir}')
+        if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        self.vs_index.storage_context.persist(persist_dir=checkpoint_dir)
+
+    def load_from_checkpoint(self, checkpoint_dir=None):
+        if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        if self.does_checkpoint_exist(checkpoint_dir=checkpoint_dir):
+            logger.info('loading checkpoint from disk')
+            from llama_index.core import StorageContext, load_index_from_storage
+            storage_context = StorageContext.from_defaults(persist_dir=checkpoint_dir)
+            self.vs_index = load_index_from_storage(storage_context, embed_model=self.embed_model)
+            return self.vs_index
+        else:
+            return self.create_new_vs()
+
+    def create_new_vs(self):
+        return GptacVectorStoreIndex.default_vector_store(embed_model=self.embed_model)
+
+    def purge(self):
+        import shutil
+        shutil.rmtree(self.checkpoint_dir, ignore_errors=True)
+        self.vs_index = self.create_new_vs()
+
+
+class LlamaIndexRagWorker(SaveLoad):
+    def __init__(self, user_name, llm_kwargs, auto_load_checkpoint=True, checkpoint_dir=None) -> None:
+        self.debug_mode = True
+        self.embed_model = OpenAiEmbeddingModel(llm_kwargs)
+        self.user_name = user_name
+        self.checkpoint_dir = checkpoint_dir
+        if auto_load_checkpoint:
+            self.vs_index = self.load_from_checkpoint(checkpoint_dir)
+        else:
+            self.vs_index = self.create_new_vs(checkpoint_dir)
+        atexit.register(lambda: self.save_to_checkpoint(checkpoint_dir))
+
+    def assign_embedding_model(self):
+        pass
+
+    def inspect_vector_store(self):
+        # This function is for debugging
+        self.vs_index.storage_context.index_store.to_dict()
+        docstore = self.vs_index.storage_context.docstore.docs
+        vector_store_preview = "\n".join([ f"{_id} | {tn.text}" for _id, tn in docstore.items() ])
+        logger.info('\n++ --------inspect_vector_store begin--------')
+        logger.info(vector_store_preview)
+        logger.info('oo --------inspect_vector_store end--------')
+        return vector_store_preview
+
+    def add_documents_to_vector_store(self, document_list):
+        documents = [Document(text=t) for t in document_list]
+        documents_nodes = run_transformations(
+                        documents,  # type: ignore
+                        self.vs_index._transformations,
+                        show_progress=True
+                    )
+        self.vs_index.insert_nodes(documents_nodes)
+        if self.debug_mode: self.inspect_vector_store()
+
+    def add_text_to_vector_store(self, text):
+        node = TextNode(text=text)
+        documents_nodes = run_transformations(
+                        [node],
+                        self.vs_index._transformations,
+                        show_progress=True
+                    )
+        self.vs_index.insert_nodes(documents_nodes)
+        if self.debug_mode: self.inspect_vector_store()
+
+    def remember_qa(self, question, answer):
+        formatted_str = QUESTION_ANSWER_RECORD.format(question=question, answer=answer)
+        self.add_text_to_vector_store(formatted_str)
+
+    def retrieve_from_store_with_query(self, query):
+        if self.debug_mode: self.inspect_vector_store()
+        retriever = self.vs_index.as_retriever()
+        return retriever.retrieve(query)
+
+    def build_prompt(self, query, nodes):
+        context_str = self.generate_node_array_preview(nodes)
+        return DEFAULT_QUERY_GENERATION_PROMPT.format(context_str=context_str, query_str=query)
+        
+    def generate_node_array_preview(self, nodes):
+        buf = "\n".join(([f"(No.{i+1} | score {n.score:.3f}): {n.text}" for i, n in enumerate(nodes)]))
+        if self.debug_mode: logger.info(buf)
+        return buf

crazy_functions/rag_fns/milvus_worker.py

@@ -0,0 +1,108 @@
+import llama_index
+import os
+import atexit
+from typing import List
+from loguru import logger
+from llama_index.core import Document
+from llama_index.core.schema import TextNode
+from request_llms.embed_models.openai_embed import OpenAiEmbeddingModel
+from shared_utils.connect_void_terminal import get_chat_default_kwargs
+from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
+from crazy_functions.rag_fns.vector_store_index import GptacVectorStoreIndex
+from llama_index.core.ingestion import run_transformations
+from llama_index.core import PromptTemplate
+from llama_index.core.response_synthesizers import TreeSummarize
+from llama_index.core import StorageContext
+from llama_index.vector_stores.milvus import MilvusVectorStore
+from crazy_functions.rag_fns.llama_index_worker import LlamaIndexRagWorker
+
+DEFAULT_QUERY_GENERATION_PROMPT = """\
+Now, you have context information as below:
+---------------------
+{context_str}
+---------------------
+Answer the user request below (use the context information if necessary, otherwise you can ignore them):
+---------------------
+{query_str}
+"""
+
+QUESTION_ANSWER_RECORD = """\
+{{
+    "type": "This is a previous conversation with the user",
+    "question": "{question}",
+    "answer": "{answer}",
+}}
+"""
+
+
+class MilvusSaveLoad():
+
+    def does_checkpoint_exist(self, checkpoint_dir=None):
+        import os, glob
+        if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        if not os.path.exists(checkpoint_dir): return False
+        if len(glob.glob(os.path.join(checkpoint_dir, "*.json"))) == 0: return False
+        return True
+
+    def save_to_checkpoint(self, checkpoint_dir=None):
+        logger.info(f'saving vector store to: {checkpoint_dir}')
+        # if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        # self.vs_index.storage_context.persist(persist_dir=checkpoint_dir)
+
+    def load_from_checkpoint(self, checkpoint_dir=None):
+        if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        if self.does_checkpoint_exist(checkpoint_dir=checkpoint_dir):
+            logger.info('loading checkpoint from disk')
+            from llama_index.core import StorageContext, load_index_from_storage
+            storage_context = StorageContext.from_defaults(persist_dir=checkpoint_dir)
+            try:
+                self.vs_index = load_index_from_storage(storage_context, embed_model=self.embed_model)
+                return self.vs_index
+            except:
+                return self.create_new_vs(checkpoint_dir)
+        else:
+            return self.create_new_vs(checkpoint_dir)
+
+    def create_new_vs(self, checkpoint_dir, overwrite=False):
+        vector_store = MilvusVectorStore(
+            uri=os.path.join(checkpoint_dir, "milvus_demo.db"), 
+            dim=self.embed_model.embedding_dimension(),
+            overwrite=overwrite
+        )
+        storage_context = StorageContext.from_defaults(vector_store=vector_store)
+        index = GptacVectorStoreIndex.default_vector_store(storage_context=storage_context, embed_model=self.embed_model)
+        return index
+
+    def purge(self):
+        self.vs_index = self.create_new_vs(self.checkpoint_dir, overwrite=True)
+
+class MilvusRagWorker(MilvusSaveLoad, LlamaIndexRagWorker):
+
+    def __init__(self, user_name, llm_kwargs, auto_load_checkpoint=True, checkpoint_dir=None) -> None:
+        self.debug_mode = True
+        self.embed_model = OpenAiEmbeddingModel(llm_kwargs)
+        self.user_name = user_name
+        self.checkpoint_dir = checkpoint_dir
+        if auto_load_checkpoint:
+            self.vs_index = self.load_from_checkpoint(checkpoint_dir)
+        else:
+            self.vs_index = self.create_new_vs(checkpoint_dir)
+        atexit.register(lambda: self.save_to_checkpoint(checkpoint_dir))
+
+    def inspect_vector_store(self):
+        # This function is for debugging
+        try:
+            self.vs_index.storage_context.index_store.to_dict()
+            docstore = self.vs_index.storage_context.docstore.docs
+            if not docstore.items():
+                raise ValueError("cannot inspect")
+            vector_store_preview = "\n".join([ f"{_id} | {tn.text}" for _id, tn in docstore.items() ])
+        except:
+            dummy_retrieve_res: List["NodeWithScore"] = self.vs_index.as_retriever().retrieve(' ')
+            vector_store_preview = "\n".join(
+                [f"{node.id_} | {node.text}" for node in dummy_retrieve_res]
+            )
+        logger.info('\n++ --------inspect_vector_store begin--------')
+        logger.info(vector_store_preview)
+        logger.info('oo --------inspect_vector_store end--------')
+        return vector_store_preview

crazy_functions/rag_fns/vector_store_index.py

@@ -0,0 +1,58 @@
+from llama_index.core import VectorStoreIndex
+from typing import Any,  List, Optional
+
+from llama_index.core.callbacks.base import CallbackManager
+from llama_index.core.schema import TransformComponent
+from llama_index.core.service_context import ServiceContext
+from llama_index.core.settings import (
+    Settings,
+    callback_manager_from_settings_or_context,
+    transformations_from_settings_or_context,
+)
+from llama_index.core.storage.storage_context import StorageContext
+
+
+class GptacVectorStoreIndex(VectorStoreIndex):
+    
+    @classmethod
+    def default_vector_store(
+        cls,
+        storage_context: Optional[StorageContext] = None,
+        show_progress: bool = False,
+        callback_manager: Optional[CallbackManager] = None,
+        transformations: Optional[List[TransformComponent]] = None,
+        # deprecated
+        service_context: Optional[ServiceContext] = None,
+        embed_model = None,
+        **kwargs: Any,
+    ):
+        """Create index from documents.
+
+        Args:
+            documents (Optional[Sequence[BaseDocument]]): List of documents to
+                build the index from.
+
+        """
+        storage_context = storage_context or StorageContext.from_defaults()
+        docstore = storage_context.docstore
+        callback_manager = (
+            callback_manager
+            or callback_manager_from_settings_or_context(Settings, service_context)
+        )
+        transformations = transformations or transformations_from_settings_or_context(
+            Settings, service_context
+        )
+
+        with callback_manager.as_trace("index_construction"):
+
+            return cls(
+                nodes=[],
+                storage_context=storage_context,
+                callback_manager=callback_manager,
+                show_progress=show_progress,
+                transformations=transformations,
+                service_context=service_context,
+                embed_model=embed_model,
+                **kwargs,
+            )
+

crazy_functions/test_project/cpp/cppipc/buffer.cpp

@@ -1,87 +0,0 @@
-#include "libipc/buffer.h"
-#include "libipc/utility/pimpl.h"
-
-#include <cstring>
-
-namespace ipc {
-
-bool operator==(buffer const & b1, buffer const & b2) {
-    return (b1.size() == b2.size()) && (std::memcmp(b1.data(), b2.data(), b1.size()) == 0);
-}
-
-bool operator!=(buffer const & b1, buffer const & b2) {
-    return !(b1 == b2);
-}
-
-class buffer::buffer_ : public pimpl<buffer_> {
-public:
-    void*       p_;
-    std::size_t s_;
-    void*       a_;
-    buffer::destructor_t d_;
-
-    buffer_(void* p, std::size_t s, buffer::destructor_t d, void* a)
-        : p_(p), s_(s), a_(a), d_(d) {
-    }
-
-    ~buffer_() {
-        if (d_ == nullptr) return;
-        d_((a_ == nullptr) ? p_ : a_, s_);
-    }
-};
-
-buffer::buffer()
-    : buffer(nullptr, 0, nullptr, nullptr) {
-}
-
-buffer::buffer(void* p, std::size_t s, destructor_t d)
-    : p_(p_->make(p, s, d, nullptr)) {
-}
-
-buffer::buffer(void* p, std::size_t s, destructor_t d, void* additional)
-    : p_(p_->make(p, s, d, additional)) {
-}
-
-buffer::buffer(void* p, std::size_t s)
-    : buffer(p, s, nullptr) {
-}
-
-buffer::buffer(char const & c)
-    : buffer(const_cast<char*>(&c), 1) {
-}
-
-buffer::buffer(buffer&& rhs)
-    : buffer() {
-    swap(rhs);
-}
-
-buffer::~buffer() {
-    p_->clear();
-}
-
-void buffer::swap(buffer& rhs) {
-    std::swap(p_, rhs.p_);
-}
-
-buffer& buffer::operator=(buffer rhs) {
-    swap(rhs);
-    return *this;
-}
-
-bool buffer::empty() const noexcept {
-    return (impl(p_)->p_ == nullptr) || (impl(p_)->s_ == 0);
-}
-
-void* buffer::data() noexcept {
-    return impl(p_)->p_;
-}
-
-void const * buffer::data() const noexcept {
-    return impl(p_)->p_;
-}
-
-std::size_t buffer::size() const noexcept {
-    return impl(p_)->s_;
-}
-
-} // namespace ipc

crazy_functions/test_project/cpp/cppipc/ipc.cpp

@@ -1,701 +0,0 @@
-
-#include <type_traits>
-#include <cstring>
-#include <algorithm>
-#include <utility>          // std::pair, std::move, std::forward
-#include <atomic>
-#include <type_traits>      // aligned_storage_t
-#include <string>
-#include <vector>
-#include <array>
-#include <cassert>
-
-#include "libipc/ipc.h"
-#include "libipc/def.h"
-#include "libipc/shm.h"
-#include "libipc/pool_alloc.h"
-#include "libipc/queue.h"
-#include "libipc/policy.h"
-#include "libipc/rw_lock.h"
-#include "libipc/waiter.h"
-
-#include "libipc/utility/log.h"
-#include "libipc/utility/id_pool.h"
-#include "libipc/utility/scope_guard.h"
-#include "libipc/utility/utility.h"
-
-#include "libipc/memory/resource.h"
-#include "libipc/platform/detail.h"
-#include "libipc/circ/elem_array.h"
-
-namespace {
-
-using msg_id_t = std::uint32_t;
-using acc_t    = std::atomic<msg_id_t>;
-
-template <std::size_t DataSize, std::size_t AlignSize>
-struct msg_t;
-
-template <std::size_t AlignSize>
-struct msg_t<0, AlignSize> {
-    msg_id_t     cc_id_;
-    msg_id_t     id_;
-    std::int32_t remain_;
-    bool         storage_;
-};
-
-template <std::size_t DataSize, std::size_t AlignSize>
-struct msg_t : msg_t<0, AlignSize> {
-    std::aligned_storage_t<DataSize, AlignSize> data_ {};
-
-    msg_t() = default;
-    msg_t(msg_id_t cc_id, msg_id_t id, std::int32_t remain, void const * data, std::size_t size)
-        : msg_t<0, AlignSize> {cc_id, id, remain, (data == nullptr) || (size == 0)} {
-        if (this->storage_) {
-            if (data != nullptr) {
-                // copy storage-id
-                *reinterpret_cast<ipc::storage_id_t*>(&data_) =
-                     *static_cast<ipc::storage_id_t const *>(data);
-            }
-        }
-        else std::memcpy(&data_, data, size);
-    }
-};
-
-template <typename T>
-ipc::buff_t make_cache(T& data, std::size_t size) {
-    auto ptr = ipc::mem::alloc(size);
-    std::memcpy(ptr, &data, (ipc::detail::min)(sizeof(data), size));
-    return { ptr, size, ipc::mem::free };
-}
-
-struct cache_t {
-    std::size_t fill_;
-    ipc::buff_t buff_;
-
-    cache_t(std::size_t f, ipc::buff_t && b)
-        : fill_(f), buff_(std::move(b))
-    {}
-
-    void append(void const * data, std::size_t size) {
-        if (fill_ >= buff_.size() || data == nullptr || size == 0) return;
-        auto new_fill = (ipc::detail::min)(fill_ + size, buff_.size());
-        std::memcpy(static_cast<ipc::byte_t*>(buff_.data()) + fill_, data, new_fill - fill_);
-        fill_ = new_fill;
-    }
-};
-
-auto cc_acc() {
-    static ipc::shm::handle acc_h("__CA_CONN__", sizeof(acc_t));
-    return static_cast<acc_t*>(acc_h.get());
-}
-
-IPC_CONSTEXPR_ std::size_t align_chunk_size(std::size_t size) noexcept {
-    return (((size - 1) / ipc::large_msg_align) + 1) * ipc::large_msg_align;
-}
-
-IPC_CONSTEXPR_ std::size_t calc_chunk_size(std::size_t size) noexcept {
-    return ipc::make_align(alignof(std::max_align_t), align_chunk_size(
-           ipc::make_align(alignof(std::max_align_t), sizeof(std::atomic<ipc::circ::cc_t>)) + size));
-}
-
-struct chunk_t {
-    std::atomic<ipc::circ::cc_t> &conns() noexcept {
-        return *reinterpret_cast<std::atomic<ipc::circ::cc_t> *>(this);
-    }
-
-    void *data() noexcept {
-        return reinterpret_cast<ipc::byte_t *>(this)
-             + ipc::make_align(alignof(std::max_align_t), sizeof(std::atomic<ipc::circ::cc_t>));
-    }
-};
-
-struct chunk_info_t {
-    ipc::id_pool<> pool_;
-    ipc::spin_lock lock_;
-
-    IPC_CONSTEXPR_ static std::size_t chunks_mem_size(std::size_t chunk_size) noexcept {
-        return ipc::id_pool<>::max_count * chunk_size;
-    }
-
-    ipc::byte_t *chunks_mem() noexcept {
-        return reinterpret_cast<ipc::byte_t *>(this + 1);
-    }
-
-    chunk_t *at(std::size_t chunk_size, ipc::storage_id_t id) noexcept {
-        if (id < 0) return nullptr;
-        return reinterpret_cast<chunk_t *>(chunks_mem() + (chunk_size * id));
-    }
-};
-
-auto& chunk_storages() {
-    class chunk_handle_t {
-        ipc::shm::handle handle_;
-
-    public:
-        chunk_info_t *get_info(std::size_t chunk_size) {
-            if (!handle_.valid() &&
-                !handle_.acquire( ("__CHUNK_INFO__" + ipc::to_string(chunk_size)).c_str(), 
-                                  sizeof(chunk_info_t) + chunk_info_t::chunks_mem_size(chunk_size) )) {
-                ipc::error("[chunk_storages] chunk_shm.id_info_.acquire failed: chunk_size = %zd\n", chunk_size);
-                return nullptr;
-            }
-            auto info = static_cast<chunk_info_t*>(handle_.get());
-            if (info == nullptr) {
-                ipc::error("[chunk_storages] chunk_shm.id_info_.get failed: chunk_size = %zd\n", chunk_size);
-                return nullptr;
-            }
-            return info;
-        }
-    };
-    static ipc::map<std::size_t, chunk_handle_t> chunk_hs;
-    return chunk_hs;
-}
-
-chunk_info_t *chunk_storage_info(std::size_t chunk_size) {
-    auto &storages = chunk_storages();
-    std::decay_t<decltype(storages)>::iterator it;
-    {
-        static ipc::rw_lock lock;
-        IPC_UNUSED_ std::shared_lock<ipc::rw_lock> guard {lock};
-        if ((it = storages.find(chunk_size)) == storages.end()) {
-            using chunk_handle_t = std::decay_t<decltype(storages)>::value_type::second_type;
-            guard.unlock();
-            IPC_UNUSED_ std::lock_guard<ipc::rw_lock> guard {lock};
-            it = storages.emplace(chunk_size, chunk_handle_t{}).first;
-        }
-    }
-    return it->second.get_info(chunk_size);
-}
-
-std::pair<ipc::storage_id_t, void*> acquire_storage(std::size_t size, ipc::circ::cc_t conns) {
-    std::size_t chunk_size = calc_chunk_size(size);
-    auto info = chunk_storage_info(chunk_size);
-    if (info == nullptr) return {};
-
-    info->lock_.lock();
-    info->pool_.prepare();
-    // got an unique id
-    auto id = info->pool_.acquire();
-    info->lock_.unlock();
-
-    auto chunk = info->at(chunk_size, id);
-    if (chunk == nullptr) return {};
-    chunk->conns().store(conns, std::memory_order_relaxed);
-    return { id, chunk->data() };
-}
-
-void *find_storage(ipc::storage_id_t id, std::size_t size) {
-    if (id < 0) {
-        ipc::error("[find_storage] id is invalid: id = %ld, size = %zd\n", (long)id, size);
-        return nullptr;
-    }
-    std::size_t chunk_size = calc_chunk_size(size);
-    auto info = chunk_storage_info(chunk_size);
-    if (info == nullptr) return nullptr;
-    return info->at(chunk_size, id)->data();
-}
-
-void release_storage(ipc::storage_id_t id, std::size_t size) {
-    if (id < 0) {
-        ipc::error("[release_storage] id is invalid: id = %ld, size = %zd\n", (long)id, size);
-        return;
-    }
-    std::size_t chunk_size = calc_chunk_size(size);
-    auto info = chunk_storage_info(chunk_size);
-    if (info == nullptr) return;
-    info->lock_.lock();
-    info->pool_.release(id);
-    info->lock_.unlock();
-}
-
-template <ipc::relat Rp, ipc::relat Rc>
-bool sub_rc(ipc::wr<Rp, Rc, ipc::trans::unicast>, 
-            std::atomic<ipc::circ::cc_t> &/*conns*/, ipc::circ::cc_t /*curr_conns*/, ipc::circ::cc_t /*conn_id*/) noexcept {
-    return true;
-}
-
-template <ipc::relat Rp, ipc::relat Rc>
-bool sub_rc(ipc::wr<Rp, Rc, ipc::trans::broadcast>, 
-            std::atomic<ipc::circ::cc_t> &conns, ipc::circ::cc_t curr_conns, ipc::circ::cc_t conn_id) noexcept {
-    auto last_conns = curr_conns & ~conn_id;
-    for (unsigned k = 0;;) {
-        auto chunk_conns  = conns.load(std::memory_order_acquire);
-        if (conns.compare_exchange_weak(chunk_conns, chunk_conns & last_conns, std::memory_order_release)) {
-            return (chunk_conns & last_conns) == 0;
-        }
-        ipc::yield(k);
-    }
-}
-
-template <typename Flag>
-void recycle_storage(ipc::storage_id_t id, std::size_t size, ipc::circ::cc_t curr_conns, ipc::circ::cc_t conn_id) {
-    if (id < 0) {
-        ipc::error("[recycle_storage] id is invalid: id = %ld, size = %zd\n", (long)id, size);
-        return;
-    }
-    std::size_t chunk_size = calc_chunk_size(size);
-    auto info = chunk_storage_info(chunk_size);
-    if (info == nullptr) return;
-
-    auto chunk = info->at(chunk_size, id);
-    if (chunk == nullptr) return;
-
-    if (!sub_rc(Flag{}, chunk->conns(), curr_conns, conn_id)) {
-        return;
-    }
-    info->lock_.lock();
-    info->pool_.release(id);
-    info->lock_.unlock();
-}
-
-template <typename MsgT>
-bool clear_message(void* p) {
-    auto msg = static_cast<MsgT*>(p);
-    if (msg->storage_) {
-        std::int32_t r_size = static_cast<std::int32_t>(ipc::data_length) + msg->remain_;
-        if (r_size <= 0) {
-            ipc::error("[clear_message] invalid msg size: %d\n", (int)r_size);
-            return true;
-        }
-        release_storage(
-            *reinterpret_cast<ipc::storage_id_t*>(&msg->data_),
-            static_cast<std::size_t>(r_size));
-    }
-    return true;
-}
-
-struct conn_info_head {
-
-    ipc::string name_;
-    msg_id_t    cc_id_; // connection-info id
-    ipc::detail::waiter cc_waiter_, wt_waiter_, rd_waiter_;
-    ipc::shm::handle acc_h_;
-
-    conn_info_head(char const * name)
-        : name_     {name}
-        , cc_id_    {(cc_acc() == nullptr) ? 0 : cc_acc()->fetch_add(1, std::memory_order_relaxed)}
-        , cc_waiter_{("__CC_CONN__" + name_).c_str()}
-        , wt_waiter_{("__WT_CONN__" + name_).c_str()}
-        , rd_waiter_{("__RD_CONN__" + name_).c_str()}
-        , acc_h_    {("__AC_CONN__" + name_).c_str(), sizeof(acc_t)} {
-    }
-
-    void quit_waiting() {
-        cc_waiter_.quit_waiting();
-        wt_waiter_.quit_waiting();
-        rd_waiter_.quit_waiting();
-    }
-
-    auto acc() {
-        return static_cast<acc_t*>(acc_h_.get());
-    }
-
-    auto& recv_cache() {
-        thread_local ipc::unordered_map<msg_id_t, cache_t> tls;
-        return tls;
-    }
-};
-
-template <typename W, typename F>
-bool wait_for(W& waiter, F&& pred, std::uint64_t tm) {
-    if (tm == 0) return !pred();
-    for (unsigned k = 0; pred();) {
-        bool ret = true;
-        ipc::sleep(k, [&k, &ret, &waiter, &pred, tm] {
-            ret = waiter.wait_if(std::forward<F>(pred), tm);
-            k   = 0;
-        });
-        if (!ret) return false; // timeout or fail
-        if (k == 0) break; // k has been reset
-    }
-    return true;
-}
-
-template <typename Policy,
-          std::size_t DataSize  = ipc::data_length,
-          std::size_t AlignSize = (ipc::detail::min)(DataSize, alignof(std::max_align_t))>
-struct queue_generator {
-
-    using queue_t = ipc::queue<msg_t<DataSize, AlignSize>, Policy>;
-
-    struct conn_info_t : conn_info_head {
-        queue_t que_;
-
-        conn_info_t(char const * name)
-            : conn_info_head{name}
-            , que_{("__QU_CONN__" +
-                    ipc::to_string(DataSize) + "__" +
-                    ipc::to_string(AlignSize) + "__" + name).c_str()} {
-        }
-
-        void disconnect_receiver() {
-            bool dis = que_.disconnect();
-            this->quit_waiting();
-            if (dis) {
-                this->recv_cache().clear();
-            }
-        }
-    };
-};
-
-template <typename Policy>
-struct detail_impl {
-
-using policy_t    = Policy;
-using flag_t      = typename policy_t::flag_t;
-using queue_t     = typename queue_generator<policy_t>::queue_t;
-using conn_info_t = typename queue_generator<policy_t>::conn_info_t;
-
-constexpr static conn_info_t* info_of(ipc::handle_t h) noexcept {
-    return static_cast<conn_info_t*>(h);
-}
-
-constexpr static queue_t* queue_of(ipc::handle_t h) noexcept {
-    return (info_of(h) == nullptr) ? nullptr : &(info_of(h)->que_);
-}
-
-/* API implementations */
-
-static void disconnect(ipc::handle_t h) {
-    auto que = queue_of(h);
-    if (que == nullptr) {
-        return;
-    }
-    que->shut_sending();
-    assert(info_of(h) != nullptr);
-    info_of(h)->disconnect_receiver();
-}
-
-static bool reconnect(ipc::handle_t * ph, bool start_to_recv) {
-    assert(ph != nullptr);
-    assert(*ph != nullptr);
-    auto que = queue_of(*ph);
-    if (que == nullptr) {
-        return false;
-    }
-    if (start_to_recv) {
-        que->shut_sending();
-        if (que->connect()) { // wouldn't connect twice
-            info_of(*ph)->cc_waiter_.broadcast();
-            return true;
-        }
-        return false;
-    }
-    // start_to_recv == false
-    if (que->connected()) {
-        info_of(*ph)->disconnect_receiver();
-    }
-    return que->ready_sending();
-}
-
-static bool connect(ipc::handle_t * ph, char const * name, bool start_to_recv) {
-    assert(ph != nullptr);
-    if (*ph == nullptr) {
-        *ph = ipc::mem::alloc<conn_info_t>(name);
-    }
-    return reconnect(ph, start_to_recv);
-}
-
-static void destroy(ipc::handle_t h) {
-    disconnect(h);
-    ipc::mem::free(info_of(h));
-}
-
-static std::size_t recv_count(ipc::handle_t h) noexcept {
-    auto que = queue_of(h);
-    if (que == nullptr) {
-        return ipc::invalid_value;
-    }
-    return que->conn_count();
-}
-
-static bool wait_for_recv(ipc::handle_t h, std::size_t r_count, std::uint64_t tm) {
-    auto que = queue_of(h);
-    if (que == nullptr) {
-        return false;
-    }
-    return wait_for(info_of(h)->cc_waiter_, [que, r_count] {
-        return que->conn_count() < r_count;
-    }, tm);
-}
-
-template <typename F>
-static bool send(F&& gen_push, ipc::handle_t h, void const * data, std::size_t size) {
-    if (data == nullptr || size == 0) {
-        ipc::error("fail: send(%p, %zd)\n", data, size);
-        return false;
-    }
-    auto que = queue_of(h);
-    if (que == nullptr) {
-        ipc::error("fail: send, queue_of(h) == nullptr\n");
-        return false;
-    }
-    if (que->elems() == nullptr) {
-        ipc::error("fail: send, queue_of(h)->elems() == nullptr\n");
-        return false;
-    }
-    if (!que->ready_sending()) {
-        ipc::error("fail: send, que->ready_sending() == false\n");
-        return false;
-    }
-    ipc::circ::cc_t conns = que->elems()->connections(std::memory_order_relaxed);
-    if (conns == 0) {
-        ipc::error("fail: send, there is no receiver on this connection.\n");
-        return false;
-    }
-    // calc a new message id
-    auto acc = info_of(h)->acc();
-    if (acc == nullptr) {
-        ipc::error("fail: send, info_of(h)->acc() == nullptr\n");
-        return false;
-    }
-    auto msg_id   = acc->fetch_add(1, std::memory_order_relaxed);
-    auto try_push = std::forward<F>(gen_push)(info_of(h), que, msg_id);
-    if (size > ipc::large_msg_limit) {
-        auto   dat = acquire_storage(size, conns);
-        void * buf = dat.second;
-        if (buf != nullptr) {
-            std::memcpy(buf, data, size);
-            return try_push(static_cast<std::int32_t>(size) - 
-                            static_cast<std::int32_t>(ipc::data_length), &(dat.first), 0);
-        }
-        // try using message fragment
-        //ipc::log("fail: shm::handle for big message. msg_id: %zd, size: %zd\n", msg_id, size);
-    }
-    // push message fragment
-    std::int32_t offset = 0;
-    for (std::int32_t i = 0; i < static_cast<std::int32_t>(size / ipc::data_length); ++i, offset += ipc::data_length) {
-        if (!try_push(static_cast<std::int32_t>(size) - offset - static_cast<std::int32_t>(ipc::data_length),
-                      static_cast<ipc::byte_t const *>(data) + offset, ipc::data_length)) {
-            return false;
-        }
-    }
-    // if remain > 0, this is the last message fragment
-    std::int32_t remain = static_cast<std::int32_t>(size) - offset;
-    if (remain > 0) {
-        if (!try_push(remain - static_cast<std::int32_t>(ipc::data_length),
-                      static_cast<ipc::byte_t const *>(data) + offset, 
-                      static_cast<std::size_t>(remain))) {
-            return false;
-        }
-    }
-    return true;
-}
-
-static bool send(ipc::handle_t h, void const * data, std::size_t size, std::uint64_t tm) {
-    return send([tm](auto info, auto que, auto msg_id) {
-        return [tm, info, que, msg_id](std::int32_t remain, void const * data, std::size_t size) {
-            if (!wait_for(info->wt_waiter_, [&] {
-                    return !que->push(
-                        [](void*) { return true; },
-                        info->cc_id_, msg_id, remain, data, size);
-                }, tm)) {
-                ipc::log("force_push: msg_id = %zd, remain = %d, size = %zd\n", msg_id, remain, size);
-                if (!que->force_push(
-                        clear_message<typename queue_t::value_t>,
-                        info->cc_id_, msg_id, remain, data, size)) {
-                    return false;
-                }
-            }
-            info->rd_waiter_.broadcast();
-            return true;
-        };
-    }, h, data, size);
-}
-
-static bool try_send(ipc::handle_t h, void const * data, std::size_t size, std::uint64_t tm) {
-    return send([tm](auto info, auto que, auto msg_id) {
-        return [tm, info, que, msg_id](std::int32_t remain, void const * data, std::size_t size) {
-            if (!wait_for(info->wt_waiter_, [&] {
-                    return !que->push(
-                        [](void*) { return true; },
-                        info->cc_id_, msg_id, remain, data, size);
-                }, tm)) {
-                return false;
-            }
-            info->rd_waiter_.broadcast();
-            return true;
-        };
-    }, h, data, size);
-}
-
-static ipc::buff_t recv(ipc::handle_t h, std::uint64_t tm) {
-    auto que = queue_of(h);
-    if (que == nullptr) {
-        ipc::error("fail: recv, queue_of(h) == nullptr\n");
-        return {};
-    }
-    if (!que->connected()) {
-        // hasn't connected yet, just return.
-        return {};
-    }
-    auto& rc = info_of(h)->recv_cache();
-    for (;;) {
-        // pop a new message
-        typename queue_t::value_t msg;
-        if (!wait_for(info_of(h)->rd_waiter_, [que, &msg] {
-                return !que->pop(msg);
-            }, tm)) {
-            // pop failed, just return.
-            return {};
-        }
-        info_of(h)->wt_waiter_.broadcast();
-        if ((info_of(h)->acc() != nullptr) && (msg.cc_id_ == info_of(h)->cc_id_)) {
-            continue; // ignore message to self
-        }
-        // msg.remain_ may minus & abs(msg.remain_) < data_length
-        std::int32_t r_size = static_cast<std::int32_t>(ipc::data_length) + msg.remain_;
-        if (r_size <= 0) {
-            ipc::error("fail: recv, r_size = %d\n", (int)r_size);
-            return {};
-        }
-        std::size_t msg_size = static_cast<std::size_t>(r_size);
-        // large message
-        if (msg.storage_) {
-            ipc::storage_id_t buf_id = *reinterpret_cast<ipc::storage_id_t*>(&msg.data_);
-            void* buf = find_storage(buf_id, msg_size);
-            if (buf != nullptr) {
-                struct recycle_t {
-                    ipc::storage_id_t storage_id;
-                    ipc::circ::cc_t   curr_conns;
-                    ipc::circ::cc_t   conn_id;
-                } *r_info = ipc::mem::alloc<recycle_t>(recycle_t{
-                    buf_id, que->elems()->connections(std::memory_order_relaxed), que->connected_id()
-                });
-                if (r_info == nullptr) {
-                    ipc::log("fail: ipc::mem::alloc<recycle_t>.\n");
-                    return ipc::buff_t{buf, msg_size}; // no recycle
-                } else {
-                    return ipc::buff_t{buf, msg_size, [](void* p_info, std::size_t size) {
-                        auto r_info = static_cast<recycle_t *>(p_info);
-                        IPC_UNUSED_ auto finally = ipc::guard([r_info] {
-                            ipc::mem::free(r_info);
-                        });
-                        recycle_storage<flag_t>(r_info->storage_id, size, r_info->curr_conns, r_info->conn_id);
-                    }, r_info};
-                }
-            } else {
-                ipc::log("fail: shm::handle for large message. msg_id: %zd, buf_id: %zd, size: %zd\n", msg.id_, buf_id, msg_size);
-                continue;
-            }
-        }
-        // find cache with msg.id_
-        auto cac_it = rc.find(msg.id_);
-        if (cac_it == rc.end()) {
-            if (msg_size <= ipc::data_length) {
-                return make_cache(msg.data_, msg_size);
-            }
-            // gc
-            if (rc.size() > 1024) {
-                std::vector<msg_id_t> need_del;
-                for (auto const & pair : rc) {
-                    auto cmp = std::minmax(msg.id_, pair.first);
-                    if (cmp.second - cmp.first > 8192) {
-                        need_del.push_back(pair.first);
-                    }
-                }
-                for (auto id : need_del) rc.erase(id);
-            }
-            // cache the first message fragment
-            rc.emplace(msg.id_, cache_t { ipc::data_length, make_cache(msg.data_, msg_size) });
-        }
-        // has cached before this message
-        else {
-            auto& cac = cac_it->second;
-            // this is the last message fragment
-            if (msg.remain_ <= 0) {
-                cac.append(&(msg.data_), msg_size);
-                // finish this message, erase it from cache
-                auto buff = std::move(cac.buff_);
-                rc.erase(cac_it);
-                return buff;
-            }
-            // there are remain datas after this message
-            cac.append(&(msg.data_), ipc::data_length);
-        }
-    }
-}
-
-static ipc::buff_t try_recv(ipc::handle_t h) {
-    return recv(h, 0);
-}
-
-}; // detail_impl<Policy>
-
-template <typename Flag>
-using policy_t = ipc::policy::choose<ipc::circ::elem_array, Flag>;
-
-} // internal-linkage
-
-namespace ipc {
-
-template <typename Flag>
-ipc::handle_t chan_impl<Flag>::inited() {
-    ipc::detail::waiter::init();
-    return nullptr;
-}
-
-template <typename Flag>
-bool chan_impl<Flag>::connect(ipc::handle_t * ph, char const * name, unsigned mode) {
-    return detail_impl<policy_t<Flag>>::connect(ph, name, mode & receiver);
-}
-
-template <typename Flag>
-bool chan_impl<Flag>::reconnect(ipc::handle_t * ph, unsigned mode) {
-    return detail_impl<policy_t<Flag>>::reconnect(ph, mode & receiver);
-}
-
-template <typename Flag>
-void chan_impl<Flag>::disconnect(ipc::handle_t h) {
-    detail_impl<policy_t<Flag>>::disconnect(h);
-}
-
-template <typename Flag>
-void chan_impl<Flag>::destroy(ipc::handle_t h) {
-    detail_impl<policy_t<Flag>>::destroy(h);
-}
-
-template <typename Flag>
-char const * chan_impl<Flag>::name(ipc::handle_t h) {
-    auto info = detail_impl<policy_t<Flag>>::info_of(h);
-    return (info == nullptr) ? nullptr : info->name_.c_str();
-}
-
-template <typename Flag>
-std::size_t chan_impl<Flag>::recv_count(ipc::handle_t h) {
-    return detail_impl<policy_t<Flag>>::recv_count(h);
-}
-
-template <typename Flag>
-bool chan_impl<Flag>::wait_for_recv(ipc::handle_t h, std::size_t r_count, std::uint64_t tm) {
-    return detail_impl<policy_t<Flag>>::wait_for_recv(h, r_count, tm);
-}
-
-template <typename Flag>
-bool chan_impl<Flag>::send(ipc::handle_t h, void const * data, std::size_t size, std::uint64_t tm) {
-    return detail_impl<policy_t<Flag>>::send(h, data, size, tm);
-}
-
-template <typename Flag>
-buff_t chan_impl<Flag>::recv(ipc::handle_t h, std::uint64_t tm) {
-    return detail_impl<policy_t<Flag>>::recv(h, tm);
-}
-
-template <typename Flag>
-bool chan_impl<Flag>::try_send(ipc::handle_t h, void const * data, std::size_t size, std::uint64_t tm) {
-    return detail_impl<policy_t<Flag>>::try_send(h, data, size, tm);
-}
-
-template <typename Flag>
-buff_t chan_impl<Flag>::try_recv(ipc::handle_t h) {
-    return detail_impl<policy_t<Flag>>::try_recv(h);
-}
-
-template struct chan_impl<ipc::wr<relat::single, relat::single, trans::unicast  >>;
-// template struct chan_impl<ipc::wr<relat::single, relat::multi , trans::unicast  >>; // TBD
-// template struct chan_impl<ipc::wr<relat::multi , relat::multi , trans::unicast  >>; // TBD
-template struct chan_impl<ipc::wr<relat::single, relat::multi , trans::broadcast>>;
-template struct chan_impl<ipc::wr<relat::multi , relat::multi , trans::broadcast>>;
-
-} // namespace ipc

crazy_functions/test_project/cpp/cppipc/policy.h

@@ -1,25 +0,0 @@
-#pragma once
-
-#include <type_traits>
-
-#include "libipc/def.h"
-#include "libipc/prod_cons.h"
-
-#include "libipc/circ/elem_array.h"
-
-namespace ipc {
-namespace policy {
-
-template <template <typename, std::size_t...> class Elems, typename Flag>
-struct choose;
-
-template <typename Flag>
-struct choose<circ::elem_array, Flag> {
-    using flag_t = Flag;
-
-    template <std::size_t DataSize, std::size_t AlignSize>
-    using elems_t = circ::elem_array<ipc::prod_cons_impl<flag_t>, DataSize, AlignSize>;
-};
-
-} // namespace policy
-} // namespace ipc

crazy_functions/test_project/cpp/cppipc/pool_alloc.cpp

@@ -1,17 +0,0 @@
-#include "libipc/pool_alloc.h"
-
-#include "libipc/memory/resource.h"
-
-namespace ipc {
-namespace mem {
-
-void* pool_alloc::alloc(std::size_t size) {
-    return async_pool_alloc::alloc(size);
-}
-
-void pool_alloc::free(void* p, std::size_t size) {
-    async_pool_alloc::free(p, size);
-}
-
-} // namespace mem
-} // namespace ipc

crazy_functions/test_project/cpp/cppipc/prod_cons.h

@@ -1,433 +0,0 @@
-#pragma once
-
-#include <atomic>
-#include <utility>
-#include <cstring>
-#include <type_traits>
-#include <cstdint>
-
-#include "libipc/def.h"
-
-#include "libipc/platform/detail.h"
-#include "libipc/circ/elem_def.h"
-#include "libipc/utility/log.h"
-#include "libipc/utility/utility.h"
-
-namespace ipc {
-
-////////////////////////////////////////////////////////////////
-/// producer-consumer implementation
-////////////////////////////////////////////////////////////////
-
-template <typename Flag>
-struct prod_cons_impl;
-
-template <>
-struct prod_cons_impl<wr<relat::single, relat::single, trans::unicast>> {
-
-    template <std::size_t DataSize, std::size_t AlignSize>
-    struct elem_t {
-        std::aligned_storage_t<DataSize, AlignSize> data_ {};
-    };
-
-    alignas(cache_line_size) std::atomic<circ::u2_t> rd_; // read index
-    alignas(cache_line_size) std::atomic<circ::u2_t> wt_; // write index
-
-    constexpr circ::u2_t cursor() const noexcept {
-        return 0;
-    }
-
-    template <typename W, typename F, typename E>
-    bool push(W* /*wrapper*/, F&& f, E* elems) {
-        auto cur_wt = circ::index_of(wt_.load(std::memory_order_relaxed));
-        if (cur_wt == circ::index_of(rd_.load(std::memory_order_acquire) - 1)) {
-            return false; // full
-        }
-        std::forward<F>(f)(&(elems[cur_wt].data_));
-        wt_.fetch_add(1, std::memory_order_release);
-        return true;
-    }
-
-    /**
-     * In single-single-unicast, 'force_push' means 'no reader' or 'the only one reader is dead'.
-     * So we could just disconnect all connections of receiver, and return false.
-    */
-    template <typename W, typename F, typename E>
-    bool force_push(W* wrapper, F&&, E*) {
-        wrapper->elems()->disconnect_receiver(~static_cast<circ::cc_t>(0u));
-        return false;
-    }
-
-    template <typename W, typename F, typename R, typename E>
-    bool pop(W* /*wrapper*/, circ::u2_t& /*cur*/, F&& f, R&& out, E* elems) {
-        auto cur_rd = circ::index_of(rd_.load(std::memory_order_relaxed));
-        if (cur_rd == circ::index_of(wt_.load(std::memory_order_acquire))) {
-            return false; // empty
-        }
-        std::forward<F>(f)(&(elems[cur_rd].data_));
-        std::forward<R>(out)(true);
-        rd_.fetch_add(1, std::memory_order_release);
-        return true;
-    }
-};
-
-template <>
-struct prod_cons_impl<wr<relat::single, relat::multi , trans::unicast>>
-     : prod_cons_impl<wr<relat::single, relat::single, trans::unicast>> {
-
-    template <typename W, typename F, typename E>
-    bool force_push(W* wrapper, F&&, E*) {
-        wrapper->elems()->disconnect_receiver(1);
-        return false;
-    }
-
-    template <typename W, typename F, typename R, 
-              template <std::size_t, std::size_t> class E, std::size_t DS, std::size_t AS>
-    bool pop(W* /*wrapper*/, circ::u2_t& /*cur*/, F&& f, R&& out, E<DS, AS>* elems) {
-        byte_t buff[DS];
-        for (unsigned k = 0;;) {
-            auto cur_rd = rd_.load(std::memory_order_relaxed);
-            if (circ::index_of(cur_rd) ==
-                circ::index_of(wt_.load(std::memory_order_acquire))) {
-                return false; // empty
-            }
-            std::memcpy(buff, &(elems[circ::index_of(cur_rd)].data_), sizeof(buff));
-            if (rd_.compare_exchange_weak(cur_rd, cur_rd + 1, std::memory_order_release)) {
-                std::forward<F>(f)(buff);
-                std::forward<R>(out)(true);
-                return true;
-            }
-            ipc::yield(k);
-        }
-    }
-};
-
-template <>
-struct prod_cons_impl<wr<relat::multi , relat::multi, trans::unicast>>
-     : prod_cons_impl<wr<relat::single, relat::multi, trans::unicast>> {
-
-    using flag_t = std::uint64_t;
-
-    template <std::size_t DataSize, std::size_t AlignSize>
-    struct elem_t {
-        std::aligned_storage_t<DataSize, AlignSize> data_ {};
-        std::atomic<flag_t> f_ct_ { 0 }; // commit flag
-    };
-
-    alignas(cache_line_size) std::atomic<circ::u2_t> ct_; // commit index
-
-    template <typename W, typename F, typename E>
-    bool push(W* /*wrapper*/, F&& f, E* elems) {
-        circ::u2_t cur_ct, nxt_ct;
-        for (unsigned k = 0;;) {
-            cur_ct = ct_.load(std::memory_order_relaxed);
-            if (circ::index_of(nxt_ct = cur_ct + 1) ==
-                circ::index_of(rd_.load(std::memory_order_acquire))) {
-                return false; // full
-            }
-            if (ct_.compare_exchange_weak(cur_ct, nxt_ct, std::memory_order_acq_rel)) {
-                break;
-            }
-            ipc::yield(k);
-        }
-        auto* el = elems + circ::index_of(cur_ct);
-        std::forward<F>(f)(&(el->data_));
-        // set flag & try update wt
-        el->f_ct_.store(~static_cast<flag_t>(cur_ct), std::memory_order_release);
-        while (1) {
-            auto cac_ct = el->f_ct_.load(std::memory_order_acquire);
-            if (cur_ct != wt_.load(std::memory_order_relaxed)) {
-                return true;
-            }
-            if ((~cac_ct) != cur_ct) {
-                return true;
-            }
-            if (!el->f_ct_.compare_exchange_strong(cac_ct, 0, std::memory_order_relaxed)) {
-                return true;
-            }
-            wt_.store(nxt_ct, std::memory_order_release);
-            cur_ct = nxt_ct;
-            nxt_ct = cur_ct + 1;
-            el = elems + circ::index_of(cur_ct);
-        }
-        return true;
-    }
-
-    template <typename W, typename F, typename E>
-    bool force_push(W* wrapper, F&&, E*) {
-        wrapper->elems()->disconnect_receiver(1);
-        return false;
-    }
-
-    template <typename W, typename F, typename R, 
-              template <std::size_t, std::size_t> class E, std::size_t DS, std::size_t AS>
-    bool pop(W* /*wrapper*/, circ::u2_t& /*cur*/, F&& f, R&& out, E<DS, AS>* elems) {
-        byte_t buff[DS];
-        for (unsigned k = 0;;) {
-            auto cur_rd = rd_.load(std::memory_order_relaxed);
-            auto cur_wt = wt_.load(std::memory_order_acquire);
-            auto id_rd  = circ::index_of(cur_rd);
-            auto id_wt  = circ::index_of(cur_wt);
-            if (id_rd == id_wt) {
-                auto* el = elems + id_wt;
-                auto cac_ct = el->f_ct_.load(std::memory_order_acquire);
-                if ((~cac_ct) != cur_wt) {
-                    return false; // empty
-                }
-                if (el->f_ct_.compare_exchange_weak(cac_ct, 0, std::memory_order_relaxed)) {
-                    wt_.store(cur_wt + 1, std::memory_order_release);
-                }
-                k = 0;
-            }
-            else {
-                std::memcpy(buff, &(elems[circ::index_of(cur_rd)].data_), sizeof(buff));
-                if (rd_.compare_exchange_weak(cur_rd, cur_rd + 1, std::memory_order_release)) {
-                    std::forward<F>(f)(buff);
-                    std::forward<R>(out)(true);
-                    return true;
-                }
-                ipc::yield(k);
-            }
-        }
-    }
-};
-
-template <>
-struct prod_cons_impl<wr<relat::single, relat::multi, trans::broadcast>> {
-
-    using rc_t = std::uint64_t;
-
-    enum : rc_t {
-        ep_mask = 0x00000000ffffffffull,
-        ep_incr = 0x0000000100000000ull
-    };
-
-    template <std::size_t DataSize, std::size_t AlignSize>
-    struct elem_t {
-        std::aligned_storage_t<DataSize, AlignSize> data_ {};
-        std::atomic<rc_t> rc_ { 0 }; // read-counter
-    };
-
-    alignas(cache_line_size) std::atomic<circ::u2_t> wt_;   // write index
-    alignas(cache_line_size) rc_t epoch_ { 0 };             // only one writer
-
-    circ::u2_t cursor() const noexcept {
-        return wt_.load(std::memory_order_acquire);
-    }
-
-    template <typename W, typename F, typename E>
-    bool push(W* wrapper, F&& f, E* elems) {
-        E* el;
-        for (unsigned k = 0;;) {
-            circ::cc_t cc = wrapper->elems()->connections(std::memory_order_relaxed);
-            if (cc == 0) return false; // no reader
-            el = elems + circ::index_of(wt_.load(std::memory_order_relaxed));
-            // check all consumers have finished reading this element
-            auto cur_rc = el->rc_.load(std::memory_order_acquire);
-            circ::cc_t rem_cc = cur_rc & ep_mask;
-            if ((cc & rem_cc) && ((cur_rc & ~ep_mask) == epoch_)) {
-                return false; // has not finished yet
-            }
-            // consider rem_cc to be 0 here
-            if (el->rc_.compare_exchange_weak(
-                        cur_rc, epoch_ | static_cast<rc_t>(cc), std::memory_order_release)) {
-                break;
-            }
-            ipc::yield(k);
-        }
-        std::forward<F>(f)(&(el->data_));
-        wt_.fetch_add(1, std::memory_order_release);
-        return true;
-    }
-
-    template <typename W, typename F, typename E>
-    bool force_push(W* wrapper, F&& f, E* elems) {
-        E* el;
-        epoch_ += ep_incr;
-        for (unsigned k = 0;;) {
-            circ::cc_t cc = wrapper->elems()->connections(std::memory_order_relaxed);
-            if (cc == 0) return false; // no reader
-            el = elems + circ::index_of(wt_.load(std::memory_order_relaxed));
-            // check all consumers have finished reading this element
-            auto cur_rc = el->rc_.load(std::memory_order_acquire);
-            circ::cc_t rem_cc = cur_rc & ep_mask;
-            if (cc & rem_cc) {
-                ipc::log("force_push: k = %u, cc = %u, rem_cc = %u\n", k, cc, rem_cc);
-                cc = wrapper->elems()->disconnect_receiver(rem_cc); // disconnect all invalid readers
-                if (cc == 0) return false; // no reader
-            }
-            // just compare & exchange
-            if (el->rc_.compare_exchange_weak(
-                        cur_rc, epoch_ | static_cast<rc_t>(cc), std::memory_order_release)) {
-                break;
-            }
-            ipc::yield(k);
-        }
-        std::forward<F>(f)(&(el->data_));
-        wt_.fetch_add(1, std::memory_order_release);
-        return true;
-    }
-
-    template <typename W, typename F, typename R, typename E>
-    bool pop(W* wrapper, circ::u2_t& cur, F&& f, R&& out, E* elems) {
-        if (cur == cursor()) return false; // acquire
-        auto* el = elems + circ::index_of(cur++);
-        std::forward<F>(f)(&(el->data_));
-        for (unsigned k = 0;;) {
-            auto cur_rc = el->rc_.load(std::memory_order_acquire);
-            if ((cur_rc & ep_mask) == 0) {
-                std::forward<R>(out)(true);
-                return true;
-            }
-            auto nxt_rc = cur_rc & ~static_cast<rc_t>(wrapper->connected_id());
-            if (el->rc_.compare_exchange_weak(cur_rc, nxt_rc, std::memory_order_release)) {
-                std::forward<R>(out)((nxt_rc & ep_mask) == 0);
-                return true;
-            }
-            ipc::yield(k);
-        }
-    }
-};
-
-template <>
-struct prod_cons_impl<wr<relat::multi, relat::multi, trans::broadcast>> {
-
-    using rc_t   = std::uint64_t;
-    using flag_t = std::uint64_t;
-
-    enum : rc_t {
-        rc_mask = 0x00000000ffffffffull,
-        ep_mask = 0x00ffffffffffffffull,
-        ep_incr = 0x0100000000000000ull,
-        ic_mask = 0xff000000ffffffffull,
-        ic_incr = 0x0000000100000000ull
-    };
-
-    template <std::size_t DataSize, std::size_t AlignSize>
-    struct elem_t {
-        std::aligned_storage_t<DataSize, AlignSize> data_ {};
-        std::atomic<rc_t  > rc_   { 0 }; // read-counter
-        std::atomic<flag_t> f_ct_ { 0 }; // commit flag
-    };
-
-    alignas(cache_line_size) std::atomic<circ::u2_t> ct_;   // commit index
-    alignas(cache_line_size) std::atomic<rc_t> epoch_ { 0 };
-
-    circ::u2_t cursor() const noexcept {
-        return ct_.load(std::memory_order_acquire);
-    }
-
-    constexpr static rc_t inc_rc(rc_t rc) noexcept {
-        return (rc & ic_mask) | ((rc + ic_incr) & ~ic_mask);
-    }
-
-    constexpr static rc_t inc_mask(rc_t rc) noexcept {
-        return inc_rc(rc) & ~rc_mask;
-    }
-
-    template <typename W, typename F, typename E>
-    bool push(W* wrapper, F&& f, E* elems) {
-        E* el;
-        circ::u2_t cur_ct;
-        rc_t epoch = epoch_.load(std::memory_order_acquire);
-        for (unsigned k = 0;;) {
-            circ::cc_t cc = wrapper->elems()->connections(std::memory_order_relaxed);
-            if (cc == 0) return false; // no reader
-            el = elems + circ::index_of(cur_ct = ct_.load(std::memory_order_relaxed));
-            // check all consumers have finished reading this element
-            auto cur_rc = el->rc_.load(std::memory_order_relaxed);
-            circ::cc_t rem_cc = cur_rc & rc_mask;
-            if ((cc & rem_cc) && ((cur_rc & ~ep_mask) == epoch)) {
-                return false; // has not finished yet
-            }
-            else if (!rem_cc) {
-                auto cur_fl = el->f_ct_.load(std::memory_order_acquire);
-                if ((cur_fl != cur_ct) && cur_fl) {
-                    return false; // full
-                }
-            }
-            // consider rem_cc to be 0 here
-            if (el->rc_.compare_exchange_weak(
-                        cur_rc, inc_mask(epoch | (cur_rc & ep_mask)) | static_cast<rc_t>(cc), std::memory_order_relaxed) &&
-                epoch_.compare_exchange_weak(epoch, epoch, std::memory_order_acq_rel)) {
-                break;
-            }
-            ipc::yield(k);
-        }
-        // only one thread/process would touch here at one time
-        ct_.store(cur_ct + 1, std::memory_order_release);
-        std::forward<F>(f)(&(el->data_));
-        // set flag & try update wt
-        el->f_ct_.store(~static_cast<flag_t>(cur_ct), std::memory_order_release);
-        return true;
-    }
-
-    template <typename W, typename F, typename E>
-    bool force_push(W* wrapper, F&& f, E* elems) {
-        E* el;
-        circ::u2_t cur_ct;
-        rc_t epoch = epoch_.fetch_add(ep_incr, std::memory_order_release) + ep_incr;
-        for (unsigned k = 0;;) {
-            circ::cc_t cc = wrapper->elems()->connections(std::memory_order_relaxed);
-            if (cc == 0) return false; // no reader
-            el = elems + circ::index_of(cur_ct = ct_.load(std::memory_order_relaxed));
-            // check all consumers have finished reading this element
-            auto cur_rc = el->rc_.load(std::memory_order_acquire);
-            circ::cc_t rem_cc = cur_rc & rc_mask;
-            if (cc & rem_cc) {
-                ipc::log("force_push: k = %u, cc = %u, rem_cc = %u\n", k, cc, rem_cc);
-                cc = wrapper->elems()->disconnect_receiver(rem_cc); // disconnect all invalid readers
-                if (cc == 0) return false; // no reader
-            }
-            // just compare & exchange
-            if (el->rc_.compare_exchange_weak(
-                        cur_rc, inc_mask(epoch | (cur_rc & ep_mask)) | static_cast<rc_t>(cc), std::memory_order_relaxed)) {
-                if (epoch == epoch_.load(std::memory_order_acquire)) {
-                    break;
-                }
-                else if (push(wrapper, std::forward<F>(f), elems)) {
-                    return true;
-                }
-                epoch = epoch_.fetch_add(ep_incr, std::memory_order_release) + ep_incr;
-            }
-            ipc::yield(k);
-        }
-        // only one thread/process would touch here at one time
-        ct_.store(cur_ct + 1, std::memory_order_release);
-        std::forward<F>(f)(&(el->data_));
-        // set flag & try update wt
-        el->f_ct_.store(~static_cast<flag_t>(cur_ct), std::memory_order_release);
-        return true;
-    }
-
-    template <typename W, typename F, typename R, typename E, std::size_t N>
-    bool pop(W* wrapper, circ::u2_t& cur, F&& f, R&& out, E(& elems)[N]) {
-        auto* el = elems + circ::index_of(cur);
-        auto cur_fl = el->f_ct_.load(std::memory_order_acquire);
-        if (cur_fl != ~static_cast<flag_t>(cur)) {
-            return false; // empty
-        }
-        ++cur;
-        std::forward<F>(f)(&(el->data_));
-        for (unsigned k = 0;;) {
-            auto cur_rc = el->rc_.load(std::memory_order_acquire);
-            if ((cur_rc & rc_mask) == 0) {
-                std::forward<R>(out)(true);
-                el->f_ct_.store(cur + N - 1, std::memory_order_release);
-                return true;
-            }
-            auto nxt_rc = inc_rc(cur_rc) & ~static_cast<rc_t>(wrapper->connected_id());
-            bool last_one = false;
-            if ((last_one = (nxt_rc & rc_mask) == 0)) {
-                el->f_ct_.store(cur + N - 1, std::memory_order_release);
-            }
-            if (el->rc_.compare_exchange_weak(cur_rc, nxt_rc, std::memory_order_release)) {
-                std::forward<R>(out)(last_one);
-                return true;
-            }
-            ipc::yield(k);
-        }
-    }
-};
-
-} // namespace ipc

crazy_functions/test_project/cpp/cppipc/queue.h

@@ -1,216 +0,0 @@
-#pragma once
-
-#include <type_traits>
-#include <new>
-#include <utility>  // [[since C++14]]: std::exchange
-#include <algorithm>
-#include <atomic>
-#include <tuple>
-#include <thread>
-#include <chrono>
-#include <string>
-#include <cassert>  // assert
-
-#include "libipc/def.h"
-#include "libipc/shm.h"
-#include "libipc/rw_lock.h"
-
-#include "libipc/utility/log.h"
-#include "libipc/platform/detail.h"
-#include "libipc/circ/elem_def.h"
-
-namespace ipc {
-namespace detail {
-
-class queue_conn {
-protected:
-    circ::cc_t connected_ = 0;
-    shm::handle elems_h_;
-
-    template <typename Elems>
-    Elems* open(char const * name) {
-        if (name == nullptr || name[0] == '\0') {
-            ipc::error("fail open waiter: name is empty!\n");
-            return nullptr;
-        }
-        if (!elems_h_.acquire(name, sizeof(Elems))) {
-            return nullptr;
-        }
-        auto elems = static_cast<Elems*>(elems_h_.get());
-        if (elems == nullptr) {
-            ipc::error("fail acquire elems: %s\n", name);
-            return nullptr;
-        }
-        elems->init();
-        return elems;
-    }
-
-    void close() {
-        elems_h_.release();
-    }
-
-public:
-    queue_conn() = default;
-    queue_conn(const queue_conn&) = delete;
-    queue_conn& operator=(const queue_conn&) = delete;
-
-    bool connected() const noexcept {
-        return connected_ != 0;
-    }
-
-    circ::cc_t connected_id() const noexcept {
-        return connected_;
-    }
-
-    template <typename Elems>
-    auto connect(Elems* elems) noexcept
-                         /*needs 'optional' here*/
-     -> std::tuple<bool, bool, decltype(std::declval<Elems>().cursor())> {
-        if (elems == nullptr) return {};
-        // if it's already connected, just return
-        if (connected()) return {connected(), false, 0};
-        connected_ = elems->connect_receiver();
-        return {connected(), true, elems->cursor()};
-    }
-
-    template <typename Elems>
-    bool disconnect(Elems* elems) noexcept {
-        if (elems == nullptr) return false;
-        // if it's already disconnected, just return false
-        if (!connected()) return false;
-        elems->disconnect_receiver(std::exchange(connected_, 0));
-        return true;
-    }
-};
-
-template <typename Elems>
-class queue_base : public queue_conn {
-    using base_t = queue_conn;
-
-public:
-    using elems_t  = Elems;
-    using policy_t = typename elems_t::policy_t;
-
-protected:
-    elems_t * elems_ = nullptr;
-    decltype(std::declval<elems_t>().cursor()) cursor_ = 0;
-    bool sender_flag_ = false;
-
-public:
-    using base_t::base_t;
-
-    queue_base() = default;
-
-    explicit queue_base(char const * name)
-        : queue_base{} {
-        elems_ = open<elems_t>(name);
-    }
-
-    explicit queue_base(elems_t * elems) noexcept
-        : queue_base{} {
-        assert(elems != nullptr);
-        elems_ = elems;
-    }
-
-    /* not virtual */ ~queue_base() {
-        base_t::close();
-    }
-
-    elems_t       * elems()       noexcept { return elems_; }
-    elems_t const * elems() const noexcept { return elems_; }
-
-    bool ready_sending() noexcept {
-        if (elems_ == nullptr) return false;
-        return sender_flag_ || (sender_flag_ = elems_->connect_sender());
-    }
-
-    void shut_sending() noexcept {
-        if (elems_ == nullptr) return;
-        if (!sender_flag_) return;
-        elems_->disconnect_sender();
-    }
-
-    bool connect() noexcept {
-        auto tp = base_t::connect(elems_);
-        if (std::get<0>(tp) && std::get<1>(tp)) {
-            cursor_ = std::get<2>(tp);
-            return true;
-        }
-        return std::get<0>(tp);
-    }
-
-    bool disconnect() noexcept {
-        return base_t::disconnect(elems_);
-    }
-
-    std::size_t conn_count() const noexcept {
-        return (elems_ == nullptr) ? static_cast<std::size_t>(invalid_value) : elems_->conn_count();
-    }
-
-    bool valid() const noexcept {
-        return elems_ != nullptr;
-    }
-
-    bool empty() const noexcept {
-        return !valid() || (cursor_ == elems_->cursor());
-    }
-
-    template <typename T, typename F, typename... P>
-    bool push(F&& prep, P&&... params) {
-        if (elems_ == nullptr) return false;
-        return elems_->push(this, [&](void* p) {
-            if (prep(p)) ::new (p) T(std::forward<P>(params)...);
-        });
-    }
-
-    template <typename T, typename F, typename... P>
-    bool force_push(F&& prep, P&&... params) {
-        if (elems_ == nullptr) return false;
-        return elems_->force_push(this, [&](void* p) {
-            if (prep(p)) ::new (p) T(std::forward<P>(params)...);
-        });
-    }
-
-    template <typename T, typename F>
-    bool pop(T& item, F&& out) {
-        if (elems_ == nullptr) {
-            return false;
-        }
-        return elems_->pop(this, &(this->cursor_), [&item](void* p) {
-            ::new (&item) T(std::move(*static_cast<T*>(p)));
-        }, std::forward<F>(out));
-    }
-};
-
-} // namespace detail
-
-template <typename T, typename Policy>
-class queue final : public detail::queue_base<typename Policy::template elems_t<sizeof(T), alignof(T)>> {
-    using base_t = detail::queue_base<typename Policy::template elems_t<sizeof(T), alignof(T)>>;
-
-public:
-    using value_t = T;
-
-    using base_t::base_t;
-
-    template <typename... P>
-    bool push(P&&... params) {
-        return base_t::template push<T>(std::forward<P>(params)...);
-    }
-
-    template <typename... P>
-    bool force_push(P&&... params) {
-        return base_t::template force_push<T>(std::forward<P>(params)...);
-    }
-
-    bool pop(T& item) {
-        return base_t::pop(item, [](bool) {});
-    }
-
-    template <typename F>
-    bool pop(T& item, F&& out) {
-        return base_t::pop(item, std::forward<F>(out));
-    }
-};
-
-} // namespace ipc

crazy_functions/test_project/cpp/cppipc/shm.cpp

@@ -1,103 +0,0 @@
-
-#include <string>
-#include <utility>
-
-#include "libipc/shm.h"
-
-#include "libipc/utility/pimpl.h"
-#include "libipc/memory/resource.h"
-
-namespace ipc {
-namespace shm {
-
-class handle::handle_ : public pimpl<handle_> {
-public:
-    shm::id_t id_ = nullptr;
-    void*     m_  = nullptr;
-
-    ipc::string n_;
-    std::size_t s_ = 0;
-};
-
-handle::handle()
-    : p_(p_->make()) {
-}
-
-handle::handle(char const * name, std::size_t size, unsigned mode)
-    : handle() {
-    acquire(name, size, mode);
-}
-
-handle::handle(handle&& rhs)
-    : handle() {
-    swap(rhs);
-}
-
-handle::~handle() {
-    release();
-    p_->clear();
-}
-
-void handle::swap(handle& rhs) {
-    std::swap(p_, rhs.p_);
-}
-
-handle& handle::operator=(handle rhs) {
-    swap(rhs);
-    return *this;
-}
-
-bool handle::valid() const noexcept {
-    return impl(p_)->m_ != nullptr;
-}
-
-std::size_t handle::size() const noexcept {
-    return impl(p_)->s_;
-}
-
-char const * handle::name() const noexcept {
-    return impl(p_)->n_.c_str();
-}
-
-std::int32_t handle::ref() const noexcept {
-    return shm::get_ref(impl(p_)->id_);
-}
-
-void handle::sub_ref() noexcept {
-    shm::sub_ref(impl(p_)->id_);
-}
-
-bool handle::acquire(char const * name, std::size_t size, unsigned mode) {
-    release();
-    impl(p_)->id_ = shm::acquire((impl(p_)->n_ = name).c_str(), size, mode);
-    impl(p_)->m_  = shm::get_mem(impl(p_)->id_, &(impl(p_)->s_));
-    return valid();
-}
-
-std::int32_t handle::release() {
-    if (impl(p_)->id_ == nullptr) return -1;
-    return shm::release(detach());
-}
-
-void* handle::get() const {
-    return impl(p_)->m_;
-}
-
-void handle::attach(id_t id) {
-    if (id == nullptr) return;
-    release();
-    impl(p_)->id_ = id;
-    impl(p_)->m_  = shm::get_mem(impl(p_)->id_, &(impl(p_)->s_));
-}
-
-id_t handle::detach() {
-    auto old = impl(p_)->id_;
-    impl(p_)->id_ = nullptr;
-    impl(p_)->m_  = nullptr;
-    impl(p_)->s_  = 0;
-    impl(p_)->n_.clear();
-    return old;
-}
-
-} // namespace shm
-} // namespace ipc

crazy_functions/test_project/cpp/cppipc/waiter.h

@@ -1,83 +0,0 @@
-#pragma once
-
-#include <utility>
-#include <string>
-#include <mutex>
-#include <atomic>
-
-#include "libipc/def.h"
-#include "libipc/mutex.h"
-#include "libipc/condition.h"
-#include "libipc/platform/detail.h"
-
-namespace ipc {
-namespace detail {
-
-class waiter {
-    ipc::sync::condition cond_;
-    ipc::sync::mutex     lock_;
-    std::atomic<bool>    quit_ {false};
-
-public:
-    static void init();
-
-    waiter() = default;
-    waiter(char const *name) {
-        open(name);
-    }
-
-    ~waiter() {
-        close();
-    }
-
-    bool valid() const noexcept {
-        return cond_.valid() && lock_.valid();
-    }
-
-    bool open(char const *name) noexcept {
-        quit_.store(false, std::memory_order_relaxed);
-        if (!cond_.open((std::string{"_waiter_cond_"} + name).c_str())) {
-            return false;
-        }
-        if (!lock_.open((std::string{"_waiter_lock_"} + name).c_str())) {
-            cond_.close();
-            return false;
-        }
-        return valid();
-    }
-
-    void close() noexcept {
-        cond_.close();
-        lock_.close();
-    }
-
-    template <typename F>
-    bool wait_if(F &&pred, std::uint64_t tm = ipc::invalid_value) noexcept {
-        IPC_UNUSED_ std::lock_guard<ipc::sync::mutex> guard {lock_};
-        while ([this, &pred] {
-                    return !quit_.load(std::memory_order_relaxed)
-                        && std::forward<F>(pred)();
-                }()) {
-            if (!cond_.wait(lock_, tm)) return false;
-        }
-        return true;
-    }
-
-    bool notify() noexcept {
-        std::lock_guard<ipc::sync::mutex>{lock_}; // barrier
-        return cond_.notify(lock_);
-    }
-
-    bool broadcast() noexcept {
-        std::lock_guard<ipc::sync::mutex>{lock_}; // barrier
-        return cond_.broadcast(lock_);
-    }
-
-    bool quit_waiting() {
-        quit_.store(true, std::memory_order_release);
-        return broadcast();
-    }
-};
-
-} // namespace detail
-} // namespace ipc

crazy_functions/test_project/cpp/cppipc/来源

@@ -1,3 +0,0 @@
-https://github.com/mutouyun/cpp-ipc
-
-A high-performance inter-process communication library using shared memory on Linux/Windows.

crazy_functions/test_project/cpp/libJPG/jpgd.h

@@ -1,316 +0,0 @@
-// jpgd.h - C++ class for JPEG decompression.
-// Public domain, Rich Geldreich <richgel99@gmail.com>
-#ifndef JPEG_DECODER_H
-#define JPEG_DECODER_H
-
-#include <stdlib.h>
-#include <stdio.h>
-#include <setjmp.h>
-
-namespace jpgd
-{
-  typedef unsigned char  uint8;
-  typedef   signed short int16;
-  typedef unsigned short uint16;
-  typedef unsigned int   uint;
-  typedef   signed int   int32;
-
-  // Loads a JPEG image from a memory buffer or a file.
-  // req_comps can be 1 (grayscale), 3 (RGB), or 4 (RGBA).
-  // On return, width/height will be set to the image's dimensions, and actual_comps will be set to the either 1 (grayscale) or 3 (RGB).
-  // Notes: For more control over where and how the source data is read, see the decompress_jpeg_image_from_stream() function below, or call the jpeg_decoder class directly.
-  // Requesting a 8 or 32bpp image is currently a little faster than 24bpp because the jpeg_decoder class itself currently always unpacks to either 8 or 32bpp.
-// BEGIN EPIC MOD
-//unsigned char *decompress_jpeg_image_from_memory(const unsigned char *pSrc_data, int src_data_size, int *width, int *height, int *actual_comps, int req_comps);
-  unsigned char *decompress_jpeg_image_from_memory(const unsigned char *pSrc_data, int src_data_size, int *width, int *height, int *actual_comps, int req_comps, int format);
-// END EPIC MOD
-  unsigned char *decompress_jpeg_image_from_file(const char *pSrc_filename, int *width, int *height, int *actual_comps, int req_comps);
-
-  // Success/failure error codes.
-  enum jpgd_status
-  {
-    JPGD_SUCCESS = 0, JPGD_FAILED = -1, JPGD_DONE = 1,
-    JPGD_BAD_DHT_COUNTS = -256, JPGD_BAD_DHT_INDEX, JPGD_BAD_DHT_MARKER, JPGD_BAD_DQT_MARKER, JPGD_BAD_DQT_TABLE, 
-    JPGD_BAD_PRECISION, JPGD_BAD_HEIGHT, JPGD_BAD_WIDTH, JPGD_TOO_MANY_COMPONENTS, 
-    JPGD_BAD_SOF_LENGTH, JPGD_BAD_VARIABLE_MARKER, JPGD_BAD_DRI_LENGTH, JPGD_BAD_SOS_LENGTH,
-    JPGD_BAD_SOS_COMP_ID, JPGD_W_EXTRA_BYTES_BEFORE_MARKER, JPGD_NO_ARITHMITIC_SUPPORT, JPGD_UNEXPECTED_MARKER,
-    JPGD_NOT_JPEG, JPGD_UNSUPPORTED_MARKER, JPGD_BAD_DQT_LENGTH, JPGD_TOO_MANY_BLOCKS,
-    JPGD_UNDEFINED_QUANT_TABLE, JPGD_UNDEFINED_HUFF_TABLE, JPGD_NOT_SINGLE_SCAN, JPGD_UNSUPPORTED_COLORSPACE,
-    JPGD_UNSUPPORTED_SAMP_FACTORS, JPGD_DECODE_ERROR, JPGD_BAD_RESTART_MARKER, JPGD_ASSERTION_ERROR,
-    JPGD_BAD_SOS_SPECTRAL, JPGD_BAD_SOS_SUCCESSIVE, JPGD_STREAM_READ, JPGD_NOTENOUGHMEM
-  };
-    
-  // Input stream interface.
-  // Derive from this class to read input data from sources other than files or memory. Set m_eof_flag to true when no more data is available.
-  // The decoder is rather greedy: it will keep on calling this method until its internal input buffer is full, or until the EOF flag is set.
-  // It the input stream contains data after the JPEG stream's EOI (end of image) marker it will probably be pulled into the internal buffer.
-  // Call the get_total_bytes_read() method to determine the actual size of the JPEG stream after successful decoding.
-  class jpeg_decoder_stream
-  {
-  public:
-    jpeg_decoder_stream() { }
-    virtual ~jpeg_decoder_stream() { }
-
-    // The read() method is called when the internal input buffer is empty.
-    // Parameters:
-    // pBuf - input buffer
-    // max_bytes_to_read - maximum bytes that can be written to pBuf
-    // pEOF_flag - set this to true if at end of stream (no more bytes remaining)
-    // Returns -1 on error, otherwise return the number of bytes actually written to the buffer (which may be 0).
-    // Notes: This method will be called in a loop until you set *pEOF_flag to true or the internal buffer is full.
-    virtual int read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag) = 0;
-  };
-
-  // stdio FILE stream class.
-  class jpeg_decoder_file_stream : public jpeg_decoder_stream
-  {
-    jpeg_decoder_file_stream(const jpeg_decoder_file_stream &);
-    jpeg_decoder_file_stream &operator =(const jpeg_decoder_file_stream &);
-
-    FILE *m_pFile;
-    bool m_eof_flag, m_error_flag;
-
-  public:
-    jpeg_decoder_file_stream();
-    virtual ~jpeg_decoder_file_stream();
-    
-    bool open(const char *Pfilename);
-    void close();
-
-    virtual int read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag);
-  };
-
-  // Memory stream class.
-  class jpeg_decoder_mem_stream : public jpeg_decoder_stream
-  {
-    const uint8 *m_pSrc_data;
-    uint m_ofs, m_size;
-
-  public:
-    jpeg_decoder_mem_stream() : m_pSrc_data(NULL), m_ofs(0), m_size(0) { }
-    jpeg_decoder_mem_stream(const uint8 *pSrc_data, uint size) : m_pSrc_data(pSrc_data), m_ofs(0), m_size(size) { }
-
-    virtual ~jpeg_decoder_mem_stream() { }
-
-    bool open(const uint8 *pSrc_data, uint size);
-    void close() { m_pSrc_data = NULL; m_ofs = 0; m_size = 0; }
-    
-    virtual int read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag);
-  };
-
-  // Loads JPEG file from a jpeg_decoder_stream.
-  unsigned char *decompress_jpeg_image_from_stream(jpeg_decoder_stream *pStream, int *width, int *height, int *actual_comps, int req_comps);
-
-  enum 
-  { 
-    JPGD_IN_BUF_SIZE = 8192, JPGD_MAX_BLOCKS_PER_MCU = 10, JPGD_MAX_HUFF_TABLES = 8, JPGD_MAX_QUANT_TABLES = 4, 
-    JPGD_MAX_COMPONENTS = 4, JPGD_MAX_COMPS_IN_SCAN = 4, JPGD_MAX_BLOCKS_PER_ROW = 8192, JPGD_MAX_HEIGHT = 16384, JPGD_MAX_WIDTH = 16384 
-  };
-          
-  typedef int16 jpgd_quant_t;
-  typedef int16 jpgd_block_t;
-
-  class jpeg_decoder
-  {
-  public:
-    // Call get_error_code() after constructing to determine if the stream is valid or not. You may call the get_width(), get_height(), etc.
-    // methods after the constructor is called. You may then either destruct the object, or begin decoding the image by calling begin_decoding(), then decode() on each scanline.
-    jpeg_decoder(jpeg_decoder_stream *pStream);
-
-    ~jpeg_decoder();
-
-    // Call this method after constructing the object to begin decompression.
-    // If JPGD_SUCCESS is returned you may then call decode() on each scanline.
-    int begin_decoding();
-
-    // Returns the next scan line.
-    // For grayscale images, pScan_line will point to a buffer containing 8-bit pixels (get_bytes_per_pixel() will return 1). 
-    // Otherwise, it will always point to a buffer containing 32-bit RGBA pixels (A will always be 255, and get_bytes_per_pixel() will return 4).
-    // Returns JPGD_SUCCESS if a scan line has been returned.
-    // Returns JPGD_DONE if all scan lines have been returned.
-    // Returns JPGD_FAILED if an error occurred. Call get_error_code() for a more info.
-    int decode(const void** pScan_line, uint* pScan_line_len);
-    
-    inline jpgd_status get_error_code() const { return m_error_code; }
-
-    inline int get_width() const { return m_image_x_size; }
-    inline int get_height() const { return m_image_y_size; }
-
-    inline int get_num_components() const { return m_comps_in_frame; }
-
-    inline int get_bytes_per_pixel() const { return m_dest_bytes_per_pixel; }
-    inline int get_bytes_per_scan_line() const { return m_image_x_size * get_bytes_per_pixel(); }
-
-    // Returns the total number of bytes actually consumed by the decoder (which should equal the actual size of the JPEG file).
-    inline int get_total_bytes_read() const { return m_total_bytes_read; }
-    
-  private:
-    jpeg_decoder(const jpeg_decoder &);
-    jpeg_decoder &operator =(const jpeg_decoder &);
-
-    typedef void (*pDecode_block_func)(jpeg_decoder *, int, int, int);
-
-    struct huff_tables
-    {
-      bool ac_table;
-      uint  look_up[256];
-      uint  look_up2[256];
-      uint8 code_size[256];
-      uint  tree[512];
-    };
-
-    struct coeff_buf
-    {
-      uint8 *pData;
-      int block_num_x, block_num_y;
-      int block_len_x, block_len_y;
-      int block_size;
-    };
-
-    struct mem_block
-    {
-      mem_block *m_pNext;
-      size_t m_used_count;
-      size_t m_size;
-      char m_data[1];
-    };
-
-    jmp_buf m_jmp_state;
-    mem_block *m_pMem_blocks;
-    int m_image_x_size;
-    int m_image_y_size;
-    jpeg_decoder_stream *m_pStream;
-    int m_progressive_flag;
-    uint8 m_huff_ac[JPGD_MAX_HUFF_TABLES];
-    uint8* m_huff_num[JPGD_MAX_HUFF_TABLES];      // pointer to number of Huffman codes per bit size
-    uint8* m_huff_val[JPGD_MAX_HUFF_TABLES];      // pointer to Huffman codes per bit size
-    jpgd_quant_t* m_quant[JPGD_MAX_QUANT_TABLES]; // pointer to quantization tables
-    int m_scan_type;                              // Gray, Yh1v1, Yh1v2, Yh2v1, Yh2v2 (CMYK111, CMYK4114 no longer supported)
-    int m_comps_in_frame;                         // # of components in frame
-    int m_comp_h_samp[JPGD_MAX_COMPONENTS];       // component's horizontal sampling factor
-    int m_comp_v_samp[JPGD_MAX_COMPONENTS];       // component's vertical sampling factor
-    int m_comp_quant[JPGD_MAX_COMPONENTS];        // component's quantization table selector
-    int m_comp_ident[JPGD_MAX_COMPONENTS];        // component's ID
-    int m_comp_h_blocks[JPGD_MAX_COMPONENTS];
-    int m_comp_v_blocks[JPGD_MAX_COMPONENTS];
-    int m_comps_in_scan;                          // # of components in scan
-    int m_comp_list[JPGD_MAX_COMPS_IN_SCAN];      // components in this scan
-    int m_comp_dc_tab[JPGD_MAX_COMPONENTS];       // component's DC Huffman coding table selector
-    int m_comp_ac_tab[JPGD_MAX_COMPONENTS];       // component's AC Huffman coding table selector
-    int m_spectral_start;                         // spectral selection start
-    int m_spectral_end;                           // spectral selection end
-    int m_successive_low;                         // successive approximation low
-    int m_successive_high;                        // successive approximation high
-    int m_max_mcu_x_size;                         // MCU's max. X size in pixels
-    int m_max_mcu_y_size;                         // MCU's max. Y size in pixels
-    int m_blocks_per_mcu;
-    int m_max_blocks_per_row;
-    int m_mcus_per_row, m_mcus_per_col;
-    int m_mcu_org[JPGD_MAX_BLOCKS_PER_MCU];
-    int m_total_lines_left;                       // total # lines left in image
-    int m_mcu_lines_left;                         // total # lines left in this MCU
-    int m_real_dest_bytes_per_scan_line;
-    int m_dest_bytes_per_scan_line;               // rounded up
-    int m_dest_bytes_per_pixel;                   // 4 (RGB) or 1 (Y)
-    huff_tables* m_pHuff_tabs[JPGD_MAX_HUFF_TABLES];
-    coeff_buf* m_dc_coeffs[JPGD_MAX_COMPONENTS];
-    coeff_buf* m_ac_coeffs[JPGD_MAX_COMPONENTS];
-    int m_eob_run;
-    int m_block_y_mcu[JPGD_MAX_COMPONENTS];
-    uint8* m_pIn_buf_ofs;
-    int m_in_buf_left;
-    int m_tem_flag;
-    bool m_eof_flag;
-    uint8 m_in_buf_pad_start[128];
-    uint8 m_in_buf[JPGD_IN_BUF_SIZE + 128];
-    uint8 m_in_buf_pad_end[128];
-    int m_bits_left;
-    uint m_bit_buf;
-    int m_restart_interval;
-    int m_restarts_left;
-    int m_next_restart_num;
-    int m_max_mcus_per_row;
-    int m_max_blocks_per_mcu;
-    int m_expanded_blocks_per_mcu;
-    int m_expanded_blocks_per_row;
-    int m_expanded_blocks_per_component;
-    bool  m_freq_domain_chroma_upsample;
-    int m_max_mcus_per_col;
-    uint m_last_dc_val[JPGD_MAX_COMPONENTS];
-    jpgd_block_t* m_pMCU_coefficients;
-    int m_mcu_block_max_zag[JPGD_MAX_BLOCKS_PER_MCU];
-    uint8* m_pSample_buf;
-    int m_crr[256];
-    int m_cbb[256];
-    int m_crg[256];
-    int m_cbg[256];
-    uint8* m_pScan_line_0;
-    uint8* m_pScan_line_1;
-    jpgd_status m_error_code;
-    bool m_ready_flag;
-    int m_total_bytes_read;
-
-    void free_all_blocks();
-    // BEGIN EPIC MOD
-    UE_NORETURN void stop_decoding(jpgd_status status);
-    // END EPIC MOD
-    void *alloc(size_t n, bool zero = false);
-    void word_clear(void *p, uint16 c, uint n);
-    void prep_in_buffer();
-    void read_dht_marker();
-    void read_dqt_marker();
-    void read_sof_marker();
-    void skip_variable_marker();
-    void read_dri_marker();
-    void read_sos_marker();
-    int next_marker();
-    int process_markers();
-    void locate_soi_marker();
-    void locate_sof_marker();
-    int locate_sos_marker();
-    void init(jpeg_decoder_stream * pStream);
-    void create_look_ups();
-    void fix_in_buffer();
-    void transform_mcu(int mcu_row);
-    void transform_mcu_expand(int mcu_row);
-    coeff_buf* coeff_buf_open(int block_num_x, int block_num_y, int block_len_x, int block_len_y);
-    inline jpgd_block_t *coeff_buf_getp(coeff_buf *cb, int block_x, int block_y);
-    void load_next_row();
-    void decode_next_row();
-    void make_huff_table(int index, huff_tables *pH);
-    void check_quant_tables();
-    void check_huff_tables();
-    void calc_mcu_block_order();
-    int init_scan();
-    void init_frame();
-    void process_restart();
-    void decode_scan(pDecode_block_func decode_block_func);
-    void init_progressive();
-    void init_sequential();
-    void decode_start();
-    void decode_init(jpeg_decoder_stream * pStream);
-    void H2V2Convert();
-    void H2V1Convert();
-    void H1V2Convert();
-    void H1V1Convert();
-    void gray_convert();
-    void expanded_convert();
-    void find_eoi();
-    inline uint get_char();
-    inline uint get_char(bool *pPadding_flag);
-    inline void stuff_char(uint8 q);
-    inline uint8 get_octet();
-    inline uint get_bits(int num_bits);
-    inline uint get_bits_no_markers(int numbits);
-    inline int huff_decode(huff_tables *pH);
-    inline int huff_decode(huff_tables *pH, int& extrabits);
-    static inline uint8 clamp(int i);
-    static void decode_block_dc_first(jpeg_decoder *pD, int component_id, int block_x, int block_y);
-    static void decode_block_dc_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y);
-    static void decode_block_ac_first(jpeg_decoder *pD, int component_id, int block_x, int block_y);
-    static void decode_block_ac_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y);
-  };
-  
-} // namespace jpgd
-
-#endif // JPEG_DECODER_H

crazy_functions/test_project/cpp/libJPG/jpge.h

@@ -1,172 +0,0 @@
-
-// jpge.h - C++ class for JPEG compression.
-// Public domain, Rich Geldreich <richgel99@gmail.com>
-// Alex Evans: Added RGBA support, linear memory allocator.
-#ifndef JPEG_ENCODER_H
-#define JPEG_ENCODER_H
-
-#include <stdint.h>
-
-namespace jpge
-{
-  typedef unsigned char  uint8;
-  typedef signed short   int16;
-  typedef signed int     int32;
-  typedef unsigned short uint16;
-  typedef unsigned int   uint32;
-  typedef unsigned int   uint;
-
-  // JPEG chroma subsampling factors. Y_ONLY (grayscale images) and H2V2 (color images) are the most common.
-  enum subsampling_t { Y_ONLY = 0, H1V1 = 1, H2V1 = 2, H2V2 = 3 };
-
-  // JPEG compression parameters structure.
-  struct params
-  {
-    inline params() : m_quality(85), m_subsampling(H2V2), m_no_chroma_discrim_flag(false), m_two_pass_flag(false) { }
-
-    inline bool check_valid() const
-    {
-      if ((m_quality < 1) || (m_quality > 100)) return false;
-      if ((uint)m_subsampling > (uint)H2V2) return false;
-      return true;
-    }
-
-    // Quality: 1-100, higher is better. Typical values are around 50-95.
-    int m_quality;
-
-    // m_subsampling:
-    // 0 = Y (grayscale) only
-    // 1 = YCbCr, no subsampling (H1V1, YCbCr 1x1x1, 3 blocks per MCU)
-    // 2 = YCbCr, H2V1 subsampling (YCbCr 2x1x1, 4 blocks per MCU)
-    // 3 = YCbCr, H2V2 subsampling (YCbCr 4x1x1, 6 blocks per MCU-- very common)
-    subsampling_t m_subsampling;
-
-    // Disables CbCr discrimination - only intended for testing.
-    // If true, the Y quantization table is also used for the CbCr channels.
-    bool m_no_chroma_discrim_flag;
-
-    bool m_two_pass_flag;
-  };
-  
-  // Writes JPEG image to a file. 
-  // num_channels must be 1 (Y) or 3 (RGB), image pitch must be width*num_channels.
-  bool compress_image_to_jpeg_file(const char *pFilename, int64_t width, int64_t height, int64_t num_channels, const uint8 *pImage_data, const params &comp_params = params());
-
-  // Writes JPEG image to memory buffer. 
-  // On entry, buf_size is the size of the output buffer pointed at by pBuf, which should be at least ~1024 bytes. 
-  // If return value is true, buf_size will be set to the size of the compressed data.
-  bool compress_image_to_jpeg_file_in_memory(void *pBuf, int64_t &buf_size, int64_t width, int64_t height, int64_t num_channels, const uint8 *pImage_data, const params &comp_params = params());
-    
-  // Output stream abstract class - used by the jpeg_encoder class to write to the output stream. 
-  // put_buf() is generally called with len==JPGE_OUT_BUF_SIZE bytes, but for headers it'll be called with smaller amounts.
-  class output_stream
-  {
-  public:
-    virtual ~output_stream() { };
-    virtual bool put_buf(const void* Pbuf, int64_t len) = 0;
-    template<class T> inline bool put_obj(const T& obj) { return put_buf(&obj, sizeof(T)); }
-  };
-    
-  // Lower level jpeg_encoder class - useful if more control is needed than the above helper functions.
-  class jpeg_encoder
-  {
-  public:
-    jpeg_encoder();
-    ~jpeg_encoder();
-
-    // Initializes the compressor.
-    // pStream: The stream object to use for writing compressed data.
-    // params - Compression parameters structure, defined above.
-    // width, height  - Image dimensions.
-    // channels - May be 1, or 3. 1 indicates grayscale, 3 indicates RGB source data.
-    // Returns false on out of memory or if a stream write fails.
-    bool init(output_stream *pStream, int64_t width, int64_t height, int64_t src_channels, const params &comp_params = params());
-    
-    const params &get_params() const { return m_params; }
-    
-    // Deinitializes the compressor, freeing any allocated memory. May be called at any time.
-    void deinit();
-
-    uint get_total_passes() const { return m_params.m_two_pass_flag ? 2 : 1; }
-    inline uint get_cur_pass() { return m_pass_num; }
-
-    // Call this method with each source scanline.
-    // width * src_channels bytes per scanline is expected (RGB or Y format).
-    // You must call with NULL after all scanlines are processed to finish compression.
-    // Returns false on out of memory or if a stream write fails.
-    bool process_scanline(const void* pScanline);
-        
-  private:
-    jpeg_encoder(const jpeg_encoder &);
-    jpeg_encoder &operator =(const jpeg_encoder &);
-
-    typedef int32 sample_array_t;
-        
-    output_stream *m_pStream;
-    params m_params;
-    uint8 m_num_components;
-    uint8 m_comp_h_samp[3], m_comp_v_samp[3];
-    int m_image_x, m_image_y, m_image_bpp, m_image_bpl;
-    int m_image_x_mcu, m_image_y_mcu;
-    int m_image_bpl_xlt, m_image_bpl_mcu;
-    int m_mcus_per_row;
-    int m_mcu_x, m_mcu_y;
-    uint8 *m_mcu_lines[16];
-    uint8 m_mcu_y_ofs;
-    sample_array_t m_sample_array[64];
-    int16 m_coefficient_array[64];
-    int32 m_quantization_tables[2][64];
-    uint m_huff_codes[4][256];
-    uint8 m_huff_code_sizes[4][256];
-    uint8 m_huff_bits[4][17];
-    uint8 m_huff_val[4][256];
-    uint32 m_huff_count[4][256];
-    int m_last_dc_val[3];
-    enum { JPGE_OUT_BUF_SIZE = 2048 };
-    uint8 m_out_buf[JPGE_OUT_BUF_SIZE];
-    uint8 *m_pOut_buf;
-    uint m_out_buf_left;
-    uint32 m_bit_buffer;
-    uint m_bits_in;
-    uint8 m_pass_num;
-    bool m_all_stream_writes_succeeded;
-        
-    void optimize_huffman_table(int table_num, int table_len);
-    void emit_byte(uint8 i);
-    void emit_word(uint i);
-    void emit_marker(int marker);
-    void emit_jfif_app0();
-    void emit_dqt();
-    void emit_sof();
-    void emit_dht(uint8 *bits, uint8 *val, int index, bool ac_flag);
-    void emit_dhts();
-    void emit_sos();
-    void emit_markers();
-    void compute_huffman_table(uint *codes, uint8 *code_sizes, uint8 *bits, uint8 *val);
-    void compute_quant_table(int32 *dst, int16 *src);
-    void adjust_quant_table(int32 *dst, int32 *src);
-    void first_pass_init();
-    bool second_pass_init();
-    bool jpg_open(int p_x_res, int p_y_res, int src_channels);
-    void load_block_8_8_grey(int x);
-    void load_block_8_8(int x, int y, int c);
-    void load_block_16_8(int x, int c);
-    void load_block_16_8_8(int x, int c);
-    void load_quantized_coefficients(int component_num);
-    void flush_output_buffer();
-    void put_bits(uint bits, uint len);
-    void code_coefficients_pass_one(int component_num);
-    void code_coefficients_pass_two(int component_num);
-    void code_block(int component_num);
-    void process_mcu_row();
-    bool terminate_pass_one();
-    bool terminate_pass_two();
-    bool process_end_of_image();
-    void load_mcu(const void* src);
-    void clear();
-    void init();
-  };
-
-} // namespace jpge
-
-#endif // JPEG_ENCODER

crazy_functions/test_project/cpp/libJPG/来源

@@ -1,3 +0,0 @@
-jpge.h - C++ class for JPEG compression.
-Public domain, Rich Geldreich <richgel99@gmail.com>
-Alex Evans: Added RGBA support, linear memory allocator.

crazy_functions/test_project/cpp/longcode/prod_cons.h

@@ -1,433 +0,0 @@
-#pragma once
-
-#include <atomic>
-#include <utility>
-#include <cstring>
-#include <type_traits>
-#include <cstdint>
-
-#include "libipc/def.h"
-
-#include "libipc/platform/detail.h"
-#include "libipc/circ/elem_def.h"
-#include "libipc/utility/log.h"
-#include "libipc/utility/utility.h"
-
-namespace ipc {
-
-////////////////////////////////////////////////////////////////
-/// producer-consumer implementation
-////////////////////////////////////////////////////////////////
-
-template <typename Flag>
-struct prod_cons_impl;
-
-template <>
-struct prod_cons_impl<wr<relat::single, relat::single, trans::unicast>> {
-
-    template <std::size_t DataSize, std::size_t AlignSize>
-    struct elem_t {
-        std::aligned_storage_t<DataSize, AlignSize> data_ {};
-    };
-
-    alignas(cache_line_size) std::atomic<circ::u2_t> rd_; // read index
-    alignas(cache_line_size) std::atomic<circ::u2_t> wt_; // write index
-
-    constexpr circ::u2_t cursor() const noexcept {
-        return 0;
-    }
-
-    template <typename W, typename F, typename E>
-    bool push(W* /*wrapper*/, F&& f, E* elems) {
-        auto cur_wt = circ::index_of(wt_.load(std::memory_order_relaxed));
-        if (cur_wt == circ::index_of(rd_.load(std::memory_order_acquire) - 1)) {
-            return false; // full
-        }
-        std::forward<F>(f)(&(elems[cur_wt].data_));
-        wt_.fetch_add(1, std::memory_order_release);
-        return true;
-    }
-
-    /**
-     * In single-single-unicast, 'force_push' means 'no reader' or 'the only one reader is dead'.
-     * So we could just disconnect all connections of receiver, and return false.
-    */
-    template <typename W, typename F, typename E>
-    bool force_push(W* wrapper, F&&, E*) {
-        wrapper->elems()->disconnect_receiver(~static_cast<circ::cc_t>(0u));
-        return false;
-    }
-
-    template <typename W, typename F, typename R, typename E>
-    bool pop(W* /*wrapper*/, circ::u2_t& /*cur*/, F&& f, R&& out, E* elems) {
-        auto cur_rd = circ::index_of(rd_.load(std::memory_order_relaxed));
-        if (cur_rd == circ::index_of(wt_.load(std::memory_order_acquire))) {
-            return false; // empty
-        }
-        std::forward<F>(f)(&(elems[cur_rd].data_));
-        std::forward<R>(out)(true);
-        rd_.fetch_add(1, std::memory_order_release);
-        return true;
-    }
-};
-
-template <>
-struct prod_cons_impl<wr<relat::single, relat::multi , trans::unicast>>
-     : prod_cons_impl<wr<relat::single, relat::single, trans::unicast>> {
-
-    template <typename W, typename F, typename E>
-    bool force_push(W* wrapper, F&&, E*) {
-        wrapper->elems()->disconnect_receiver(1);
-        return false;
-    }
-
-    template <typename W, typename F, typename R, 
-              template <std::size_t, std::size_t> class E, std::size_t DS, std::size_t AS>
-    bool pop(W* /*wrapper*/, circ::u2_t& /*cur*/, F&& f, R&& out, E<DS, AS>* elems) {
-        byte_t buff[DS];
-        for (unsigned k = 0;;) {
-            auto cur_rd = rd_.load(std::memory_order_relaxed);
-            if (circ::index_of(cur_rd) ==
-                circ::index_of(wt_.load(std::memory_order_acquire))) {
-                return false; // empty
-            }
-            std::memcpy(buff, &(elems[circ::index_of(cur_rd)].data_), sizeof(buff));
-            if (rd_.compare_exchange_weak(cur_rd, cur_rd + 1, std::memory_order_release)) {
-                std::forward<F>(f)(buff);
-                std::forward<R>(out)(true);
-                return true;
-            }
-            ipc::yield(k);
-        }
-    }
-};
-
-template <>
-struct prod_cons_impl<wr<relat::multi , relat::multi, trans::unicast>>
-     : prod_cons_impl<wr<relat::single, relat::multi, trans::unicast>> {
-
-    using flag_t = std::uint64_t;
-
-    template <std::size_t DataSize, std::size_t AlignSize>
-    struct elem_t {
-        std::aligned_storage_t<DataSize, AlignSize> data_ {};
-        std::atomic<flag_t> f_ct_ { 0 }; // commit flag
-    };
-
-    alignas(cache_line_size) std::atomic<circ::u2_t> ct_; // commit index
-
-    template <typename W, typename F, typename E>
-    bool push(W* /*wrapper*/, F&& f, E* elems) {
-        circ::u2_t cur_ct, nxt_ct;
-        for (unsigned k = 0;;) {
-            cur_ct = ct_.load(std::memory_order_relaxed);
-            if (circ::index_of(nxt_ct = cur_ct + 1) ==
-                circ::index_of(rd_.load(std::memory_order_acquire))) {
-                return false; // full
-            }
-            if (ct_.compare_exchange_weak(cur_ct, nxt_ct, std::memory_order_acq_rel)) {
-                break;
-            }
-            ipc::yield(k);
-        }
-        auto* el = elems + circ::index_of(cur_ct);
-        std::forward<F>(f)(&(el->data_));
-        // set flag & try update wt
-        el->f_ct_.store(~static_cast<flag_t>(cur_ct), std::memory_order_release);
-        while (1) {
-            auto cac_ct = el->f_ct_.load(std::memory_order_acquire);
-            if (cur_ct != wt_.load(std::memory_order_relaxed)) {
-                return true;
-            }
-            if ((~cac_ct) != cur_ct) {
-                return true;
-            }
-            if (!el->f_ct_.compare_exchange_strong(cac_ct, 0, std::memory_order_relaxed)) {
-                return true;
-            }
-            wt_.store(nxt_ct, std::memory_order_release);
-            cur_ct = nxt_ct;
-            nxt_ct = cur_ct + 1;
-            el = elems + circ::index_of(cur_ct);
-        }
-        return true;
-    }
-
-    template <typename W, typename F, typename E>
-    bool force_push(W* wrapper, F&&, E*) {
-        wrapper->elems()->disconnect_receiver(1);
-        return false;
-    }
-
-    template <typename W, typename F, typename R, 
-              template <std::size_t, std::size_t> class E, std::size_t DS, std::size_t AS>
-    bool pop(W* /*wrapper*/, circ::u2_t& /*cur*/, F&& f, R&& out, E<DS, AS>* elems) {
-        byte_t buff[DS];
-        for (unsigned k = 0;;) {
-            auto cur_rd = rd_.load(std::memory_order_relaxed);
-            auto cur_wt = wt_.load(std::memory_order_acquire);
-            auto id_rd  = circ::index_of(cur_rd);
-            auto id_wt  = circ::index_of(cur_wt);
-            if (id_rd == id_wt) {
-                auto* el = elems + id_wt;
-                auto cac_ct = el->f_ct_.load(std::memory_order_acquire);
-                if ((~cac_ct) != cur_wt) {
-                    return false; // empty
-                }
-                if (el->f_ct_.compare_exchange_weak(cac_ct, 0, std::memory_order_relaxed)) {
-                    wt_.store(cur_wt + 1, std::memory_order_release);
-                }
-                k = 0;
-            }
-            else {
-                std::memcpy(buff, &(elems[circ::index_of(cur_rd)].data_), sizeof(buff));
-                if (rd_.compare_exchange_weak(cur_rd, cur_rd + 1, std::memory_order_release)) {
-                    std::forward<F>(f)(buff);
-                    std::forward<R>(out)(true);
-                    return true;
-                }
-                ipc::yield(k);
-            }
-        }
-    }
-};
-
-template <>
-struct prod_cons_impl<wr<relat::single, relat::multi, trans::broadcast>> {
-
-    using rc_t = std::uint64_t;
-
-    enum : rc_t {
-        ep_mask = 0x00000000ffffffffull,
-        ep_incr = 0x0000000100000000ull
-    };
-
-    template <std::size_t DataSize, std::size_t AlignSize>
-    struct elem_t {
-        std::aligned_storage_t<DataSize, AlignSize> data_ {};
-        std::atomic<rc_t> rc_ { 0 }; // read-counter
-    };
-
-    alignas(cache_line_size) std::atomic<circ::u2_t> wt_;   // write index
-    alignas(cache_line_size) rc_t epoch_ { 0 };             // only one writer
-
-    circ::u2_t cursor() const noexcept {
-        return wt_.load(std::memory_order_acquire);
-    }
-
-    template <typename W, typename F, typename E>
-    bool push(W* wrapper, F&& f, E* elems) {
-        E* el;
-        for (unsigned k = 0;;) {
-            circ::cc_t cc = wrapper->elems()->connections(std::memory_order_relaxed);
-            if (cc == 0) return false; // no reader
-            el = elems + circ::index_of(wt_.load(std::memory_order_relaxed));
-            // check all consumers have finished reading this element
-            auto cur_rc = el->rc_.load(std::memory_order_acquire);
-            circ::cc_t rem_cc = cur_rc & ep_mask;
-            if ((cc & rem_cc) && ((cur_rc & ~ep_mask) == epoch_)) {
-                return false; // has not finished yet
-            }
-            // consider rem_cc to be 0 here
-            if (el->rc_.compare_exchange_weak(
-                        cur_rc, epoch_ | static_cast<rc_t>(cc), std::memory_order_release)) {
-                break;
-            }
-            ipc::yield(k);
-        }
-        std::forward<F>(f)(&(el->data_));
-        wt_.fetch_add(1, std::memory_order_release);
-        return true;
-    }
-
-    template <typename W, typename F, typename E>
-    bool force_push(W* wrapper, F&& f, E* elems) {
-        E* el;
-        epoch_ += ep_incr;
-        for (unsigned k = 0;;) {
-            circ::cc_t cc = wrapper->elems()->connections(std::memory_order_relaxed);
-            if (cc == 0) return false; // no reader
-            el = elems + circ::index_of(wt_.load(std::memory_order_relaxed));
-            // check all consumers have finished reading this element
-            auto cur_rc = el->rc_.load(std::memory_order_acquire);
-            circ::cc_t rem_cc = cur_rc & ep_mask;
-            if (cc & rem_cc) {
-                ipc::log("force_push: k = %u, cc = %u, rem_cc = %u\n", k, cc, rem_cc);
-                cc = wrapper->elems()->disconnect_receiver(rem_cc); // disconnect all invalid readers
-                if (cc == 0) return false; // no reader
-            }
-            // just compare & exchange
-            if (el->rc_.compare_exchange_weak(
-                        cur_rc, epoch_ | static_cast<rc_t>(cc), std::memory_order_release)) {
-                break;
-            }
-            ipc::yield(k);
-        }
-        std::forward<F>(f)(&(el->data_));
-        wt_.fetch_add(1, std::memory_order_release);
-        return true;
-    }
-
-    template <typename W, typename F, typename R, typename E>
-    bool pop(W* wrapper, circ::u2_t& cur, F&& f, R&& out, E* elems) {
-        if (cur == cursor()) return false; // acquire
-        auto* el = elems + circ::index_of(cur++);
-        std::forward<F>(f)(&(el->data_));
-        for (unsigned k = 0;;) {
-            auto cur_rc = el->rc_.load(std::memory_order_acquire);
-            if ((cur_rc & ep_mask) == 0) {
-                std::forward<R>(out)(true);
-                return true;
-            }
-            auto nxt_rc = cur_rc & ~static_cast<rc_t>(wrapper->connected_id());
-            if (el->rc_.compare_exchange_weak(cur_rc, nxt_rc, std::memory_order_release)) {
-                std::forward<R>(out)((nxt_rc & ep_mask) == 0);
-                return true;
-            }
-            ipc::yield(k);
-        }
-    }
-};
-
-template <>
-struct prod_cons_impl<wr<relat::multi, relat::multi, trans::broadcast>> {
-
-    using rc_t   = std::uint64_t;
-    using flag_t = std::uint64_t;
-
-    enum : rc_t {
-        rc_mask = 0x00000000ffffffffull,
-        ep_mask = 0x00ffffffffffffffull,
-        ep_incr = 0x0100000000000000ull,
-        ic_mask = 0xff000000ffffffffull,
-        ic_incr = 0x0000000100000000ull
-    };
-
-    template <std::size_t DataSize, std::size_t AlignSize>
-    struct elem_t {
-        std::aligned_storage_t<DataSize, AlignSize> data_ {};
-        std::atomic<rc_t  > rc_   { 0 }; // read-counter
-        std::atomic<flag_t> f_ct_ { 0 }; // commit flag
-    };
-
-    alignas(cache_line_size) std::atomic<circ::u2_t> ct_;   // commit index
-    alignas(cache_line_size) std::atomic<rc_t> epoch_ { 0 };
-
-    circ::u2_t cursor() const noexcept {
-        return ct_.load(std::memory_order_acquire);
-    }
-
-    constexpr static rc_t inc_rc(rc_t rc) noexcept {
-        return (rc & ic_mask) | ((rc + ic_incr) & ~ic_mask);
-    }
-
-    constexpr static rc_t inc_mask(rc_t rc) noexcept {
-        return inc_rc(rc) & ~rc_mask;
-    }
-
-    template <typename W, typename F, typename E>
-    bool push(W* wrapper, F&& f, E* elems) {
-        E* el;
-        circ::u2_t cur_ct;
-        rc_t epoch = epoch_.load(std::memory_order_acquire);
-        for (unsigned k = 0;;) {
-            circ::cc_t cc = wrapper->elems()->connections(std::memory_order_relaxed);
-            if (cc == 0) return false; // no reader
-            el = elems + circ::index_of(cur_ct = ct_.load(std::memory_order_relaxed));
-            // check all consumers have finished reading this element
-            auto cur_rc = el->rc_.load(std::memory_order_relaxed);
-            circ::cc_t rem_cc = cur_rc & rc_mask;
-            if ((cc & rem_cc) && ((cur_rc & ~ep_mask) == epoch)) {
-                return false; // has not finished yet
-            }
-            else if (!rem_cc) {
-                auto cur_fl = el->f_ct_.load(std::memory_order_acquire);
-                if ((cur_fl != cur_ct) && cur_fl) {
-                    return false; // full
-                }
-            }
-            // consider rem_cc to be 0 here
-            if (el->rc_.compare_exchange_weak(
-                        cur_rc, inc_mask(epoch | (cur_rc & ep_mask)) | static_cast<rc_t>(cc), std::memory_order_relaxed) &&
-                epoch_.compare_exchange_weak(epoch, epoch, std::memory_order_acq_rel)) {
-                break;
-            }
-            ipc::yield(k);
-        }
-        // only one thread/process would touch here at one time
-        ct_.store(cur_ct + 1, std::memory_order_release);
-        std::forward<F>(f)(&(el->data_));
-        // set flag & try update wt
-        el->f_ct_.store(~static_cast<flag_t>(cur_ct), std::memory_order_release);
-        return true;
-    }
-
-    template <typename W, typename F, typename E>
-    bool force_push(W* wrapper, F&& f, E* elems) {
-        E* el;
-        circ::u2_t cur_ct;
-        rc_t epoch = epoch_.fetch_add(ep_incr, std::memory_order_release) + ep_incr;
-        for (unsigned k = 0;;) {
-            circ::cc_t cc = wrapper->elems()->connections(std::memory_order_relaxed);
-            if (cc == 0) return false; // no reader
-            el = elems + circ::index_of(cur_ct = ct_.load(std::memory_order_relaxed));
-            // check all consumers have finished reading this element
-            auto cur_rc = el->rc_.load(std::memory_order_acquire);
-            circ::cc_t rem_cc = cur_rc & rc_mask;
-            if (cc & rem_cc) {
-                ipc::log("force_push: k = %u, cc = %u, rem_cc = %u\n", k, cc, rem_cc);
-                cc = wrapper->elems()->disconnect_receiver(rem_cc); // disconnect all invalid readers
-                if (cc == 0) return false; // no reader
-            }
-            // just compare & exchange
-            if (el->rc_.compare_exchange_weak(
-                        cur_rc, inc_mask(epoch | (cur_rc & ep_mask)) | static_cast<rc_t>(cc), std::memory_order_relaxed)) {
-                if (epoch == epoch_.load(std::memory_order_acquire)) {
-                    break;
-                }
-                else if (push(wrapper, std::forward<F>(f), elems)) {
-                    return true;
-                }
-                epoch = epoch_.fetch_add(ep_incr, std::memory_order_release) + ep_incr;
-            }
-            ipc::yield(k);
-        }
-        // only one thread/process would touch here at one time
-        ct_.store(cur_ct + 1, std::memory_order_release);
-        std::forward<F>(f)(&(el->data_));
-        // set flag & try update wt
-        el->f_ct_.store(~static_cast<flag_t>(cur_ct), std::memory_order_release);
-        return true;
-    }
-
-    template <typename W, typename F, typename R, typename E, std::size_t N>
-    bool pop(W* wrapper, circ::u2_t& cur, F&& f, R&& out, E(& elems)[N]) {
-        auto* el = elems + circ::index_of(cur);
-        auto cur_fl = el->f_ct_.load(std::memory_order_acquire);
-        if (cur_fl != ~static_cast<flag_t>(cur)) {
-            return false; // empty
-        }
-        ++cur;
-        std::forward<F>(f)(&(el->data_));
-        for (unsigned k = 0;;) {
-            auto cur_rc = el->rc_.load(std::memory_order_acquire);
-            if ((cur_rc & rc_mask) == 0) {
-                std::forward<R>(out)(true);
-                el->f_ct_.store(cur + N - 1, std::memory_order_release);
-                return true;
-            }
-            auto nxt_rc = inc_rc(cur_rc) & ~static_cast<rc_t>(wrapper->connected_id());
-            bool last_one = false;
-            if ((last_one = (nxt_rc & rc_mask) == 0)) {
-                el->f_ct_.store(cur + N - 1, std::memory_order_release);
-            }
-            if (el->rc_.compare_exchange_weak(cur_rc, nxt_rc, std::memory_order_release)) {
-                std::forward<R>(out)(last_one);
-                return true;
-            }
-            ipc::yield(k);
-        }
-    }
-};
-
-} // namespace ipc

crazy_functions/test_project/latex/attention/background.tex

@@ -1,58 +0,0 @@
-The goal of reducing sequential computation also forms the foundation of the Extended Neural GPU \citep{extendedngpu}, ByteNet \citep{NalBytenet2017} and ConvS2S \citep{JonasFaceNet2017}, all of which use convolutional neural networks as basic building block, computing hidden representations in parallel for all input and output positions. In these models, the number of operations required to relate signals from two arbitrary input or output positions grows in the distance between positions, linearly for ConvS2S and logarithmically for ByteNet. This makes it more difficult to learn dependencies between distant positions \citep{hochreiter2001gradient}. In the Transformer this is reduced to a constant number of operations, albeit at the cost of reduced effective resolution due to averaging attention-weighted positions, an effect we counteract with Multi-Head Attention as described in section~\ref{sec:attention}. 
-
-Self-attention, sometimes called intra-attention is an attention mechanism relating different positions of a single sequence in order to compute a representation of the sequence. Self-attention has been used successfully in a variety of tasks including reading comprehension, abstractive summarization, textual entailment and learning task-independent sentence representations \citep{cheng2016long, decomposableAttnModel, paulus2017deep, lin2017structured}.
-
-End-to-end memory networks are based on a recurrent attention mechanism instead of sequence-aligned recurrence and have been shown to perform well on simple-language question answering and language modeling tasks \citep{sukhbaatar2015}.
-
-To the best of our knowledge, however, the Transformer is the first transduction model relying entirely on self-attention to compute representations of its input and output without using sequence-aligned RNNs or convolution.
-In the following sections, we will describe the Transformer, motivate self-attention and discuss its advantages over models such as \citep{neural_gpu, NalBytenet2017} and \citep{JonasFaceNet2017}.
-
-
-%\citep{JonasFaceNet2017} report new SOTA on machine translation for English-to-German (EnDe), Enlish-to-French (EnFr) and English-to-Romanian language pairs. 
-
-%For example,! in MT, we must draw information from both input and previous output words to translate an output word accurately. An attention layer \citep{bahdanau2014neural} can connect a very large number of positions at low computation cost, making it an essential ingredient in competitive recurrent models for machine translation.
-
-%A natural question to ask then is, "Could we replace recurrence with attention?". \marginpar{Don't know if it's the most natural question to ask given the previous statements. Also, need to say that the complexity table summarizes these statements} Such a model would be blessed with the computational efficiency of attention and the power of cross-positional communication. In this work, show that pure attention models work remarkably well for MT, achieving new SOTA results on EnDe and EnFr, and can be trained in under $2$ days on xyz architecture. 
-
-%After the seminal models introduced in \citep{sutskever14, bahdanau2014neural, cho2014learning}, recurrent models have become the dominant solution for both sequence modeling and sequence-to-sequence transduction. Many efforts such as \citep{wu2016google,luong2015effective,jozefowicz2016exploring} have pushed the boundaries of machine translation (MT) and language modeling with recurrent endoder-decoder and recurrent language models. Recent effort \citep{shazeer2017outrageously} has successfully combined the power of conditional computation with sequence models to train very large models for MT, pushing SOTA at lower computational cost.
-
-%Recurrent models compute a vector of hidden states $h_t$, for each time step $t$ of computation. $h_t$ is a function of both the input at time $t$ and the previous hidden state $h_t$. This dependence on the previous hidden state precludes processing all timesteps at once, instead requiring long sequences of sequential operations.  In practice, this results in greatly reduced computational efficiency, as on modern computing hardware, a single operation on a large batch is much faster than a large number of operations on small batches.  The problem gets worse at longer sequence lengths. Although sequential computation is not a severe bottleneck at inference time, as autoregressively generating each output requires all previous outputs, the inability to compute scores at all output positions at once hinders us from rapidly training our models over large datasets. Although impressive work such as \citep{Kuchaiev2017Factorization} is able to significantly accelerate the training of LSTMs with factorization tricks, we are still bound by the linear dependence on sequence length.
-
-%If the model could compute hidden states at each time step using only the inputs and outputs,  it would be liberated from the dependence on results from previous time steps during training. This line of thought is the foundation of recent efforts such as the Markovian neural GPU \citep{neural_gpu}, ByteNet \citep{NalBytenet2017} and ConvS2S \citep{JonasFaceNet2017}, all of which use convolutional neural networks as a building block to compute hidden representations simultaneously for all timesteps, resulting in $O(1)$ sequential time complexity. \citep{JonasFaceNet2017} report new SOTA on machine translation for English-to-German (EnDe), Enlish-to-French (EnFr) and English-to-Romanian language pairs. 
-
-%A crucial component for accurate sequence prediction is modeling cross-positional communication. For example, in MT, we must draw information from both input and previous output words to translate an output word accurately. An attention layer \citep{bahdanau2014neural} can connect a very large number of positions at a low computation cost, also $O(1)$ sequential time complexity, making it an essential ingredient in recurrent encoder-decoder architectures for MT. A natural question to ask then is, "Could we replace recurrence with attention?". \marginpar{Don't know if it's the most natural question to ask given the previous statements. Also, need to say that the complexity table summarizes these statements} Such a model would be blessed with the computational efficiency of attention and the power of cross-positional communication. In this work, show that pure attention models work remarkably well for MT, achieving new SOTA results on EnDe and EnFr, and can be trained in under $2$ days on xyz architecture. 
-
-
-
-%Note: Facebook model is no better than RNNs in this regard, since it requires a number of layers proportional to the distance you want to communicate.  Bytenet is more promising, since it requires a logarithmnic number of layers (does bytenet have SOTA results)?   
-
-%Note: An attention  layer can connect a very large number of positions at a low computation cost in O(1) sequential operations.  This is why encoder-decoder attention has been so successful in seq-to-seq models so far.  It is only natural, then, to also use attention to connect the timesteps of the same sequence.
-
-%Note: I wouldn't say that long sequences are not a problem during inference.  It would be great if we could infer with no long sequences.  We could just say later on that, while our training graph is constant-depth, our model still requires sequential operations in the decoder part during inference due to the autoregressive nature of the model.   
-
-%\begin{table}[h!]
-%\caption{Attention models are quite efficient for cross-positional communications when sequence length is smaller than channel depth. $n$ represents the sequence length and $d$ represents the channel depth.}
-%\label{tab:op_complexities}
-%\begin{center}
-%\vspace{-5pt}
-%\scalebox{0.75}{
-
-%\begin{tabular}{l|c|c|c}
-%\hline \hline
-%Layer Type & Receptive & Complexity & Sequential  \\
-%           & Field     &            & Operations  \\
-%\hline
-%Pointwise Feed-Forward & $1$ & $O(n \cdot d^2)$ & $O(1)$ \\
-%\hline
-%Recurrent & $n$ & $O(n \cdot d^2)$ & $O(n)$ \\
-%\hline
-%Convolutional & $r$ & $O(r \cdot n \cdot d^2)$ & $O(1)$ \\
-%\hline
-%Convolutional (separable) & $r$ & $O(r \cdot n \cdot d + n %\cdot d^2)$ & $O(1)$ \\
-%\hline
-%Attention & $r$ & $O(r \cdot n \cdot d)$ & $O(1)$ \\
-%\hline \hline
-%\end{tabular}
-%}
-%\end{center}
-%\end{table}

crazy_functions/test_project/latex/attention/introduction.tex

@@ -1,18 +0,0 @@
-Recurrent neural networks, long short-term memory \citep{hochreiter1997} and gated recurrent \citep{gruEval14} neural networks in particular, have been firmly established as state of the art approaches in sequence modeling and transduction problems such as language modeling and machine translation \citep{sutskever14, bahdanau2014neural, cho2014learning}. Numerous efforts have since continued to push the boundaries of recurrent language models and encoder-decoder architectures \citep{wu2016google,luong2015effective,jozefowicz2016exploring}.
-
-Recurrent models typically factor computation along the symbol positions of the input and output sequences. Aligning the positions to steps in computation time, they generate a sequence of hidden states $h_t$, as a function of the previous hidden state $h_{t-1}$ and the input for position $t$. This inherently sequential nature precludes parallelization within training examples, which becomes critical at longer sequence lengths, as memory constraints limit batching across examples.
-%\marginpar{not sure if the memory constraints are understandable here}
-Recent work has achieved significant improvements in computational efficiency through factorization tricks \citep{Kuchaiev2017Factorization} and conditional computation \citep{shazeer2017outrageously}, while also improving model performance in case of the latter. The fundamental constraint of sequential computation, however, remains.
-
-%\marginpar{@all: there is work on analyzing what attention really does in seq2seq models, couldn't find it right away} 
-
-Attention mechanisms have become an integral part of compelling sequence modeling and transduction models in various tasks, allowing modeling of dependencies without regard to their distance in the input or output sequences \citep{bahdanau2014neural, structuredAttentionNetworks}. In all but a few cases \citep{decomposableAttnModel}, however, such attention mechanisms are used in conjunction with a recurrent network.
-
-%\marginpar{not sure if "cross-positional communication" is understandable without explanation}
-%\marginpar{insert exact training times and stats for the model that reaches sota earliest, maybe even a single GPU model?}
-
-In this work we propose the Transformer, a model architecture eschewing recurrence and instead relying entirely on an attention mechanism to draw global dependencies between input and output. The Transformer allows for significantly more parallelization and can reach a new state of the art in translation quality after being trained for as little as twelve hours on eight P100 GPUs. 
-%\marginpar{you removed the constant number of repetitions part. I wrote it because I wanted to make it clear that the model does not only perform attention once, while it's also not recurrent. I thought that might be important to get across early.}
-
-% Just a standard paragraph with citations, rewrite.
-%After the seminal papers of \citep{sutskever14}, \citep{bahdanau2014neural}, and \citep{cho2014learning}, recurrent models have become the dominant solution for both sequence modeling and sequence-to-sequence transduction. Many efforts such as \citep{wu2016google,luong2015effective,jozefowicz2016exploring} have pushed the boundaries of machine translation and language modeling with recurrent sequence models. Recent effort \citep{shazeer2017outrageously} has combined the power of conditional computation with sequence models to train very large models for machine translation, pushing SOTA at lower computational cost. Recurrent models compute a vector of hidden states $h_t$, for each time step $t$ of computation. $h_t$ is a function of both the input at time $t$ and the previous hidden state $h_t$. This dependence on the previous hidden state encumbers recurrnet models to process multiple inputs at once, and their time complexity is a linear function of the length of the input and output, both during training and inference. [What I want to say here is that although this is fine during decoding, at training time, we are given both input and output and this linear nature does not allow the RNN to process all inputs and outputs simultaneously and haven't been used on datasets that are the of the scale of the web. What's the largest dataset we have ? . Talk about Nividia and possibly other's effors to speed up things, and possibly other efforts that alleviate this, but are still limited by it's comptuational nature]. Rest of the intro: What if you could construct the state based on the actual inputs and outputs, then you could construct them all at once. This has been the foundation of many promising recent efforts, bytenet,facenet (Also talk about quasi rnn here). Now we talk about attention!! Along with cell architectures such as long short-term meory (LSTM) \citep{hochreiter1997}, and gated recurrent units (GRUs) \citep{cho2014learning}, attention has emerged as an essential ingredient in successful sequence models, in particular for machine translation. In recent years, many, if not all, state-of-the-art (SOTA) results in machine translation have been achieved with attention-based sequence models \citep{wu2016google,luong2015effective,jozefowicz2016exploring}. Talk about the neon work on how it played with attention to do self attention! Then talk about what we do.

crazy_functions/test_project/latex/attention/model_architecture.tex

@@ -1,155 +0,0 @@
-
-\begin{figure}
-  \centering
-  \includegraphics[scale=0.6]{Figures/ModalNet-21}
-  \caption{The Transformer - model architecture.}
-  \label{fig:model-arch}
-\end{figure}
-
-% Although the primary workhorse of our model is attention, 
-%Our model maintains the encoder-decoder structure that is common to many so-called sequence-to-sequence models \citep{bahdanau2014neural,sutskever14}.  As in all such architectures, the encoder computes a representation of the input sequence, and the decoder consumes these representations along with the output tokens to autoregressively produce the output sequence.  Where, traditionally, the encoder and decoder contain stacks of recurrent or convolutional layers, our encoder and decoder stacks are composed of attention layers and position-wise feed-forward layers (Figure~\ref{fig:model-arch}).  The following sections describe the gross architecture and these particular components in detail.
-
-Most competitive neural sequence transduction models have an encoder-decoder structure \citep{cho2014learning,bahdanau2014neural,sutskever14}. Here, the encoder maps an input sequence of symbol representations $(x_1, ..., x_n)$ to a sequence of continuous representations $\mathbf{z} = (z_1, ..., z_n)$. Given $\mathbf{z}$, the decoder then generates an output sequence $(y_1,...,y_m)$ of symbols one element at a time. At each step the model is auto-regressive \citep{graves2013generating}, consuming the previously generated symbols as additional input when generating the next.
-
-The Transformer follows this overall architecture using stacked self-attention and point-wise, fully connected layers for both the encoder and decoder, shown in the left and right halves of Figure~\ref{fig:model-arch}, respectively.
-
-\subsection{Encoder and Decoder Stacks}
-
-\paragraph{Encoder:}The encoder is composed of a stack of $N=6$ identical layers. Each layer has two sub-layers. The first is a multi-head self-attention mechanism, and the second is a simple, position-wise fully connected feed-forward network.   We employ a residual connection \citep{he2016deep} around each of the two sub-layers, followed by layer normalization \cite{layernorm2016}.  That is, the output of each sub-layer is $\mathrm{LayerNorm}(x + \mathrm{Sublayer}(x))$, where $\mathrm{Sublayer}(x)$ is the function implemented by the sub-layer itself.  To facilitate these residual connections, all sub-layers in the model, as well as the embedding layers, produce outputs of dimension $\dmodel=512$.
-
-\paragraph{Decoder:}The decoder is also composed of a stack of $N=6$ identical layers.  In addition to the two sub-layers in each encoder layer, the decoder inserts a third sub-layer, which performs multi-head attention over the output of the encoder stack.  Similar to the encoder, we employ residual connections around each of the sub-layers, followed by layer normalization.  We also modify the self-attention sub-layer in the decoder stack to prevent positions from attending to subsequent positions.  This masking, combined with fact that the output embeddings are offset by one position, ensures that the predictions for position $i$ can depend only on the known outputs at positions less than $i$.
-
-% In our model (Figure~\ref{fig:model-arch}), the encoder and decoder are composed of stacks of alternating self-attention layers (for cross-positional communication) and position-wise feed-forward layers (for in-place computation).  In addition, the decoder stack contains encoder-decoder attention layers.  Since attention is agnostic to the distances between words, our model requires a "positional encoding" to be added to the encoder and decoder input. The following sections describe all of these components in detail.
-
-\subsection{Attention} \label{sec:attention}
-An attention function can be described as mapping a query and a set of key-value pairs to an output, where the query, keys, values, and output are all vectors.  The output is computed as a weighted sum of the values, where the weight assigned to each value is computed by a compatibility function of the query with the corresponding key.
-
-\subsubsection{Scaled Dot-Product Attention} \label{sec:scaled-dot-prod}
-
-% \begin{figure}
-%   \centering
-%   \includegraphics[scale=0.6]{Figures/ModalNet-19}
-%   \caption{Scaled Dot-Product Attention.}
-%   \label{fig:multi-head-att}
-% \end{figure}
-
-We call our particular attention "Scaled Dot-Product Attention" (Figure~\ref{fig:multi-head-att}).   The input consists of queries and keys of dimension $d_k$, and values of dimension $d_v$.  We compute the dot products of the query with all keys, divide each by $\sqrt{d_k}$, and apply a softmax function to obtain the weights on the values.
-
-In practice, we compute the attention function on a set of queries simultaneously, packed together into a matrix $Q$.   The keys and values are also packed together into matrices $K$ and $V$.  We compute the matrix of outputs as:
-
-\begin{equation}
-   \mathrm{Attention}(Q, K, V) = \mathrm{softmax}(\frac{QK^T}{\sqrt{d_k}})V
-\end{equation}
-
-The two most commonly used attention functions are additive attention \citep{bahdanau2014neural}, and dot-product (multiplicative) attention.  Dot-product attention is identical to our algorithm, except for the scaling factor of $\frac{1}{\sqrt{d_k}}$. Additive attention computes the compatibility function using a feed-forward network with a single hidden layer.  While the two are similar in theoretical complexity, dot-product attention is much faster and more space-efficient in practice, since it can be implemented using highly optimized matrix multiplication code. 
-
-%We scale the dot products by $1/\sqrt{d_k}$ to limit the magnitude of the dot products, which works well in practice. Otherwise, we found applying the softmax to often result in weights very close to 0 or 1, and hence minuscule gradients.
-
-% Already described in the subsequent section
-%When used as part of decoder self-attention, an optional mask function is applied just before the softmax to prevent positions from attending to subsequent positions.   This mask simply sets the logits corresponding to all illegal connections (those outside of the lower triangle) to $-\infty$.
-
-%\paragraph{Comparison to Additive Attention: } We choose dot product attention over additive attention \citep{bahdanau2014neural} since it can be computed using highly optimized matrix multiplication code.  This optimization is particularly important to us, as we employ many attention layers in our model.
-
-While for small values of $d_k$ the two mechanisms perform similarly, additive attention outperforms dot product attention without scaling for larger values of $d_k$ \citep{DBLP:journals/corr/BritzGLL17}. We suspect that for large values of $d_k$, the dot products grow large in magnitude, pushing the softmax function into regions where it has extremely small gradients  \footnote{To illustrate why the dot products get large, assume that the components of $q$ and $k$ are independent random variables with mean $0$ and variance $1$.  Then their dot product, $q \cdot k = \sum_{i=1}^{d_k} q_ik_i$, has mean $0$ and variance $d_k$.}. To counteract this effect, we scale the dot products by $\frac{1}{\sqrt{d_k}}$.
-
-
-%We suspect this to be caused by the dot products growing too large in magnitude to result in useful gradients after applying the softmax function.  To counteract this, we scale the dot product by $1/\sqrt{d_k}$.
-
-
-\subsubsection{Multi-Head Attention} \label{sec:multihead}
-
-\begin{figure}
-\begin{minipage}[t]{0.5\textwidth}
-  \centering
-  Scaled Dot-Product Attention \\
-  \vspace{0.5cm}
-  \includegraphics[scale=0.6]{Figures/ModalNet-19}
-\end{minipage}
-\begin{minipage}[t]{0.5\textwidth}
-  \centering 
-  Multi-Head Attention \\
-  \vspace{0.1cm}
-  \includegraphics[scale=0.6]{Figures/ModalNet-20}  
-\end{minipage}
-
-
-  % \centering
-
-  \caption{(left) Scaled Dot-Product Attention. (right) Multi-Head Attention consists of several attention layers running in parallel.}
-  \label{fig:multi-head-att}
-\end{figure}
-
-Instead of performing a single attention function with $\dmodel$-dimensional keys, values and queries, we found it beneficial to linearly project the queries, keys and values $h$ times with different, learned linear projections to $d_k$, $d_k$ and $d_v$ dimensions, respectively.
-On each of these projected versions of queries, keys and values we then perform the attention function in parallel, yielding $d_v$-dimensional output values. These are concatenated and once again projected, resulting in the final values, as depicted in Figure~\ref{fig:multi-head-att}.
-
-Multi-head attention allows the model to jointly attend to information from different representation subspaces at different positions. With a single attention head, averaging inhibits this.
-
-\begin{align*}
-    \mathrm{MultiHead}(Q, K, V) &= \mathrm{Concat}(\mathrm{head_1}, ..., \mathrm{head_h})W^O\\
-%    \mathrm{where} \mathrm{head_i} &= \mathrm{Attention}(QW_Q_i^{\dmodel \times d_q}, KW_K_i^{\dmodel \times d_k}, VW^V_i^{\dmodel \times d_v})\\
-    \text{where}~\mathrm{head_i} &= \mathrm{Attention}(QW^Q_i, KW^K_i, VW^V_i)\\
-\end{align*}
-
-Where the projections are parameter matrices $W^Q_i \in \mathbb{R}^{\dmodel \times d_k}$, $W^K_i \in \mathbb{R}^{\dmodel \times d_k}$, $W^V_i \in \mathbb{R}^{\dmodel \times d_v}$ and $W^O \in \mathbb{R}^{hd_v \times \dmodel}$.
-
-
-%find it better (and no more expensive) to have multiple parallel attention layers (each over the full set of positions) with proportionally lower-dimensional keys, values and queries.  We call this "Multi-Head Attention" (Figure~\ref{fig:multi-head-att}).  The keys, values, and queries for each of these parallel attention layers are computed by learned linear transformations of the inputs to the multi-head attention.  We use different linear transformations across different parallel attention layers.  The output of the parallel attention layers are concatenated, and then passed through a final learned linear transformation. 
-
-In this work we employ $h=8$ parallel attention layers, or heads. For each of these we use $d_k=d_v=\dmodel/h=64$.
-Due to the reduced dimension of each head, the total computational cost is similar to that of single-head attention with full dimensionality.
-
-\subsubsection{Applications of Attention in our Model}
-
-The Transformer uses multi-head attention in three different ways: 
-\begin{itemize}
- \item In "encoder-decoder attention" layers, the queries come from the previous decoder layer, and the memory keys and values come from the output of the encoder.   This allows every position in the decoder to attend over all positions in the input sequence.  This mimics the typical encoder-decoder attention mechanisms in sequence-to-sequence models such as \citep{wu2016google, bahdanau2014neural,JonasFaceNet2017}.
-
- \item The encoder contains self-attention layers.  In a self-attention layer all of the keys, values and queries come from the same place, in this case, the output of the previous layer in the encoder.   Each position in the encoder can attend to all positions in the previous layer of the encoder.
-
- \item Similarly, self-attention layers in the decoder allow each position in the decoder to attend to all positions in the decoder up to and including that position.  We need to prevent leftward information flow in the decoder to preserve the auto-regressive property.  We implement this inside of scaled dot-product attention by masking out (setting to $-\infty$) all values in the input of the softmax which correspond to illegal connections.  See Figure~\ref{fig:multi-head-att}.
-
-\end{itemize}
-
-\subsection{Position-wise Feed-Forward Networks}\label{sec:ffn}
-
-In addition to attention sub-layers, each of the layers in our encoder and decoder contains a fully connected feed-forward network, which is applied to each position separately and identically.  This consists of two linear transformations with a ReLU activation in between.
-
-\begin{equation}
-   \mathrm{FFN}(x)=\max(0, xW_1 + b_1) W_2 + b_2
-\end{equation}
-
-While the linear transformations are the same across different positions, they use different parameters from layer to layer. Another way of describing this is as two convolutions with kernel size 1.  The dimensionality of input and output is $\dmodel=512$, and the inner-layer has dimensionality $d_{ff}=2048$.
-
-
-
-%In the appendix, we describe how the position-wise feed-forward network can also be seen as a form of attention.
-
-%from Jakob: The number of operations required for the model to relate signals from two arbitrary input or output positions grows in the distance between positions in input or output, linearly for ConvS2S and logarithmically for ByteNet, making it harder to learn dependencies between these positions \citep{hochreiter2001gradient}. In the transformer this is reduced to a constant number of operations, albeit at the cost of effective resolution caused by averaging attention-weighted positions, an effect we aim to counteract with multi-headed attention.
-
-
-%Figure~\ref{fig:simple-att} presents a simple attention function, $A$, with a single head, that forms the basis of our multi-head attention. $A$ takes a query key vector $\kq$, matrices of memory keys $\km$ and memory values $\vm$ ,and produces a query value vector $\vq$ as 
-%\begin{equation*} \label{eq:attention}
-%    A(\kq, \km, \vm) = {\vm}^T (Softmax(\km \kq).
-%\end{equation*}
-%We linearly transform $\kq,\,\km$, and $\vm$ with learned matrices ${\Wkq \text{,} \, \Wkm}$, and ${\Wvm}$ before calling the attention function, and transform the output query with $\Wvq$ before handing it to the feed forward layer. Each attention layer has it's own set of transformation matrices, which are shared across all query positions. $A$ is applied in parallel for each query position, and is implemented very efficiently as a batch of matrix multiplies. The self-attention and encoder-decoder attention layers use $A$, but with different arguments. For example, in encdoder self-attention, queries in encoder layer $i$ attention to memories in encoder layer $i-1$. To ensure that decoder self-attention layers do not look at future words, we add $- \inf$ to the softmax logits in positions $j+1$ to query length for query position $l$.  
-
-%In simple attention, the query value is a weighted combination of the memory values where the attention weights sum to one. Although this function performs well in practice, the constraint on attention weights can restrict the amount of information that flows from memories to queries because the query cannot focus on multiple memory positions at once, which might be desirable when translating long sequences. \marginpar{@usz, could you think of an example of this ?} We remedy this by maintaining multiple attention heads at each query position that attend to all memory positions in parallel, with a different set of parameters  per attention head $h$. 
-%\marginpar{}
-
-\subsection{Embeddings and Softmax}
-Similarly to other sequence transduction models, we use learned embeddings to convert the input tokens and output tokens to vectors of dimension $\dmodel$.  We also use the usual learned linear transformation and softmax function to convert the decoder output to predicted next-token probabilities.  In our model, we share the same weight matrix between the two embedding layers and the pre-softmax linear transformation, similar to \citep{press2016using}.   In the embedding layers, we multiply those weights by $\sqrt{\dmodel}$.
-
-
-\subsection{Positional Encoding}
-Since our model contains no recurrence and no convolution, in order for the model to make use of the order of the sequence, we must inject some information about the relative or absolute position of the tokens in the sequence.  To this end, we add "positional encodings" to the input embeddings at the bottoms of the encoder and decoder stacks.  The positional encodings have the same dimension $\dmodel$ as the embeddings, so that the two can be summed.   There are many choices of positional encodings, learned and fixed \citep{JonasFaceNet2017}.
-
-In this work, we use sine and cosine functions of different frequencies:
-
-\begin{align*}
-    PE_{(pos,2i)} = sin(pos / 10000^{2i/\dmodel}) \\
-    PE_{(pos,2i+1)} = cos(pos / 10000^{2i/\dmodel})
-\end{align*}
-
-where $pos$ is the position and $i$ is the dimension.  That is, each dimension of the positional encoding corresponds to a sinusoid.  The wavelengths form a geometric progression from $2\pi$ to $10000 \cdot 2\pi$.  We chose this function because we hypothesized it would allow the model to easily learn to attend by relative positions, since for any fixed offset $k$, $PE_{pos+k}$ can be represented as a linear function of $PE_{pos}$.
-
-We also experimented with using learned positional embeddings \citep{JonasFaceNet2017} instead, and found that the two versions produced nearly identical results (see Table~\ref{tab:variations} row (E)).  We chose the sinusoidal version because it may allow the model to extrapolate to sequence lengths longer than the ones encountered during training.

crazy_functions/test_project/latex/attention/parameter_attention.tex

@@ -1,45 +0,0 @@
-\pagebreak
-\section*{Two Feed-Forward Layers = Attention over Parameters}\label{sec:parameter_attention}
-
-In addition to attention layers, our model contains position-wise feed-forward networks (Section \ref{sec:ffn}), which consist of two linear transformations with a ReLU activation in between.  In fact, these networks too can be seen as a form of attention.  Compare the formula for such a network with the formula for a simple dot-product attention layer (biases and scaling factors omitted):
-
-\begin{align*}
-    FFN(x, W_1, W_2) = ReLU(xW_1)W_2 \\
-    A(q, K, V) = Softmax(qK^T)V
-\end{align*}
-
-Based on the similarity of these formulae, the two-layer feed-forward network can be seen as a kind of attention, where the keys and values are the rows of the trainable parameter matrices $W_1$ and $W_2$, and where we use ReLU instead of Softmax in the compatibility function.
-
-%the compatablity function is $compat(q, k_i) = ReLU(q \cdot k_i)$ instead of $Softmax(qK_T)_i$.
-
-Given this similarity, we experimented with replacing the position-wise feed-forward networks with attention layers similar to the ones we use everywhere else our model. The multi-head-attention-over-parameters sublayer is identical to the multi-head attention described in \ref{sec:multihead}, except that the "keys" and "values" inputs to each attention head are trainable model parameters, as opposed to being linear projections of a previous layer.  These parameters are scaled up by a factor of $\sqrt{d_{model}}$ in order to be more similar to activations.
-
-In our first experiment, we replaced each position-wise feed-forward network with a multi-head-attention-over-parameters sublayer with $h_p=8$ heads, key-dimensionality $d_{pk}=64$, and value-dimensionality $d_{pv}=64$, using $n_p=1536$ key-value pairs for each attention head.  The sublayer has a total of $2097152$ parameters, including the parameters in the query projection and the output projection.  This matches the number of parameters in the position-wise feed-forward network that we replaced.  While the theoretical amount of computation is also the same, in practice, the attention version caused the step times to be about 30\% longer.
-
-In our second experiment, we used $h_p=8$ heads, and $n_p=512$ key-value pairs for each attention head, again matching the total number of parameters in the base model.
-
-Results for the first experiment were slightly worse than for the base model, and results for the second experiment were slightly better, see Table~\ref{tab:parameter_attention}.
-
-\begin{table}[h]
-\caption{Replacing the position-wise feed-forward networks with multihead-attention-over-parameters produces similar results to the base model.  All metrics are on the English-to-German translation development set, newstest2013.}
-\label{tab:parameter_attention}
-\begin{center}
-\vspace{-2mm}
-%\scalebox{1.0}{
-\begin{tabular}{c|cccccc|cccc}
-\hline\rule{0pt}{2.0ex}
- & \multirow{2}{*}{$\dmodel$} & \multirow{2}{*}{$\dff$} &
-\multirow{2}{*}{$h_p$} & \multirow{2}{*}{$d_{pk}$} & \multirow{2}{*}{$d_{pv}$} &
- \multirow{2}{*}{$n_p$} &
- PPL & BLEU & params & training\\
- & & & & & &  & (dev) & (dev) & $\times10^6$ & time \\
-\hline\rule{0pt}{2.0ex}
-base & 512 & 2048 & & & & & 4.92 & 25.8 & 65 & 12 hours\\
-\hline\rule{0pt}{2.0ex}
-AOP$_1$ & 512 & & 8 & 64 & 64 & 1536 & 4.92& 25.5  & 65 & 16 hours\\
-AOP$_2$ & 512 & & 16 & 64 & 64 & 512 & \textbf{4.86} & \textbf{25.9}  & 65 & 16 hours \\
-\hline
-\end{tabular}
-%}
-\end{center}
-\end{table}

crazy_functions/test_project/latex/attention/来源

@@ -1,8 +0,0 @@
-chatgpt的老祖宗《Attention is all you need》
-
-Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, Illia Polosukhin
-
-真实的摘要如下
-The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data.
-
-https://arxiv.org/abs/1706.03762

crazy_functions/test_project/python/dqn/__init__.py

@@ -1,2 +0,0 @@
-from stable_baselines3.dqn.dqn import DQN
-from stable_baselines3.dqn.policies import CnnPolicy, MlpPolicy

crazy_functions/test_project/python/dqn/dqn.py

@@ -1,245 +0,0 @@
-from typing import Any, Dict, List, Optional, Tuple, Type, Union
-
-import gym
-import numpy as np
-import torch as th
-from torch.nn import functional as F
-
-from stable_baselines3.common import logger
-from stable_baselines3.common.off_policy_algorithm import OffPolicyAlgorithm
-from stable_baselines3.common.preprocessing import maybe_transpose
-from stable_baselines3.common.type_aliases import GymEnv, MaybeCallback, Schedule
-from stable_baselines3.common.utils import get_linear_fn, is_vectorized_observation, polyak_update
-from stable_baselines3.dqn.policies import DQNPolicy
-
-
-class DQN(OffPolicyAlgorithm):
-    """
-    Deep Q-Network (DQN)
-
-    Paper: https://arxiv.org/abs/1312.5602, https://www.nature.com/articles/nature14236
-    Default hyperparameters are taken from the nature paper,
-    except for the optimizer and learning rate that were taken from Stable Baselines defaults.
-
-    :param policy: The policy model to use (MlpPolicy, CnnPolicy, ...)
-    :param env: The environment to learn from (if registered in Gym, can be str)
-    :param learning_rate: The learning rate, it can be a function
-        of the current progress remaining (from 1 to 0)
-    :param buffer_size: size of the replay buffer
-    :param learning_starts: how many steps of the model to collect transitions for before learning starts
-    :param batch_size: Minibatch size for each gradient update
-    :param tau: the soft update coefficient ("Polyak update", between 0 and 1) default 1 for hard update
-    :param gamma: the discount factor
-    :param train_freq: Update the model every ``train_freq`` steps. Alternatively pass a tuple of frequency and unit
-        like ``(5, "step")`` or ``(2, "episode")``.
-    :param gradient_steps: How many gradient steps to do after each rollout (see ``train_freq``)
-        Set to ``-1`` means to do as many gradient steps as steps done in the environment
-        during the rollout.
-    :param optimize_memory_usage: Enable a memory efficient variant of the replay buffer
-        at a cost of more complexity.
-        See https://github.com/DLR-RM/stable-baselines3/issues/37#issuecomment-637501195
-    :param target_update_interval: update the target network every ``target_update_interval``
-        environment steps.
-    :param exploration_fraction: fraction of entire training period over which the exploration rate is reduced
-    :param exploration_initial_eps: initial value of random action probability
-    :param exploration_final_eps: final value of random action probability
-    :param max_grad_norm: The maximum value for the gradient clipping
-    :param tensorboard_log: the log location for tensorboard (if None, no logging)
-    :param create_eval_env: Whether to create a second environment that will be
-        used for evaluating the agent periodically. (Only available when passing string for the environment)
-    :param policy_kwargs: additional arguments to be passed to the policy on creation
-    :param verbose: the verbosity level: 0 no output, 1 info, 2 debug
-    :param seed: Seed for the pseudo random generators
-    :param device: Device (cpu, cuda, ...) on which the code should be run.
-        Setting it to auto, the code will be run on the GPU if possible.
-    :param _init_setup_model: Whether or not to build the network at the creation of the instance
-    """
-
-    def __init__(
-        self,
-        policy: Union[str, Type[DQNPolicy]],
-        env: Union[GymEnv, str],
-        learning_rate: Union[float, Schedule] = 1e-4,
-        buffer_size: int = 1000000,
-        learning_starts: int = 50000,
-        batch_size: Optional[int] = 32,
-        tau: float = 1.0,
-        gamma: float = 0.99,
-        train_freq: Union[int, Tuple[int, str]] = 4,
-        gradient_steps: int = 1,
-        optimize_memory_usage: bool = False,
-        target_update_interval: int = 10000,
-        exploration_fraction: float = 0.1,
-        exploration_initial_eps: float = 1.0,
-        exploration_final_eps: float = 0.05,
-        max_grad_norm: float = 10,
-        tensorboard_log: Optional[str] = None,
-        create_eval_env: bool = False,
-        policy_kwargs: Optional[Dict[str, Any]] = None,
-        verbose: int = 0,
-        seed: Optional[int] = None,
-        device: Union[th.device, str] = "auto",
-        _init_setup_model: bool = True,
-    ):
-
-        super(DQN, self).__init__(
-            policy,
-            env,
-            DQNPolicy,
-            learning_rate,
-            buffer_size,
-            learning_starts,
-            batch_size,
-            tau,
-            gamma,
-            train_freq,
-            gradient_steps,
-            action_noise=None,  # No action noise
-            policy_kwargs=policy_kwargs,
-            tensorboard_log=tensorboard_log,
-            verbose=verbose,
-            device=device,
-            create_eval_env=create_eval_env,
-            seed=seed,
-            sde_support=False,
-            optimize_memory_usage=optimize_memory_usage,
-            supported_action_spaces=(gym.spaces.Discrete,),
-        )
-
-        self.exploration_initial_eps = exploration_initial_eps
-        self.exploration_final_eps = exploration_final_eps
-        self.exploration_fraction = exploration_fraction
-        self.target_update_interval = target_update_interval
-        self.max_grad_norm = max_grad_norm
-        # "epsilon" for the epsilon-greedy exploration
-        self.exploration_rate = 0.0
-        # Linear schedule will be defined in `_setup_model()`
-        self.exploration_schedule = None
-        self.q_net, self.q_net_target = None, None
-
-        if _init_setup_model:
-            self._setup_model()
-
-    def _setup_model(self) -> None:
-        super(DQN, self)._setup_model()
-        self._create_aliases()
-        self.exploration_schedule = get_linear_fn(
-            self.exploration_initial_eps, self.exploration_final_eps, self.exploration_fraction
-        )
-
-    def _create_aliases(self) -> None:
-        self.q_net = self.policy.q_net
-        self.q_net_target = self.policy.q_net_target
-
-    def _on_step(self) -> None:
-        """
-        Update the exploration rate and target network if needed.
-        This method is called in ``collect_rollouts()`` after each step in the environment.
-        """
-        if self.num_timesteps % self.target_update_interval == 0:
-            polyak_update(self.q_net.parameters(), self.q_net_target.parameters(), self.tau)
-
-        self.exploration_rate = self.exploration_schedule(self._current_progress_remaining)
-        logger.record("rollout/exploration rate", self.exploration_rate)
-
-    def train(self, gradient_steps: int, batch_size: int = 100) -> None:
-        # Update learning rate according to schedule
-        self._update_learning_rate(self.policy.optimizer)
-
-        losses = []
-        for _ in range(gradient_steps):
-            # Sample replay buffer
-            replay_data = self.replay_buffer.sample(batch_size, env=self._vec_normalize_env)
-
-            with th.no_grad():
-                # Compute the next Q-values using the target network
-                next_q_values = self.q_net_target(replay_data.next_observations)
-                # Follow greedy policy: use the one with the highest value
-                next_q_values, _ = next_q_values.max(dim=1)
-                # Avoid potential broadcast issue
-                next_q_values = next_q_values.reshape(-1, 1)
-                # 1-step TD target
-                target_q_values = replay_data.rewards + (1 - replay_data.dones) * self.gamma * next_q_values
-
-            # Get current Q-values estimates
-            current_q_values = self.q_net(replay_data.observations)
-
-            # Retrieve the q-values for the actions from the replay buffer
-            current_q_values = th.gather(current_q_values, dim=1, index=replay_data.actions.long())
-
-            # Compute Huber loss (less sensitive to outliers)
-            loss = F.smooth_l1_loss(current_q_values, target_q_values)
-            losses.append(loss.item())
-
-            # Optimize the policy
-            self.policy.optimizer.zero_grad()
-            loss.backward()
-            # Clip gradient norm
-            th.nn.utils.clip_grad_norm_(self.policy.parameters(), self.max_grad_norm)
-            self.policy.optimizer.step()
-
-        # Increase update counter
-        self._n_updates += gradient_steps
-
-        logger.record("train/n_updates", self._n_updates, exclude="tensorboard")
-        logger.record("train/loss", np.mean(losses))
-
-    def predict(
-        self,
-        observation: np.ndarray,
-        state: Optional[np.ndarray] = None,
-        mask: Optional[np.ndarray] = None,
-        deterministic: bool = False,
-    ) -> Tuple[np.ndarray, Optional[np.ndarray]]:
-        """
-        Overrides the base_class predict function to include epsilon-greedy exploration.
-
-        :param observation: the input observation
-        :param state: The last states (can be None, used in recurrent policies)
-        :param mask: The last masks (can be None, used in recurrent policies)
-        :param deterministic: Whether or not to return deterministic actions.
-        :return: the model's action and the next state
-            (used in recurrent policies)
-        """
-        if not deterministic and np.random.rand() < self.exploration_rate:
-            if is_vectorized_observation(maybe_transpose(observation, self.observation_space), self.observation_space):
-                n_batch = observation.shape[0]
-                action = np.array([self.action_space.sample() for _ in range(n_batch)])
-            else:
-                action = np.array(self.action_space.sample())
-        else:
-            action, state = self.policy.predict(observation, state, mask, deterministic)
-        return action, state
-
-    def learn(
-        self,
-        total_timesteps: int,
-        callback: MaybeCallback = None,
-        log_interval: int = 4,
-        eval_env: Optional[GymEnv] = None,
-        eval_freq: int = -1,
-        n_eval_episodes: int = 5,
-        tb_log_name: str = "DQN",
-        eval_log_path: Optional[str] = None,
-        reset_num_timesteps: bool = True,
-    ) -> OffPolicyAlgorithm:
-
-        return super(DQN, self).learn(
-            total_timesteps=total_timesteps,
-            callback=callback,
-            log_interval=log_interval,
-            eval_env=eval_env,
-            eval_freq=eval_freq,
-            n_eval_episodes=n_eval_episodes,
-            tb_log_name=tb_log_name,
-            eval_log_path=eval_log_path,
-            reset_num_timesteps=reset_num_timesteps,
-        )
-
-    def _excluded_save_params(self) -> List[str]:
-        return super(DQN, self)._excluded_save_params() + ["q_net", "q_net_target"]
-
-    def _get_torch_save_params(self) -> Tuple[List[str], List[str]]:
-        state_dicts = ["policy", "policy.optimizer"]
-
-        return state_dicts, []

crazy_functions/test_project/python/dqn/policies.py

@@ -1,237 +0,0 @@
-from typing import Any, Dict, List, Optional, Type
-
-import gym
-import torch as th
-from torch import nn
-
-from stable_baselines3.common.policies import BasePolicy, register_policy
-from stable_baselines3.common.torch_layers import BaseFeaturesExtractor, FlattenExtractor, NatureCNN, create_mlp
-from stable_baselines3.common.type_aliases import Schedule
-
-
-class QNetwork(BasePolicy):
-    """
-    Action-Value (Q-Value) network for DQN
-
-    :param observation_space: Observation space
-    :param action_space: Action space
-    :param net_arch: The specification of the policy and value networks.
-    :param activation_fn: Activation function
-    :param normalize_images: Whether to normalize images or not,
-         dividing by 255.0 (True by default)
-    """
-
-    def __init__(
-        self,
-        observation_space: gym.spaces.Space,
-        action_space: gym.spaces.Space,
-        features_extractor: nn.Module,
-        features_dim: int,
-        net_arch: Optional[List[int]] = None,
-        activation_fn: Type[nn.Module] = nn.ReLU,
-        normalize_images: bool = True,
-    ):
-        super(QNetwork, self).__init__(
-            observation_space,
-            action_space,
-            features_extractor=features_extractor,
-            normalize_images=normalize_images,
-        )
-
-        if net_arch is None:
-            net_arch = [64, 64]
-
-        self.net_arch = net_arch
-        self.activation_fn = activation_fn
-        self.features_extractor = features_extractor
-        self.features_dim = features_dim
-        self.normalize_images = normalize_images
-        action_dim = self.action_space.n  # number of actions
-        q_net = create_mlp(self.features_dim, action_dim, self.net_arch, self.activation_fn)
-        self.q_net = nn.Sequential(*q_net)
-
-    def forward(self, obs: th.Tensor) -> th.Tensor:
-        """
-        Predict the q-values.
-
-        :param obs: Observation
-        :return: The estimated Q-Value for each action.
-        """
-        return self.q_net(self.extract_features(obs))
-
-    def _predict(self, observation: th.Tensor, deterministic: bool = True) -> th.Tensor:
-        q_values = self.forward(observation)
-        # Greedy action
-        action = q_values.argmax(dim=1).reshape(-1)
-        return action
-
-    def _get_constructor_parameters(self) -> Dict[str, Any]:
-        data = super()._get_constructor_parameters()
-
-        data.update(
-            dict(
-                net_arch=self.net_arch,
-                features_dim=self.features_dim,
-                activation_fn=self.activation_fn,
-                features_extractor=self.features_extractor,
-            )
-        )
-        return data
-
-
-class DQNPolicy(BasePolicy):
-    """
-    Policy class with Q-Value Net and target net for DQN
-
-    :param observation_space: Observation space
-    :param action_space: Action space
-    :param lr_schedule: Learning rate schedule (could be constant)
-    :param net_arch: The specification of the policy and value networks.
-    :param activation_fn: Activation function
-    :param features_extractor_class: Features extractor to use.
-    :param features_extractor_kwargs: Keyword arguments
-        to pass to the features extractor.
-    :param normalize_images: Whether to normalize images or not,
-         dividing by 255.0 (True by default)
-    :param optimizer_class: The optimizer to use,
-        ``th.optim.Adam`` by default
-    :param optimizer_kwargs: Additional keyword arguments,
-        excluding the learning rate, to pass to the optimizer
-    """
-
-    def __init__(
-        self,
-        observation_space: gym.spaces.Space,
-        action_space: gym.spaces.Space,
-        lr_schedule: Schedule,
-        net_arch: Optional[List[int]] = None,
-        activation_fn: Type[nn.Module] = nn.ReLU,
-        features_extractor_class: Type[BaseFeaturesExtractor] = FlattenExtractor,
-        features_extractor_kwargs: Optional[Dict[str, Any]] = None,
-        normalize_images: bool = True,
-        optimizer_class: Type[th.optim.Optimizer] = th.optim.Adam,
-        optimizer_kwargs: Optional[Dict[str, Any]] = None,
-    ):
-        super(DQNPolicy, self).__init__(
-            observation_space,
-            action_space,
-            features_extractor_class,
-            features_extractor_kwargs,
-            optimizer_class=optimizer_class,
-            optimizer_kwargs=optimizer_kwargs,
-        )
-
-        if net_arch is None:
-            if features_extractor_class == FlattenExtractor:
-                net_arch = [64, 64]
-            else:
-                net_arch = []
-
-        self.net_arch = net_arch
-        self.activation_fn = activation_fn
-        self.normalize_images = normalize_images
-
-        self.net_args = {
-            "observation_space": self.observation_space,
-            "action_space": self.action_space,
-            "net_arch": self.net_arch,
-            "activation_fn": self.activation_fn,
-            "normalize_images": normalize_images,
-        }
-
-        self.q_net, self.q_net_target = None, None
-        self._build(lr_schedule)
-
-    def _build(self, lr_schedule: Schedule) -> None:
-        """
-        Create the network and the optimizer.
-
-        :param lr_schedule: Learning rate schedule
-            lr_schedule(1) is the initial learning rate
-        """
-
-        self.q_net = self.make_q_net()
-        self.q_net_target = self.make_q_net()
-        self.q_net_target.load_state_dict(self.q_net.state_dict())
-
-        # Setup optimizer with initial learning rate
-        self.optimizer = self.optimizer_class(self.parameters(), lr=lr_schedule(1), **self.optimizer_kwargs)
-
-    def make_q_net(self) -> QNetwork:
-        # Make sure we always have separate networks for features extractors etc
-        net_args = self._update_features_extractor(self.net_args, features_extractor=None)
-        return QNetwork(**net_args).to(self.device)
-
-    def forward(self, obs: th.Tensor, deterministic: bool = True) -> th.Tensor:
-        return self._predict(obs, deterministic=deterministic)
-
-    def _predict(self, obs: th.Tensor, deterministic: bool = True) -> th.Tensor:
-        return self.q_net._predict(obs, deterministic=deterministic)
-
-    def _get_constructor_parameters(self) -> Dict[str, Any]:
-        data = super()._get_constructor_parameters()
-
-        data.update(
-            dict(
-                net_arch=self.net_args["net_arch"],
-                activation_fn=self.net_args["activation_fn"],
-                lr_schedule=self._dummy_schedule,  # dummy lr schedule, not needed for loading policy alone
-                optimizer_class=self.optimizer_class,
-                optimizer_kwargs=self.optimizer_kwargs,
-                features_extractor_class=self.features_extractor_class,
-                features_extractor_kwargs=self.features_extractor_kwargs,
-            )
-        )
-        return data
-
-
-MlpPolicy = DQNPolicy
-
-
-class CnnPolicy(DQNPolicy):
-    """
-    Policy class for DQN when using images as input.
-
-    :param observation_space: Observation space
-    :param action_space: Action space
-    :param lr_schedule: Learning rate schedule (could be constant)
-    :param net_arch: The specification of the policy and value networks.
-    :param activation_fn: Activation function
-    :param features_extractor_class: Features extractor to use.
-    :param normalize_images: Whether to normalize images or not,
-         dividing by 255.0 (True by default)
-    :param optimizer_class: The optimizer to use,
-        ``th.optim.Adam`` by default
-    :param optimizer_kwargs: Additional keyword arguments,
-        excluding the learning rate, to pass to the optimizer
-    """
-
-    def __init__(
-        self,
-        observation_space: gym.spaces.Space,
-        action_space: gym.spaces.Space,
-        lr_schedule: Schedule,
-        net_arch: Optional[List[int]] = None,
-        activation_fn: Type[nn.Module] = nn.ReLU,
-        features_extractor_class: Type[BaseFeaturesExtractor] = NatureCNN,
-        features_extractor_kwargs: Optional[Dict[str, Any]] = None,
-        normalize_images: bool = True,
-        optimizer_class: Type[th.optim.Optimizer] = th.optim.Adam,
-        optimizer_kwargs: Optional[Dict[str, Any]] = None,
-    ):
-        super(CnnPolicy, self).__init__(
-            observation_space,
-            action_space,
-            lr_schedule,
-            net_arch,
-            activation_fn,
-            features_extractor_class,
-            features_extractor_kwargs,
-            normalize_images,
-            optimizer_class,
-            optimizer_kwargs,
-        )
-
-
-register_policy("MlpPolicy", MlpPolicy)
-register_policy("CnnPolicy", CnnPolicy)

crazy_functions/test_project/python/dqn/来源

@@ -1,2 +0,0 @@
-github stablebaseline3
-https://github.com/DLR-RM/stable-baselines3

crazy_functions/test_project/其他测试

@@ -1,27 +0,0 @@
-"In practice, we found that a high-entropy initial state is more likely to increase the speed of training.
-The entropy is calculated by:
-$$H=-\sum_{k= 1}^{n_k} p(k) \cdot \log p(k), p(k)=\frac{|A_k|}{|\mathcal{A}|}$$
-where $H$ is the entropy, $|A_k|$ is the number of agent nodes in $k$-th cluster, $|\mathcal{A}|$ is the total number of agents.
-To ensure the Cooperation Graph initialization has higher entropy,
-we will randomly generate multiple initial states, 
-rank by their entropy and then pick the one with maximum $H$."
-
-```
-FROM ubuntu:latest
-
-RUN apt-get update && \
-    apt-get install -y python3 python3-pip && \
-    rm -rf /var/lib/apt/lists/*
-
-RUN echo '[global]' > /etc/pip.conf && \
-    echo 'index-url = https://mirrors.aliyun.com/pypi/simple/' >> /etc/pip.conf && \
-    echo 'trusted-host = mirrors.aliyun.com' >> /etc/pip.conf
-
-RUN pip3 install gradio requests[socks] mdtex2html
-
-COPY . /gpt
-WORKDIR /gpt
-
-
-CMD ["python3", "main.py"]
-```

crazy_functions/vector_fns/general_file_loader.py

@@ -0,0 +1,70 @@
+# From project chatglm-langchain
+
+
+from langchain.document_loaders import UnstructuredFileLoader
+from langchain.text_splitter import CharacterTextSplitter
+import re
+from typing import List
+
+class ChineseTextSplitter(CharacterTextSplitter):
+    def __init__(self, pdf: bool = False, sentence_size: int = None, **kwargs):
+        super().__init__(**kwargs)
+        self.pdf = pdf
+        self.sentence_size = sentence_size
+
+    def split_text1(self, text: str) -> List[str]:
+        if self.pdf:
+            text = re.sub(r"\n{3,}", "\n", text)
+            text = re.sub('\s', ' ', text)
+            text = text.replace("\n\n", "")
+        sent_sep_pattern = re.compile('([﹒﹔﹖﹗．。！？]["’”」』]{0,2}|(?=["‘“「『]{1,2}|$))')  # del ：；
+        sent_list = []
+        for ele in sent_sep_pattern.split(text):
+            if sent_sep_pattern.match(ele) and sent_list:
+                sent_list[-1] += ele
+            elif ele:
+                sent_list.append(ele)
+        return sent_list
+
+    def split_text(self, text: str) -> List[str]:   ##此处需要进一步优化逻辑
+        if self.pdf:
+            text = re.sub(r"\n{3,}", r"\n", text)
+            text = re.sub('\s', " ", text)
+            text = re.sub("\n\n", "", text)
+
+        text = re.sub(r'([;；.!?。！？\?])([^”’])', r"\1\n\2", text)  # 单字符断句符
+        text = re.sub(r'(\.{6})([^"’”」』])', r"\1\n\2", text)  # 英文省略号
+        text = re.sub(r'(\…{2})([^"’”」』])', r"\1\n\2", text)  # 中文省略号
+        text = re.sub(r'([;；!?。！？\?]["’”」』]{0,2})([^;；!?，。！？\?])', r'\1\n\2', text)
+        # 如果双引号前有终止符，那么双引号才是句子的终点，把分句符\n放到双引号后，注意前面的几句都小心保留了双引号
+        text = text.rstrip()  # 段尾如果有多余的\n就去掉它
+        # 很多规则中会考虑分号;，但是这里我把它忽略不计，破折号、英文双引号等同样忽略，需要的再做些简单调整即可。
+        ls = [i for i in text.split("\n") if i]
+        for ele in ls:
+            if len(ele) > self.sentence_size:
+                ele1 = re.sub(r'([,，.]["’”」』]{0,2})([^,，.])', r'\1\n\2', ele)
+                ele1_ls = ele1.split("\n")
+                for ele_ele1 in ele1_ls:
+                    if len(ele_ele1) > self.sentence_size:
+                        ele_ele2 = re.sub(r'([\n]{1,}| {2,}["’”」』]{0,2})([^\s])', r'\1\n\2', ele_ele1)
+                        ele2_ls = ele_ele2.split("\n")
+                        for ele_ele2 in ele2_ls:
+                            if len(ele_ele2) > self.sentence_size:
+                                ele_ele3 = re.sub('( ["’”」』]{0,2})([^ ])', r'\1\n\2', ele_ele2)
+                                ele2_id = ele2_ls.index(ele_ele2)
+                                ele2_ls = ele2_ls[:ele2_id] + [i for i in ele_ele3.split("\n") if i] + ele2_ls[
+                                                                                                       ele2_id + 1:]
+                        ele_id = ele1_ls.index(ele_ele1)
+                        ele1_ls = ele1_ls[:ele_id] + [i for i in ele2_ls if i] + ele1_ls[ele_id + 1:]
+
+                id = ls.index(ele)
+                ls = ls[:id] + [i for i in ele1_ls if i] + ls[id + 1:]
+        return ls
+
+def load_file(filepath, sentence_size):
+    loader = UnstructuredFileLoader(filepath, mode="elements")
+    textsplitter = ChineseTextSplitter(pdf=False, sentence_size=sentence_size)
+    docs = loader.load_and_split(text_splitter=textsplitter)
+    # write_check_file(filepath, docs)
+    return docs
+

crazy_functions/vector_fns/vector_database.py

@@ -0,0 +1,339 @@
+# From project chatglm-langchain
+
+import os
+import os
+import uuid
+import tqdm
+import shutil
+import threading
+import numpy as np
+from toolbox import Singleton
+from loguru import logger
+from langchain.vectorstores import FAISS
+from langchain.docstore.document import Document
+from typing import List, Tuple
+from crazy_functions.vector_fns.general_file_loader import load_file
+
+embedding_model_dict = {
+    "ernie-tiny": "nghuyong/ernie-3.0-nano-zh",
+    "ernie-base": "nghuyong/ernie-3.0-base-zh",
+    "text2vec-base": "shibing624/text2vec-base-chinese",
+    "text2vec": "GanymedeNil/text2vec-large-chinese",
+}
+
+# Embedding model name
+EMBEDDING_MODEL = "text2vec"
+
+# Embedding running device
+EMBEDDING_DEVICE = "cpu"
+
+# 基于上下文的prompt模版，请务必保留"{question}"和"{context}"
+PROMPT_TEMPLATE = """已知信息：
+{context}
+
+根据上述已知信息，简洁和专业的来回答用户的问题。如果无法从中得到答案，请说 “根据已知信息无法回答该问题” 或 “没有提供足够的相关信息”，不允许在答案中添加编造成分，答案请使用中文。 问题是：{question}"""
+
+# 文本分句长度
+SENTENCE_SIZE = 100
+
+# 匹配后单段上下文长度
+CHUNK_SIZE = 250
+
+# LLM input history length
+LLM_HISTORY_LEN = 3
+
+# return top-k text chunk from vector store
+VECTOR_SEARCH_TOP_K = 5
+
+# 知识检索内容相关度 Score, 数值范围约为0-1100，如果为0，则不生效，经测试设置为小于500时，匹配结果更精准
+VECTOR_SEARCH_SCORE_THRESHOLD = 0
+
+NLTK_DATA_PATH = os.path.join(os.path.dirname(os.path.dirname(__file__)), "nltk_data")
+
+FLAG_USER_NAME = uuid.uuid4().hex
+
+# 是否开启跨域，默认为False，如果需要开启，请设置为True
+# is open cross domain
+OPEN_CROSS_DOMAIN = False
+
+def similarity_search_with_score_by_vector(
+        self, embedding: List[float], k: int = 4
+) -> List[Tuple[Document, float]]:
+
+    def seperate_list(ls: List[int]) -> List[List[int]]:
+        lists = []
+        ls1 = [ls[0]]
+        for i in range(1, len(ls)):
+            if ls[i - 1] + 1 == ls[i]:
+                ls1.append(ls[i])
+            else:
+                lists.append(ls1)
+                ls1 = [ls[i]]
+        lists.append(ls1)
+        return lists
+
+    scores, indices = self.index.search(np.array([embedding], dtype=np.float32), k)
+    docs = []
+    id_set = set()
+    store_len = len(self.index_to_docstore_id)
+    for j, i in enumerate(indices[0]):
+        if i == -1 or 0 < self.score_threshold < scores[0][j]:
+            # This happens when not enough docs are returned.
+            continue
+        _id = self.index_to_docstore_id[i]
+        doc = self.docstore.search(_id)
+        if not self.chunk_conent:
+            if not isinstance(doc, Document):
+                raise ValueError(f"Could not find document for id {_id}, got {doc}")
+            doc.metadata["score"] = int(scores[0][j])
+            docs.append(doc)
+            continue
+        id_set.add(i)
+        docs_len = len(doc.page_content)
+        for k in range(1, max(i, store_len - i)):
+            break_flag = False
+            for l in [i + k, i - k]:
+                if 0 <= l < len(self.index_to_docstore_id):
+                    _id0 = self.index_to_docstore_id[l]
+                    doc0 = self.docstore.search(_id0)
+                    if docs_len + len(doc0.page_content) > self.chunk_size:
+                        break_flag = True
+                        break
+                    elif doc0.metadata["source"] == doc.metadata["source"]:
+                        docs_len += len(doc0.page_content)
+                        id_set.add(l)
+            if break_flag:
+                break
+    if not self.chunk_conent:
+        return docs
+    if len(id_set) == 0 and self.score_threshold > 0:
+        return []
+    id_list = sorted(list(id_set))
+    id_lists = seperate_list(id_list)
+    for id_seq in id_lists:
+        for id in id_seq:
+            if id == id_seq[0]:
+                _id = self.index_to_docstore_id[id]
+                doc = self.docstore.search(_id)
+            else:
+                _id0 = self.index_to_docstore_id[id]
+                doc0 = self.docstore.search(_id0)
+                doc.page_content += " " + doc0.page_content
+        if not isinstance(doc, Document):
+            raise ValueError(f"Could not find document for id {_id}, got {doc}")
+        doc_score = min([scores[0][id] for id in [indices[0].tolist().index(i) for i in id_seq if i in indices[0]]])
+        doc.metadata["score"] = int(doc_score)
+        docs.append(doc)
+    return docs
+
+
+class LocalDocQA:
+    llm: object = None
+    embeddings: object = None
+    top_k: int = VECTOR_SEARCH_TOP_K
+    chunk_size: int = CHUNK_SIZE
+    chunk_conent: bool = True
+    score_threshold: int = VECTOR_SEARCH_SCORE_THRESHOLD
+
+    def init_cfg(self,
+                 top_k=VECTOR_SEARCH_TOP_K,
+                 ):
+
+        self.llm = None
+        self.top_k = top_k
+
+    def init_knowledge_vector_store(self,
+                                    filepath,
+                                    vs_path: str or os.PathLike = None,
+                                    sentence_size=SENTENCE_SIZE,
+                                    text2vec=None):
+        loaded_files = []
+        failed_files = []
+        if isinstance(filepath, str):
+            if not os.path.exists(filepath):
+                logger.error("路径不存在")
+                return None
+            elif os.path.isfile(filepath):
+                file = os.path.split(filepath)[-1]
+                try:
+                    docs = load_file(filepath, SENTENCE_SIZE)
+                    logger.info(f"{file} 已成功加载")
+                    loaded_files.append(filepath)
+                except Exception as e:
+                    logger.error(e)
+                    logger.error(f"{file} 未能成功加载")
+                    return None
+            elif os.path.isdir(filepath):
+                docs = []
+                for file in tqdm(os.listdir(filepath), desc="加载文件"):
+                    fullfilepath = os.path.join(filepath, file)
+                    try:
+                        docs += load_file(fullfilepath, SENTENCE_SIZE)
+                        loaded_files.append(fullfilepath)
+                    except Exception as e:
+                        logger.error(e)
+                        failed_files.append(file)
+
+                if len(failed_files) > 0:
+                    logger.error("以下文件未能成功加载：")
+                    for file in failed_files:
+                        logger.error(f"{file}\n")
+
+        else:
+            docs = []
+            for file in filepath:
+                docs += load_file(file, SENTENCE_SIZE)
+                logger.info(f"{file} 已成功加载")
+                loaded_files.append(file)
+
+        if len(docs) > 0:
+            logger.info("文件加载完毕，正在生成向量库")
+            if vs_path and os.path.isdir(vs_path):
+                try:
+                    self.vector_store = FAISS.load_local(vs_path, text2vec)
+                    self.vector_store.add_documents(docs)
+                except:
+                    self.vector_store = FAISS.from_documents(docs, text2vec)
+            else:
+                self.vector_store = FAISS.from_documents(docs, text2vec)  # docs 为Document列表
+
+            self.vector_store.save_local(vs_path)
+            return vs_path, loaded_files
+        else:
+            raise RuntimeError("文件加载失败，请检查文件格式是否正确")
+
+    def get_loaded_file(self, vs_path):
+        ds = self.vector_store.docstore
+        return set([ds._dict[k].metadata['source'].split(vs_path)[-1] for k in ds._dict])
+
+
+    # query      查询内容
+    # vs_path    知识库路径
+    # chunk_conent   是否启用上下文关联
+    # score_threshold    搜索匹配score阈值
+    # vector_search_top_k   搜索知识库内容条数，默认搜索5条结果
+    # chunk_sizes    匹配单段内容的连接上下文长度
+    def get_knowledge_based_conent_test(self, query, vs_path, chunk_conent,
+                                        score_threshold=VECTOR_SEARCH_SCORE_THRESHOLD,
+                                        vector_search_top_k=VECTOR_SEARCH_TOP_K, chunk_size=CHUNK_SIZE,
+                                        text2vec=None):
+        self.vector_store = FAISS.load_local(vs_path, text2vec)
+        self.vector_store.chunk_conent = chunk_conent
+        self.vector_store.score_threshold = score_threshold
+        self.vector_store.chunk_size = chunk_size
+
+        embedding = self.vector_store.embedding_function.embed_query(query)
+        related_docs_with_score = similarity_search_with_score_by_vector(self.vector_store, embedding, k=vector_search_top_k)
+
+        if not related_docs_with_score:
+            response = {"query": query,
+                        "source_documents": []}
+            return response, ""
+        # prompt = f"{query}. You should answer this question using information from following documents: \n\n"
+        prompt = f"{query}. 你必须利用以下文档中包含的信息回答这个问题: \n\n---\n\n"
+        prompt += "\n\n".join([f"({k}): " + doc.page_content for k, doc in enumerate(related_docs_with_score)])
+        prompt += "\n\n---\n\n"
+        prompt = prompt.encode('utf-8', 'ignore').decode()   # avoid reading non-utf8 chars
+        # logger.info(prompt)
+        response = {"query": query, "source_documents": related_docs_with_score}
+        return response, prompt
+
+
+
+
+def construct_vector_store(vs_id, vs_path, files, sentence_size, history, one_conent, one_content_segmentation, text2vec):
+    for file in files:
+        assert os.path.exists(file), "输入文件不存在：" + file
+    import nltk
+    if NLTK_DATA_PATH not in nltk.data.path: nltk.data.path = [NLTK_DATA_PATH] + nltk.data.path
+    local_doc_qa = LocalDocQA()
+    local_doc_qa.init_cfg()
+    filelist = []
+    if not os.path.exists(os.path.join(vs_path, vs_id)):
+        os.makedirs(os.path.join(vs_path, vs_id))
+    for file in files:
+        file_name = file.name if not isinstance(file, str) else file
+        filename = os.path.split(file_name)[-1]
+        shutil.copyfile(file_name, os.path.join(vs_path, vs_id, filename))
+        filelist.append(os.path.join(vs_path, vs_id, filename))
+    vs_path, loaded_files = local_doc_qa.init_knowledge_vector_store(filelist, os.path.join(vs_path, vs_id), sentence_size, text2vec)
+
+    if len(loaded_files):
+        file_status = f"已添加 {'、'.join([os.path.split(i)[-1] for i in loaded_files if i])} 内容至知识库，并已加载知识库，请开始提问"
+    else:
+        pass
+        # file_status = "文件未成功加载，请重新上传文件"
+    # logger.info(file_status)
+    return local_doc_qa, vs_path
+
+@Singleton
+class knowledge_archive_interface():
+    def __init__(self) -> None:
+        self.threadLock = threading.Lock()
+        self.current_id = ""
+        self.kai_path = None
+        self.qa_handle = None
+        self.text2vec_large_chinese = None
+
+    def get_chinese_text2vec(self):
+        if self.text2vec_large_chinese is None:
+            # < -------------------预热文本向量化模组--------------- >
+            from toolbox import ProxyNetworkActivate
+            logger.info('Checking Text2vec ...')
+            from langchain.embeddings.huggingface import HuggingFaceEmbeddings
+            with ProxyNetworkActivate('Download_LLM'):    # 临时地激活代理网络
+                self.text2vec_large_chinese = HuggingFaceEmbeddings(model_name="GanymedeNil/text2vec-large-chinese")
+
+        return self.text2vec_large_chinese
+
+
+    def feed_archive(self, file_manifest, vs_path, id="default"):
+        self.threadLock.acquire()
+        # import uuid
+        self.current_id = id
+        self.qa_handle, self.kai_path = construct_vector_store(
+            vs_id=self.current_id,
+            vs_path=vs_path,
+            files=file_manifest,
+            sentence_size=100,
+            history=[],
+            one_conent="",
+            one_content_segmentation="",
+            text2vec = self.get_chinese_text2vec(),
+        )
+        self.threadLock.release()
+
+    def get_current_archive_id(self):
+        return self.current_id
+
+    def get_loaded_file(self, vs_path):
+        return self.qa_handle.get_loaded_file(vs_path)
+
+    def answer_with_archive_by_id(self, txt, id, vs_path):
+        self.threadLock.acquire()
+        if not self.current_id == id:
+            self.current_id = id
+            self.qa_handle, self.kai_path = construct_vector_store(
+                vs_id=self.current_id,
+                vs_path=vs_path,
+                files=[],
+                sentence_size=100,
+                history=[],
+                one_conent="",
+                one_content_segmentation="",
+                text2vec = self.get_chinese_text2vec(),
+            )
+        VECTOR_SEARCH_SCORE_THRESHOLD = 0
+        VECTOR_SEARCH_TOP_K = 4
+        CHUNK_SIZE = 512
+        resp, prompt = self.qa_handle.get_knowledge_based_conent_test(
+            query = txt,
+            vs_path = self.kai_path,
+            score_threshold=VECTOR_SEARCH_SCORE_THRESHOLD,
+            vector_search_top_k=VECTOR_SEARCH_TOP_K,
+            chunk_conent=True,
+            chunk_size=CHUNK_SIZE,
+            text2vec = self.get_chinese_text2vec(),
+        )
+        self.threadLock.release()
+        return resp, prompt

crazy_functions/vt_fns/vt_call_plugin.py

@@ -0,0 +1,114 @@
+from pydantic import BaseModel, Field
+from typing import List
+from toolbox import update_ui_lastest_msg, disable_auto_promotion
+from request_llms.bridge_all import predict_no_ui_long_connection
+from crazy_functions.json_fns.pydantic_io import GptJsonIO, JsonStringError
+import copy, json, pickle, os, sys, time
+
+
+def read_avail_plugin_enum():
+    from crazy_functional import get_crazy_functions
+    plugin_arr = get_crazy_functions()
+    # remove plugins with out explaination
+    plugin_arr = {k:v for k, v in plugin_arr.items() if ('Info' in v) and ('Function' in v)}
+    plugin_arr_info = {"F_{:04d}".format(i):v["Info"] for i, v in enumerate(plugin_arr.values(), start=1)}
+    plugin_arr_dict = {"F_{:04d}".format(i):v for i, v in enumerate(plugin_arr.values(), start=1)}
+    plugin_arr_dict_parse = {"F_{:04d}".format(i):v for i, v in enumerate(plugin_arr.values(), start=1)}
+    plugin_arr_dict_parse.update({f"F_{i}":v for i, v in enumerate(plugin_arr.values(), start=1)})
+    prompt = json.dumps(plugin_arr_info, ensure_ascii=False, indent=2)
+    prompt = "\n\nThe defination of PluginEnum:\nPluginEnum=" + prompt
+    return prompt, plugin_arr_dict, plugin_arr_dict_parse
+
+def wrap_code(txt):
+    txt = txt.replace('```','')
+    return f"\n```\n{txt}\n```\n"
+
+def have_any_recent_upload_files(chatbot):
+    _5min = 5 * 60
+    if not chatbot: return False    # chatbot is None
+    most_recent_uploaded = chatbot._cookies.get("most_recent_uploaded", None)
+    if not most_recent_uploaded: return False   # most_recent_uploaded is None
+    if time.time() - most_recent_uploaded["time"] < _5min: return True # most_recent_uploaded is new
+    else: return False  # most_recent_uploaded is too old
+
+def get_recent_file_prompt_support(chatbot):
+    most_recent_uploaded = chatbot._cookies.get("most_recent_uploaded", None)
+    path = most_recent_uploaded['path']
+    prompt =   "\nAdditional Information:\n"
+    prompt =   "In case that this plugin requires a path or a file as argument,"
+    prompt += f"it is important for you to know that the user has recently uploaded a file, located at: `{path}`"
+    prompt += f"Only use it when necessary, otherwise, you can ignore this file."
+    return prompt
+
+def get_inputs_show_user(inputs, plugin_arr_enum_prompt):
+    # remove plugin_arr_enum_prompt from inputs string
+    inputs_show_user = inputs.replace(plugin_arr_enum_prompt, "")
+    inputs_show_user += plugin_arr_enum_prompt[:200] + '...'
+    inputs_show_user += '\n...\n'
+    inputs_show_user += '...\n'
+    inputs_show_user += '...}'
+    return inputs_show_user
+
+def execute_plugin(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_intention):
+    plugin_arr_enum_prompt, plugin_arr_dict, plugin_arr_dict_parse = read_avail_plugin_enum()
+    class Plugin(BaseModel):
+        plugin_selection: str = Field(description="The most related plugin from one of the PluginEnum.", default="F_0000")
+        reason_of_selection: str = Field(description="The reason why you should select this plugin.", default="This plugin satisfy user requirement most")
+    # ⭐ ⭐ ⭐ 选择插件
+    yield from update_ui_lastest_msg(lastmsg=f"正在执行任务: {txt}\n\n查找可用插件中...", chatbot=chatbot, history=history, delay=0)
+    gpt_json_io = GptJsonIO(Plugin)
+    gpt_json_io.format_instructions = "The format of your output should be a json that can be parsed by json.loads.\n"
+    gpt_json_io.format_instructions += """Output example: {"plugin_selection":"F_1234", "reason_of_selection":"F_1234 plugin satisfy user requirement most"}\n"""
+    gpt_json_io.format_instructions += "The plugins you are authorized to use are listed below:\n"
+    gpt_json_io.format_instructions += plugin_arr_enum_prompt
+    inputs = "Choose the correct plugin according to user requirements, the user requirement is: \n\n" + \
+             ">> " + txt.rstrip('\n').replace('\n','\n>> ') + '\n\n' + gpt_json_io.format_instructions
+
+    run_gpt_fn = lambda inputs, sys_prompt: predict_no_ui_long_connection(
+        inputs=inputs, llm_kwargs=llm_kwargs, history=[], sys_prompt=sys_prompt, observe_window=[])
+    try:
+        gpt_reply = run_gpt_fn(inputs, "")
+        plugin_sel = gpt_json_io.generate_output_auto_repair(gpt_reply, run_gpt_fn)
+    except JsonStringError:
+        msg = f"抱歉, {llm_kwargs['llm_model']}无法理解您的需求。"
+        msg += "请求的Prompt为：\n" + wrap_code(get_inputs_show_user(inputs, plugin_arr_enum_prompt))
+        msg += "语言模型回复为：\n" + wrap_code(gpt_reply)
+        msg += "\n但您可以尝试再试一次\n"
+        yield from update_ui_lastest_msg(lastmsg=msg, chatbot=chatbot, history=history, delay=2)
+        return
+    if plugin_sel.plugin_selection not in plugin_arr_dict_parse:
+        msg = f"抱歉, 找不到合适插件执行该任务, 或者{llm_kwargs['llm_model']}无法理解您的需求。"
+        msg += f"语言模型{llm_kwargs['llm_model']}选择了不存在的插件：\n" + wrap_code(gpt_reply)
+        msg += "\n但您可以尝试再试一次\n"
+        yield from update_ui_lastest_msg(lastmsg=msg, chatbot=chatbot, history=history, delay=2)
+        return
+
+    # ⭐ ⭐ ⭐ 确认插件参数
+    if not have_any_recent_upload_files(chatbot):
+        appendix_info = ""
+    else:
+        appendix_info = get_recent_file_prompt_support(chatbot)
+
+    plugin = plugin_arr_dict_parse[plugin_sel.plugin_selection]
+    yield from update_ui_lastest_msg(lastmsg=f"正在执行任务: {txt}\n\n提取插件参数...", chatbot=chatbot, history=history, delay=0)
+    class PluginExplicit(BaseModel):
+        plugin_selection: str = plugin_sel.plugin_selection
+        plugin_arg: str = Field(description="The argument of the plugin.", default="")
+    gpt_json_io = GptJsonIO(PluginExplicit)
+    gpt_json_io.format_instructions += "The information about this plugin is:" + plugin["Info"]
+    inputs = f"A plugin named {plugin_sel.plugin_selection} is selected, " + \
+             "you should extract plugin_arg from the user requirement, the user requirement is: \n\n" + \
+             ">> " + (txt + appendix_info).rstrip('\n').replace('\n','\n>> ') + '\n\n' + \
+             gpt_json_io.format_instructions
+    run_gpt_fn = lambda inputs, sys_prompt: predict_no_ui_long_connection(
+        inputs=inputs, llm_kwargs=llm_kwargs, history=[], sys_prompt=sys_prompt, observe_window=[])
+    plugin_sel = gpt_json_io.generate_output_auto_repair(run_gpt_fn(inputs, ""), run_gpt_fn)
+
+
+    # ⭐ ⭐ ⭐ 执行插件
+    fn = plugin['Function']
+    fn_name = fn.__name__
+    msg = f'{llm_kwargs["llm_model"]}为您选择了插件: `{fn_name}`\n\n插件说明：{plugin["Info"]}\n\n插件参数：{plugin_sel.plugin_arg}\n\n假如偏离了您的要求，按停止键终止。'
+    yield from update_ui_lastest_msg(lastmsg=msg, chatbot=chatbot, history=history, delay=2)
+    yield from fn(plugin_sel.plugin_arg, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, -1)
+    return

crazy_functions/vt_fns/vt_modify_config.py

@@ -0,0 +1,81 @@
+from pydantic import BaseModel, Field
+from typing import List
+from toolbox import update_ui_lastest_msg, get_conf
+from request_llms.bridge_all import predict_no_ui_long_connection
+from crazy_functions.json_fns.pydantic_io import GptJsonIO
+import copy, json, pickle, os, sys
+
+
+def modify_configuration_hot(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_intention):
+    ALLOW_RESET_CONFIG = get_conf('ALLOW_RESET_CONFIG')
+    if not ALLOW_RESET_CONFIG:
+        yield from update_ui_lastest_msg(
+            lastmsg=f"当前配置不允许被修改！如需激活本功能，请在config.py中设置ALLOW_RESET_CONFIG=True后重启软件。",
+            chatbot=chatbot, history=history, delay=2
+        )
+        return
+
+    # ⭐ ⭐ ⭐ 读取可配置项目条目
+    names = {}
+    from enum import Enum
+    import config
+    for k, v in config.__dict__.items():
+        if k.startswith('__'): continue
+        names.update({k:k})
+        # if len(names) > 20: break   # 限制最多前10个配置项，如果太多了会导致gpt无法理解
+
+    ConfigOptions = Enum('ConfigOptions', names)
+    class ModifyConfigurationIntention(BaseModel):
+        which_config_to_modify: ConfigOptions = Field(description="the name of the configuration to modify, you must choose from one of the ConfigOptions enum.", default=None)
+        new_option_value: str = Field(description="the new value of the option", default=None)
+
+    # ⭐ ⭐ ⭐ 分析用户意图
+    yield from update_ui_lastest_msg(lastmsg=f"正在执行任务: {txt}\n\n读取新配置中", chatbot=chatbot, history=history, delay=0)
+    gpt_json_io = GptJsonIO(ModifyConfigurationIntention)
+    inputs = "Analyze how to change configuration according to following user input, answer me with json: \n\n" + \
+             ">> " + txt.rstrip('\n').replace('\n','\n>> ') + '\n\n' + \
+             gpt_json_io.format_instructions
+
+    run_gpt_fn = lambda inputs, sys_prompt: predict_no_ui_long_connection(
+        inputs=inputs, llm_kwargs=llm_kwargs, history=[], sys_prompt=sys_prompt, observe_window=[])
+    user_intention = gpt_json_io.generate_output_auto_repair(run_gpt_fn(inputs, ""), run_gpt_fn)
+
+    explicit_conf = user_intention.which_config_to_modify.value
+
+    ok = (explicit_conf in txt)
+    if ok:
+        yield from update_ui_lastest_msg(
+            lastmsg=f"正在执行任务: {txt}\n\n新配置{explicit_conf}={user_intention.new_option_value}",
+            chatbot=chatbot, history=history, delay=1
+        )
+        yield from update_ui_lastest_msg(
+            lastmsg=f"正在执行任务: {txt}\n\n新配置{explicit_conf}={user_intention.new_option_value}\n\n正在修改配置中",
+            chatbot=chatbot, history=history, delay=2
+        )
+
+        # ⭐ ⭐ ⭐ 立即应用配置
+        from toolbox import set_conf
+        set_conf(explicit_conf, user_intention.new_option_value)
+
+        yield from update_ui_lastest_msg(
+            lastmsg=f"正在执行任务: {txt}\n\n配置修改完成，重新页面即可生效。", chatbot=chatbot, history=history, delay=1
+        )
+    else:
+        yield from update_ui_lastest_msg(
+            lastmsg=f"失败，如果需要配置{explicit_conf}，您需要明确说明并在指令中提到它。", chatbot=chatbot, history=history, delay=5
+        )
+
+def modify_configuration_reboot(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_intention):
+    ALLOW_RESET_CONFIG = get_conf('ALLOW_RESET_CONFIG')
+    if not ALLOW_RESET_CONFIG:
+        yield from update_ui_lastest_msg(
+            lastmsg=f"当前配置不允许被修改！如需激活本功能，请在config.py中设置ALLOW_RESET_CONFIG=True后重启软件。",
+            chatbot=chatbot, history=history, delay=2
+        )
+        return
+
+    yield from modify_configuration_hot(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, user_intention)
+    yield from update_ui_lastest_msg(
+        lastmsg=f"正在执行任务: {txt}\n\n配置修改完成，五秒后即将重启！若出现报错请无视即可。", chatbot=chatbot, history=history, delay=5
+    )
+    os.execl(sys.executable, sys.executable, *sys.argv)