RYDE-WORK/Langchain-Chatchat
1
0
Fork 0
mirror of https://github.com/RYDE-WORK/Langchain-Chatchat.git synced 2026-02-08 16:10:18 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #2754 from chatchat-space/dev

Browse Source 
更新文档细节和readme细节
This commit is contained in:
zR 2024-01-22 13:44:36 +08:00 committed by GitHub
commit 2c92f426d9
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
35 changed files with 302 additions and 366 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

82
README.md
View File

@ -1,12 +1,13 @@
![](img/logo-long-chatchat-trans-v2.png) ![](img/logo-long-chatchat-trans-v2.png)
🌍 [READ THIS IN ENGLISH](README_en.md) 🌍 [READ THIS IN ENGLISH](README_en.md)
📃 **LangChain-Chatchat** (原 Langchain-ChatGLM) 📃 **LangChain-Chatchat** (原 Langchain-ChatGLM)
基于 ChatGLM 等大语言模型与 Langchain 等应用框架实现，开源、可离线部署的检索增强生成(RAG)大模型知识库项目。 基于 ChatGLM 等大语言模型与 Langchain 等应用框架实现，开源、可离线部署的检索增强生成(RAG)大模型知识库项目。
⚠️`0.2.10`将会是`0.2.x`系列的最后一个版本，`0.2.x`系列版本将会停止更新和技术支持，全力研发具有更强应用性的 `Langchain-Chatchat 0.3.x`。
--- ---
## 目录 ## 目录
@ -14,23 +15,31 @@
* [介绍](README.md#介绍) * [介绍](README.md#介绍)
* [解决的痛点](README.md#解决的痛点) * [解决的痛点](README.md#解决的痛点)
* [快速上手](README.md#快速上手) * [快速上手](README.md#快速上手)
* [1. 环境配置](README.md#1-环境配置) * [1. 环境配置](README.md#1-环境配置)
* [2. 模型下载](README.md#2-模型下载) * [2. 模型下载](README.md#2-模型下载)
* [3. 初始化知识库和配置文件](README.md#3-初始化知识库和配置文件) * [3. 初始化知识库和配置文件](README.md#3-初始化知识库和配置文件)
* [4. 一键启动](README.md#4-一键启动) * [4. 一键启动](README.md#4-一键启动)
* [5. 启动界面示例](README.md#5-启动界面示例) * [5. 启动界面示例](README.md#5-启动界面示例)
* [联系我们](README.md#联系我们) * [联系我们](README.md#联系我们)
## 介绍 ## 介绍
🤖️ 一种利用 [langchain](https://github.com/hwchase17/langchain) 思想实现的基于本地知识库的问答应用，目标期望建立一套对中文场景与开源模型支持友好、可离线运行的知识库问答解决方案。 🤖️ 一种利用 [langchain](https://github.com/langchain-ai/langchain)
思想实现的基于本地知识库的问答应用，目标期望建立一套对中文场景与开源模型支持友好、可离线运行的知识库问答解决方案。
💡 受 [GanymedeNil](https://github.com/GanymedeNil) 的项目 [document.ai](https://github.com/GanymedeNil/document.ai) 和 [AlexZhangji](https://github.com/AlexZhangji) 创建的 [ChatGLM-6B Pull Request](https://github.com/THUDM/ChatGLM-6B/pull/216) 启发，建立了全流程可使用开源模型实现的本地知识库问答应用。本项目的最新版本中通过使用 [FastChat](https://github.com/lm-sys/FastChat) 接入 Vicuna, Alpaca, LLaMA, Koala, RWKV 等模型，依托于 [langchain](https://github.com/langchain-ai/langchain) 框架支持通过基于 [FastAPI](https://github.com/tiangolo/fastapi) 提供的 API 调用服务，或使用基于 [Streamlit](https://github.com/streamlit/streamlit) 的 WebUI 进行操作。 💡 受 [GanymedeNil](https://github.com/GanymedeNil) 的项目 [document.ai](https://github.com/GanymedeNil/document.ai)
和 [AlexZhangji](https://github.com/AlexZhangji)
创建的 [ChatGLM-6B Pull Request](https://github.com/THUDM/ChatGLM-6B/pull/216)
启发，建立了全流程可使用开源模型实现的本地知识库问答应用。本项目的最新版本中通过使用 [FastChat](https://github.com/lm-sys/FastChat)
接入 Vicuna, Alpaca, LLaMA, Koala, RWKV 等模型，依托于 [langchain](https://github.com/langchain-ai/langchain)
框架支持通过基于 [FastAPI](https://github.com/tiangolo/fastapi) 提供的 API
调用服务，或使用基于 [Streamlit](https://github.com/streamlit/streamlit) 的 WebUI 进行操作。
✅ 依托于本项目支持的开源 LLM 与 Embedding 模型，本项目可实现全部使用**开源**模型**离线私有部署**。与此同时，本项目也支持 OpenAI GPT API 的调用，并将在后续持续扩充对各类模型及模型 API 的接入。 ✅ 依托于本项目支持的开源 LLM 与 Embedding 模型，本项目可实现全部使用**开源**模型**离线私有部署**。与此同时，本项目也支持
OpenAI GPT API 的调用，并将在后续持续扩充对各类模型及模型 API 的接入。
⛓️ 本项目实现原理如下图所示，过程包括加载文件 -> 读取文本 -> 文本分割 -> 文本向量化 -> 问句向量化 -> 在文本向量中匹配出与问句向量最相似的 `top k`个 -> 匹配出的文本作为上下文和问题一起添加到 `prompt`中 -> 提交给 `LLM`生成回答。 ⛓️ 本项目实现原理如下图所示，过程包括加载文件 -> 读取文本 -> 文本分割 -> 文本向量化 -> 问句向量化 ->
在文本向量中匹配出与问句向量最相似的 `top k`个 -> 匹配出的文本作为上下文和问题一起添加到 `prompt`中 -> 提交给 `LLM`生成回答。
📺 [原理介绍视频](https://www.bilibili.com/video/BV13M4y1e7cN/?share_source=copy_web&vd_source=e6c5aafe684f30fbe41925d61ca6d514) 📺 [原理介绍视频](https://www.bilibili.com/video/BV13M4y1e7cN/?share_source=copy_web&vd_source=e6c5aafe684f30fbe41925d61ca6d514)
@ -42,7 +51,8 @@
🚩 本项目未涉及微调、训练过程，但可利用微调或训练对本项目效果进行优化。 🚩 本项目未涉及微调、训练过程，但可利用微调或训练对本项目效果进行优化。
🌐 [AutoDL 镜像](https://www.codewithgpu.com/i/chatchat-space/Langchain-Chatchat/Langchain-Chatchat) 中 `v13` 版本所使用代码已更新至本项目 `v0.2.9` 版本。 🌐 [AutoDL 镜像](https://www.codewithgpu.com/i/chatchat-space/Langchain-Chatchat/Langchain-Chatchat) 中 `v13`
版本所使用代码已更新至本项目 `v0.2.9` 版本。
🐳 [Docker 镜像](registry.cn-beijing.aliyuncs.com/chatchat/chatchat:0.2.6) 已经更新到 ```0.2.7``` 版本。 🐳 [Docker 镜像](registry.cn-beijing.aliyuncs.com/chatchat/chatchat:0.2.6) 已经更新到 ```0.2.7``` 版本。
@ -52,7 +62,10 @@
docker run -d --gpus all -p 80:8501 registry.cn-beijing.aliyuncs.com/chatchat/chatchat:0.2.7 docker run -d --gpus all -p 80:8501 registry.cn-beijing.aliyuncs.com/chatchat/chatchat:0.2.7
``` ```
🧩 本项目有一个非常完整的[Wiki](https://github.com/chatchat-space/Langchain-Chatchat/wiki/) ， README只是一个简单的介绍，__仅仅是入门教程，能够基础运行__。 如果你想要更深入的了解本项目，或者想对本项目做出贡献。请移步 [Wiki](https://github.com/chatchat-space/Langchain-Chatchat/wiki/) 界面 🧩 本项目有一个非常完整的[Wiki](https://github.com/chatchat-space/Langchain-Chatchat/wiki/) ， README只是一个简单的介绍，_
_仅仅是入门教程，能够基础运行__。
如果你想要更深入的了解本项目，或者想对本项目做出贡献。请移步 [Wiki](https://github.com/chatchat-space/Langchain-Chatchat/wiki/)
界面
## 解决的痛点 ## 解决的痛点
@ -62,17 +75,19 @@ docker run -d --gpus all -p 80:8501 registry.cn-beijing.aliyuncs.com/chatchat/ch
我们支持市面上主流的本地大语言模型和Embedding模型，支持开源的本地向量数据库。 我们支持市面上主流的本地大语言模型和Embedding模型，支持开源的本地向量数据库。
支持列表详见[Wiki](https://github.com/chatchat-space/Langchain-Chatchat/wiki/) 支持列表详见[Wiki](https://github.com/chatchat-space/Langchain-Chatchat/wiki/)
## 快速上手 ## 快速上手
### 1. 环境配置 ### 1. 环境配置
+ 首先，确保你的机器安装了 Python 3.8 - 3.11 + 首先，确保你的机器安装了 Python 3.8 - 3.11 (我们强烈推荐使用 Python3.11)。
``` ```
$ python --version $ python --version
Python 3.11.7 Python 3.11.7
``` ```
接着，创建一个虚拟环境，并在虚拟环境内安装项目的依赖 接着，创建一个虚拟环境，并在虚拟环境内安装项目的依赖
```shell ```shell
# 拉取仓库 # 拉取仓库
@ -88,33 +103,44 @@ $ pip install -r requirements_webui.txt
# 默认依赖包括基本运行环境（FAISS向量库）。如果要使用 milvus/pg_vector 等向量库，请将 requirements.txt 中相应依赖取消注释再安装。 # 默认依赖包括基本运行环境（FAISS向量库）。如果要使用 milvus/pg_vector 等向量库，请将 requirements.txt 中相应依赖取消注释再安装。
``` ```
请注意，LangChain-Chatchat `0.2.x` 系列是针对 Langchain `0.0.x` 系列版本的，如果你使用的是 Langchain `0.1.x` 系列版本，需要降级。
请注意，LangChain-Chatchat `0.2.x` 系列是针对 Langchain `0.0.x` 系列版本的，如果你使用的是 Langchain `0.1.x`
系列版本，需要降级您的`Langchain`版本。
### 2， 模型下载 ### 2， 模型下载
如需在本地或离线环境下运行本项目，需要首先将项目所需的模型下载至本地，通常开源 LLM 与 Embedding 模型可以从 [HuggingFace](https://huggingface.co/models) 下载。 如需在本地或离线环境下运行本项目，需要首先将项目所需的模型下载至本地，通常开源 LLM 与 Embedding
模型可以从 [HuggingFace](https://huggingface.co/models) 下载。
以本项目中默认使用的 LLM 模型 [THUDM/ChatGLM3-6B](https://huggingface.co/THUDM/chatglm3-6b) 与 Embedding 模型 [BAAI/bge-large-zh](https://huggingface.co/BAAI/bge-large-zh) 为例： 以本项目中默认使用的 LLM 模型 [THUDM/ChatGLM3-6B](https://huggingface.co/THUDM/chatglm3-6b) 与 Embedding
模型 [BAAI/bge-large-zh](https://huggingface.co/BAAI/bge-large-zh) 为例：
下载模型需要先[安装 Git LFS](https://docs.github.com/zh/repositories/working-with-files/managing-large-files/installing-git-large-file-storage)，然后运行 下载模型需要先[安装 Git LFS](https://docs.github.com/zh/repositories/working-with-files/managing-large-files/installing-git-large-file-storage)
，然后运行
```Shell ```Shell
$ git lfs install $ git lfs install
$ git clone https://huggingface.co/THUDM/chatglm3-6b $ git clone https://huggingface.co/THUDM/chatglm3-6b
$ git clone https://huggingface.co/BAAI/bge-large-zh $ git clone https://huggingface.co/BAAI/bge-large-zh
``` ```
### 3. 初始化知识库和配置文件 ### 3. 初始化知识库和配置文件
按照下列方式初始化自己的知识库和简单的复制配置文件 按照下列方式初始化自己的知识库和简单的复制配置文件
```shell ```shell
$ python copy_config_example.py $ python copy_config_example.py
$ python init_database.py --recreate-vs $ python init_database.py --recreate-vs
``` ```
### 4. 一键启动 ### 4. 一键启动
按照以下命令启动项目 按照以下命令启动项目
```shell ```shell
$ python startup.py -a $ python startup.py -a
``` ```
### 5. 启动界面示例 ### 5. 启动界面示例
如果正常启动，你将能看到以下界面 如果正常启动，你将能看到以下界面
@ -133,29 +159,37 @@ $ python startup.py -a
![](img/init_knowledge_base.jpg) ![](img/init_knowledge_base.jpg)
### 注意 ### 注意
以上方式只是为了快速上手，如果需要更多的功能和自定义启动方式 ，请参考[Wiki](https://github.com/chatchat-space/Langchain-Chatchat/wiki/) 以上方式只是为了快速上手，如果需要更多的功能和自定义启动方式
，请参考[Wiki](https://github.com/chatchat-space/Langchain-Chatchat/wiki/)
--- ---
## 项目里程碑 ## 项目里程碑
+ `2023年4月`: `Langchain-ChatGLM 0.1.0` 发布，支持基于 ChatGLM-6B 模型的本地知识库问答。 + `2023年4月`: `Langchain-ChatGLM 0.1.0` 发布，支持基于 ChatGLM-6B 模型的本地知识库问答。
+ `2023年8月`: `Langchain-ChatGLM` 改名为 `Langchain-Chatchat`，`0.2.0` 发布，使用 `fastchat` 作为模型加载方案，支持更多的模型和数据库。 + `2023年8月`: `Langchain-ChatGLM` 改名为 `Langchain-Chatchat`，`0.2.0` 发布，使用 `fastchat` 作为模型加载方案，支持更多的模型和数据库。
+ `2023年10月`: `Langchain-Chatchat 0.2.5` 发布，推出 Agent 内容，开源项目在`Founder Park & Zhipu AI & Zilliz` 举办的黑客马拉松获得三等奖。 + `2023年10月`: `Langchain-Chatchat 0.2.5` 发布，推出 Agent 内容，开源项目在`Founder Park & Zhipu AI & Zilliz`
举办的黑客马拉松获得三等奖。
+ `2023年12月`: `Langchain-Chatchat` 开源项目获得超过 **20K** stars. + `2023年12月`: `Langchain-Chatchat` 开源项目获得超过 **20K** stars.
+ `2024年1月`: `LangChain 0.1.x` 推出，`Langchain-Chatchat 0.2.x` 停止更新和技术支持，全力研发具有更强应用性的 `Langchain-Chatchat 0.3.x`。 + `2024年1月`: `LangChain 0.1.x` 推出，`Langchain-Chatchat 0.2.x` 发布稳定版本`0.2.10`
后将停止更新和技术支持，全力研发具有更强应用性的 `Langchain-Chatchat 0.3.x`。
+ 🔥 让我们一起期待未来 Chatchat 的故事 ··· + 🔥 让我们一起期待未来 Chatchat 的故事 ···
--- ---
## 联系我们 ## 联系我们
### Telegram ### Telegram
[![Telegram](https://img.shields.io/badge/Telegram-2CA5E0?style=for-the-badge&logo=telegram&logoColor=white "langchain-chatglm")](https://t.me/+RjliQ3jnJ1YyN2E9) [![Telegram](https://img.shields.io/badge/Telegram-2CA5E0?style=for-the-badge&logo=telegram&logoColor=white "langchain-chatglm")](https://t.me/+RjliQ3jnJ1YyN2E9)
### 项目交流群 ### 项目交流群
<img src="img/qr_code_85.jpg" alt="二维码" width="300" />
<img src="img/qr_code_87.jpg" alt="二维码" width="300" />
🎉 Langchain-Chatchat 项目微信交流群，如果你也对本项目感兴趣，欢迎加入群聊参与讨论交流。 🎉 Langchain-Chatchat 项目微信交流群，如果你也对本项目感兴趣，欢迎加入群聊参与讨论交流。

34
README_en.md
View File

@ -7,6 +7,10 @@
A LLM application aims to implement knowledge and search engine based QA based on Langchain and open-source or remote A LLM application aims to implement knowledge and search engine based QA based on Langchain and open-source or remote
LLM API. LLM API.
⚠️`0.2.10` will be the last version of the `0.2.x` series. The `0.2.x` series will stop updating and technical support,
and strive to develop `Langchain-Chachat 0.3.x with stronger applicability. `.
--- ---
## Table of Contents ## Table of Contents
@ -24,7 +28,8 @@ LLM API.
## Introduction ## Introduction
🤖️ A Q&A application based on local knowledge base implemented using the idea 🤖️ A Q&A application based on local knowledge base implemented using the idea
of [langchain](https://github.com/hwchase17/langchain). The goal is to build a KBQA(Knowledge based Q&A) solution that of [langchain](https://github.com/langchain-ai/langchain). The goal is to build a KBQA(Knowledge based Q&A) solution
that
is friendly to Chinese scenarios and open source models and can run both offline and online. is friendly to Chinese scenarios and open source models and can run both offline and online.
💡 Inspired by [document.ai](https://github.com/GanymedeNil/document.ai) 💡 Inspired by [document.ai](https://github.com/GanymedeNil/document.ai)
@ -55,10 +60,9 @@ The main process analysis from the aspect of document process:
🚩 The training or fine-tuning are not involved in the project, but still, one always can improve performance by do 🚩 The training or fine-tuning are not involved in the project, but still, one always can improve performance by do
these. these.
🌐 [AutoDL image](registry.cn-beijing.aliyuncs.com/chatchat/chatchat:0.2.5) is supported, and in v13 the codes are update 🌐 [AutoDL image](https://www.codewithgpu.com/i/chatchat-space/Langchain-Chatchat/Langchain-Chatchat) is supported, and in v13 the codes are update to v0.2.9.
to v0.2.9.
🐳 [Docker image](registry.cn-beijing.aliyuncs.com/chatchat/chatchat:0.2.7) 🐳 [Docker image](registry.cn-beijing.aliyuncs.com/chatchat/chatchat:0.2.7) is supported to 0.2.7
## Pain Points Addressed ## Pain Points Addressed
@ -98,7 +102,9 @@ $ pip install -r requirements_webui.txt
# 默认依赖包括基本运行环境（FAISS向量库）。如果要使用 milvus/pg_vector 等向量库，请将 requirements.txt 中相应依赖取消注释再安装。 # 默认依赖包括基本运行环境（FAISS向量库）。如果要使用 milvus/pg_vector 等向量库，请将 requirements.txt 中相应依赖取消注释再安装。
``` ```
Please note that the LangChain-Chachat `0.2.x` series is for the Langchain `0.0.x` series version. If you are using the Langchain `0.1.x` series version, you need to downgrade.
Please note that the LangChain-Chachat `0.2.x` series is for the Langchain `0.0.x` series version. If you are using the
Langchain `0.1.x` series version, you need to downgrade.
### Model Download ### Model Download
@ -157,15 +163,23 @@ The above instructions are provided for a quick start. If you need more features
please refer to the [Wiki](https://github.com/chatchat-space/Langchain-Chatchat/wiki/). please refer to the [Wiki](https://github.com/chatchat-space/Langchain-Chatchat/wiki/).
--- ---
## Project Milestones ## Project Milestones
+ `April 2023`: `Langchain-ChatGLM 0.1.0` released, supporting local knowledge base question and answer based on the ChatGLM-6B model. + `April 2023`: `Langchain-ChatGLM 0.1.0` released, supporting local knowledge base question and answer based on the
+ `August 2023`: `Langchain-ChatGLM` was renamed to `Langchain-Chatchat`, `0.2.0` was released, using `fastchat` as the model loading solution, supporting more models and databases. ChatGLM-6B model.
+ `October 2023`: `Langchain-Chachat 0.2.5` was released, Agent content was launched, and the open source project won the third prize in the hackathon held by `Founder Park & Zhipu AI & Zilliz`. + `August 2023`: `Langchain-ChatGLM` was renamed to `Langchain-Chatchat`, `0.2.0` was released, using `fastchat` as the
model loading solution, supporting more models and databases.
+ `October 2023`: `Langchain-Chachat 0.2.5` was released, Agent content was launched, and the open source project won
the third prize in the hackathon held by `Founder Park & Zhipu AI & Zilliz`.
+ `December 2023`: `Langchain-Chachat` open source project received more than **20K** stars. + `December 2023`: `Langchain-Chachat` open source project received more than **20K** stars.
+ `January 2024`: `LangChain 0.1.x` is launched, `Langchain-Chatchat 0.2.x` will stop updating and technical support, and all efforts will be made to develop `Langchain-Chatchat 0.3.x` with stronger applicability. + `January 2024`: `LangChain 0.1.x` is launched, `Langchain-Chachat 0.2.x` is released. After the stable
version `0.2.10` is released, updates and technical support will be stopped, and all efforts will be made to
develop `Langchain with stronger applicability -Chat 0.3.x`.
+ 🔥 Let’s look forward to the future Chatchat stories together··· + 🔥 Let’s look forward to the future Chatchat stories together···
--- ---
## Contact Us ## Contact Us
@ -176,7 +190,7 @@ please refer to the [Wiki](https://github.com/chatchat-space/Langchain-Chatchat/
### WeChat Group ### WeChat Group
<img src="img/qr_code_85.jpg" alt="二维码" width="300" height="300" /> <img src="img/qr_code_87.jpg" alt="二维码" width="300" height="300" />
### WeChat Official Account ### WeChat Official Account

2
configs/model_config.py.example
View File

@ -120,7 +120,7 @@ ONLINE_LLM_MODEL = {
"secret_key": "", "secret_key": "",
"provider": "TianGongWorker", "provider": "TianGongWorker",
}, },
# Gemini API (开发组未测试，由社群提供，只支持pro） # Gemini API https://makersuite.google.com/app/apikey
"gemini-api": { "gemini-api": {
"api_key": "", "api_key": "",
"provider": "GeminiWorker", "provider": "GeminiWorker",

17
configs/server_config.py.example
View File

@ -92,11 +92,10 @@ FSCHAT_MODEL_WORKERS = {
# 'disable_log_requests': False # 'disable_log_requests': False
}, },
# 可以如下示例方式更改默认配置 "Qwen-1_8B-Chat": {
# "Qwen-1_8B-Chat": { # 使用default中的IP和端口 "device": "cpu",
# "device": "cpu", },
# }, "chatglm3-6b": {
"chatglm3-6b": { # 使用default中的IP和端口
"device": "cuda", "device": "cuda",
}, },
@ -129,16 +128,10 @@ FSCHAT_MODEL_WORKERS = {
"port": 21009, "port": 21009,
}, },
"gemini-api": { "gemini-api": {
"port": 21012, "port": 21010,
}, },
} }
# fastchat multi model worker server
FSCHAT_MULTI_MODEL_WORKERS = {
# TODO:
}
# fastchat controller server
FSCHAT_CONTROLLER = { FSCHAT_CONTROLLER = {
"host": DEFAULT_BIND_HOST, "host": DEFAULT_BIND_HOST,
"port": 20001, "port": 20001,

3
document_loaders/mypdfloader.py
View File

@ -16,11 +16,8 @@ class RapidOCRPDFLoader(UnstructuredFileLoader):
b_unit = tqdm.tqdm(total=doc.page_count, desc="RapidOCRPDFLoader context page index: 0") b_unit = tqdm.tqdm(total=doc.page_count, desc="RapidOCRPDFLoader context page index: 0")
for i, page in enumerate(doc): for i, page in enumerate(doc):
# 更新描述
b_unit.set_description("RapidOCRPDFLoader context page index: {}".format(i)) b_unit.set_description("RapidOCRPDFLoader context page index: {}".format(i))
# 立即显示进度条更新结果
b_unit.refresh() b_unit.refresh()
# TODO: 依据文本与图片顺序调整处理方式
text = page.get_text("") text = page.get_text("")
resp += text + "\n" resp += text + "\n"

BIN
img/qr_code_85.jpg
View File

Binary file not shown.
Side by Side

Before

Width:  |  Height:  |  Size: 272 KiB

BIN
img/qr_code_87.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 318 KiB

94
requirements.txt
View File

@ -1,79 +1,61 @@
# API requirements
torch~=2.1.2 torch~=2.1.2
torchvision~=0.16.2 torchvision~=0.16.2
torchaudio~=2.1.2 torchaudio~=2.1.2
xformers==0.0.23.post1 xformers~=0.0.23.post1
transformers==4.36.2 transformers~=4.36.2
sentence_transformers==2.2.2 sentence_transformers~=2.2.2
langchain==0.0.354 langchain==0.0.354
langchain-experimental==0.0.47 langchain-experimental==0.0.47
pydantic==1.10.13 pydantic==1.10.13
fschat==0.2.35 fschat~=0.2.35
openai~=1.7.1 openai~=1.9.0
fastapi~=0.108.0 fastapi~=0.109.0
sse_starlette==1.8.2 sse_starlette~=1.8.2
nltk>=3.8.1 nltk~=3.8.1
uvicorn>=0.24.0.post1 uvicorn~=0.24.0.post1
starlette~=0.32.0 starlette~=0.32.0
unstructured[all-docs]==0.11.0 unstructured[all-docs]~=0.12.0
python-magic-bin; sys_platform == 'win32' python-magic-bin; sys_platform ~= 'win32'
SQLAlchemy==2.0.19 SQLAlchemy~=2.0.25
faiss-cpu~=1.7.4 # `conda install faiss-gpu -c conda-forge` if you want to accelerate with gpus faiss-cpu~=1.7.4
accelerate~=0.24.1 accelerate~=0.24.1
spacy~=3.7.2 spacy~=3.7.2
PyMuPDF~=1.23.8 PyMuPDF~=1.23.16
rapidocr_onnxruntime==1.3.8 rapidocr_onnxruntime~=1.3.8
requests~=2.31.0 requests~=2.31.0
pathlib~=1.0.1 pathlib~=1.0.1
pytest~=7.4.3 pytest~=7.4.3
numexpr~=2.8.6 # max version for py38 numexpr~=2.8.6
strsimpy~=0.2.1 strsimpy~=0.2.1
markdownify~=0.11.6 markdownify~=0.11.6
tiktoken~=0.5.2 tiktoken~=0.5.2
tqdm>=4.66.1 tqdm~=4.66.1
websockets>=12.0 websockets~=12.0
numpy~=1.24.4 numpy~=1.24.4
pandas~=2.0.3 pandas~=2.0.3
einops>=0.7.0 einops~=0.7.0
transformers_stream_generator==0.0.4 transformers_stream_generator~=0.0.4
vllm==0.2.7; sys_platform == "linux" vllm~=0.2.7; sys_platform ~= "linux"
jq~=1.6.0
# flash-attn>=2.4.3 # For Orion-14B-Chat and Qwen-14B-Chat beautifulsoup4~=4.12.2
# optional document loaders
#rapidocr_paddle[gpu]>=1.3.0.post5 # gpu accelleration for ocr of pdf and image files
jq==1.6.0 # for .json and .jsonl files. suggest `conda install jq` on windows
beautifulsoup4~=4.12.2 # for .mhtml files
pysrt~=1.1.2 pysrt~=1.1.2
dashscope~=1.13.6 # qwen
# Online api libs dependencies # volcengine~=1.0.119 # fangzhou
# zhipuAI sdk is not supported on our platform, so use http instead
dashscope==1.13.6 # qwen
# volcengine>=1.0.119 # fangzhou
# uncomment libs if you want to use corresponding vector store # uncomment libs if you want to use corresponding vector store
# pymilvus>=2.3.4 # pymilvus~=2.3.4
# psycopg2==2.9.9 # psycopg2~=2.9.9
# pgvector>=0.2.4 # pgvector~=0.2.4
# flash-attn~=2.4.3 # For Orion-14B-Chat and Qwen-14B-Chat
# Agent and Search Tools #rapidocr_paddle[gpu]~=1.3.0.post5 # gpu accelleration for ocr of pdf and image files
arxiv~=2.1.0 arxiv~=2.1.0
youtube-search~=2.1.2 youtube-search~=2.1.2
duckduckgo-search~=3.9.9 duckduckgo-search~=3.9.9
metaphor-python~=0.1.23 metaphor-python~=0.1.23
streamlit~=1.30.0
# WebUI requirements streamlit-option-menu~=0.3.12
streamlit-antd-components~=0.3.1
streamlit==1.30.0 streamlit-chatbox~=1.1.11
streamlit-option-menu==0.3.6 streamlit-modal~=0.1.0
streamlit-antd-components==0.3.1 streamlit-aggrid~=0.3.4.post3
streamlit-chatbox==1.1.11 httpx~=0.26.0
streamlit-modal==0.1.0 watchdog~=3.0.0
streamlit-aggrid==0.3.4.post3
httpx==0.26.0
watchdog==3.0.0

40
requirements_api.txt
View File

@ -1,15 +1,14 @@
torch~=2.1.2 torch~=2.1.2
torchvision~=0.16.2 torchvision~=0.16.2
torchaudio~=2.1.2 torchaudio~=2.1.2
xformers==0.0.23.post1 xformers>=0.0.23.post1
transformers==4.36.2 transformers==4.36.2
sentence_transformers==2.2.2 sentence_transformers==2.2.2
langchain==0.0.354 langchain==0.0.354
langchain-experimental==0.0.47 langchain-experimental==0.0.47
pydantic==1.10.13 pydantic==1.10.13
fschat==0.2.35 fschat==0.2.35
openai~=1.7.1 openai~=1.9.0
fastapi~=0.108.0 fastapi~=0.108.0
sse_starlette==1.8.2 sse_starlette==1.8.2
nltk>=3.8.1 nltk>=3.8.1
@ -18,7 +17,7 @@ starlette~=0.32.0
unstructured[all-docs]==0.11.0 unstructured[all-docs]==0.11.0
python-magic-bin; sys_platform == 'win32' python-magic-bin; sys_platform == 'win32'
SQLAlchemy==2.0.19 SQLAlchemy==2.0.19
faiss-cpu~=1.7.4 # `conda install faiss-gpu -c conda-forge` if you want to accelerate with gpus faiss-cpu~=1.7.4
accelerate~=0.24.1 accelerate~=0.24.1
spacy~=3.7.2 spacy~=3.7.2
PyMuPDF~=1.23.8 PyMuPDF~=1.23.8
@ -26,7 +25,7 @@ rapidocr_onnxruntime==1.3.8
requests~=2.31.0 requests~=2.31.0
pathlib~=1.0.1 pathlib~=1.0.1
pytest~=7.4.3 pytest~=7.4.3
numexpr~=2.8.6 # max version for py38 numexpr~=2.8.6
strsimpy~=0.2.1 strsimpy~=0.2.1
markdownify~=0.11.6 markdownify~=0.11.6
tiktoken~=0.5.2 tiktoken~=0.5.2
@ -39,29 +38,18 @@ transformers_stream_generator==0.0.4
vllm==0.2.7; sys_platform == "linux" vllm==0.2.7; sys_platform == "linux"
httpx==0.26.0 httpx==0.26.0
llama-index llama-index
# flash-attn>=2.4.3 # For Orion-14B-Chat and Qwen-14B-Chat jq==1.6.0
beautifulsoup4~=4.12.2
# optional document loaders
# rapidocr_paddle[gpu]>=1.3.0.post5 # gpu accelleration for ocr of pdf and image files
jq==1.6.0 # for .json and .jsonl files. suggest `conda install jq` on windows
beautifulsoup4~=4.12.2 # for .mhtml files
pysrt~=1.1.2 pysrt~=1.1.2
dashscope==1.13.6
# Online api libs dependencies
# zhipuAI sdk is not supported on our platform, so use http instead
dashscope==1.13.6 # qwen
# volcengine>=1.0.119 # fangzhou
# uncomment libs if you want to use corresponding vector store
# pymilvus>=2.3.4
# psycopg2==2.9.9
# pgvector>=0.2.4
# Agent and Search Tools
arxiv~=2.1.0 arxiv~=2.1.0
youtube-search~=2.1.2 youtube-search~=2.1.2
duckduckgo-search~=3.9.9 duckduckgo-search~=3.9.9
metaphor-python~=0.1.23 metaphor-python~=0.1.23
# volcengine>=1.0.119
# pymilvus>=2.3.4
# psycopg2==2.9.9
# pgvector>=0.2.4
# flash-attn>=2.4.3 # For Orion-14B-Chat and Qwen-14B-Chat
# rapidocr_paddle[gpu]>=1.3.0.post5

70
requirements_lite.txt
View File

@ -1,64 +1,32 @@
# API requirements
langchain==0.0.354 langchain==0.0.354
langchain-experimental==0.0.47 langchain-experimental==0.0.47
pydantic==1.10.13 pydantic==1.10.13
fschat==0.2.35 fschat~=0.2.35
openai~=1.7.1 openai~=1.9.0
fastapi~=0.108.0 fastapi~=0.109.0
sse_starlette==1.8.2 sse_starlette~=1.8.2
nltk>=3.8.1 nltk~=3.8.1
uvicorn>=0.24.0.post1 uvicorn~=0.24.0.post1
starlette~=0.32.0 starlette~=0.32.0
unstructured[all-docs]==0.11.0 unstructured[all-docs]~=0.12.0
python-magic-bin; sys_platform == 'win32' python-magic-bin; sys_platform ~= 'win32'
SQLAlchemy==2.0.19 SQLAlchemy~=2.0.25
faiss-cpu~=1.7.4 faiss-cpu~=1.7.4
accelerate~=0.24.1
spacy~=3.7.2
PyMuPDF~=1.23.16
rapidocr_onnxruntime~=1.3.8
requests~=2.31.0 requests~=2.31.0
pathlib~=1.0.1 pathlib~=1.0.1
pytest~=7.4.3 pytest~=7.4.3
numexpr~=2.8.6 # max version for py38
strsimpy~=0.2.1
markdownify~=0.11.6
tiktoken~=0.5.2
tqdm>=4.66.1
websockets>=12.0
numpy~=1.24.4
pandas~=2.0.3
einops>=0.7.0
transformers_stream_generator==0.0.4
vllm==0.2.7; sys_platform == "linux"
httpx[brotli,http2,socks]==0.25.2
requests
pathlib
pytest
# Online api libs dependencies
# zhipuAI sdk is not supported on our platform, so use http instead
dashscope==1.13.6 dashscope==1.13.6
# volcengine>=1.0.119
# uncomment libs if you want to use corresponding vector store
# pymilvus>=2.3.4
# psycopg2==2.9.9
# pgvector>=0.2.4
# Agent and Search Tools
arxiv~=2.1.0 arxiv~=2.1.0
youtube-search~=2.1.2 youtube-search~=2.1.2
duckduckgo-search~=3.9.9 duckduckgo-search~=3.9.9
metaphor-python~=0.1.23 metaphor-python~=0.1.23
watchdog~=3.0.0
# WebUI requirements # volcengine>=1.0.119
# pymilvus>=2.3.4
streamlit==1.30.0 # psycopg2==2.9.9
streamlit-option-menu==0.3.6 # pgvector>=0.2.4
streamlit-antd-components==0.3.1 # flash-attn>=2.4.3 # For Orion-14B-Chat and Qwen-14B-Chat
streamlit-chatbox==1.1.11
streamlit-modal==0.1.0
streamlit-aggrid==0.3.4.post3
httpx==0.26.0
watchdog==3.0.0

18
requirements_webui.txt
View File

@ -1,10 +1,8 @@
# WebUI requirements streamlit~=1.30.0
streamlit-option-menu~=0.3.12
streamlit==1.30.0 streamlit-antd-components~=0.3.1
streamlit-option-menu==0.3.6 streamlit-chatbox~=1.1.11
streamlit-antd-components==0.3.1 streamlit-modal~=0.1.0
streamlit-chatbox==1.1.11 streamlit-aggrid~=0.3.4.post3
streamlit-modal==0.1.0 httpx~=0.26.0
streamlit-aggrid==0.3.4.post3 watchdog~=3.0.0
httpx==0.26.0
watchdog==3.0.0

1
server/embeddings_api.py
View File

@ -16,7 +16,6 @@ def embed_texts(
) -> BaseResponse: ) -> BaseResponse:
''' '''
对文本进行向量化。返回数据格式：BaseResponse(data=List[List[float]]) 对文本进行向量化。返回数据格式：BaseResponse(data=List[List[float]])
TODO: 也许需要加入缓存机制，减少 token 消耗
''' '''
try: try:
if embed_model in list_embed_models(): # 使用本地Embeddings模型 if embed_model in list_embed_models(): # 使用本地Embeddings模型

10
server/knowledge_base/kb_api.py
View File

@ -13,9 +13,9 @@ def list_kbs():
def create_kb(knowledge_base_name: str = Body(..., examples=["samples"]), def create_kb(knowledge_base_name: str = Body(..., examples=["samples"]),
vector_store_type: str = Body("faiss"), vector_store_type: str = Body("faiss"),
embed_model: str = Body(EMBEDDING_MODEL), embed_model: str = Body(EMBEDDING_MODEL),
) -> BaseResponse: ) -> BaseResponse:
# Create selected knowledge base # Create selected knowledge base
if not validate_kb_name(knowledge_base_name): if not validate_kb_name(knowledge_base_name):
return BaseResponse(code=403, msg="Don't attack me") return BaseResponse(code=403, msg="Don't attack me")
@ -39,8 +39,8 @@ def create_kb(knowledge_base_name: str = Body(..., examples=["samples"]),
def delete_kb( def delete_kb(
knowledge_base_name: str = Body(..., examples=["samples"]) knowledge_base_name: str = Body(..., examples=["samples"])
) -> BaseResponse: ) -> BaseResponse:
# Delete selected knowledge base # Delete selected knowledge base
if not validate_kb_name(knowledge_base_name): if not validate_kb_name(knowledge_base_name):
return BaseResponse(code=403, msg="Don't attack me") return BaseResponse(code=403, msg="Don't attack me")

2
server/knowledge_base/kb_cache/faiss_cache.py
View File

@ -55,8 +55,6 @@ class _FaissPool(CachePool):
embed_model: str = EMBEDDING_MODEL, embed_model: str = EMBEDDING_MODEL,
embed_device: str = embedding_device(), embed_device: str = embedding_device(),
) -> FAISS: ) -> FAISS:
# TODO: 整个Embeddings加载逻辑有些混乱，待清理
# create an empty vector store
embeddings = EmbeddingsFunAdapter(embed_model) embeddings = EmbeddingsFunAdapter(embed_model)
doc = Document(page_content="init", metadata={}) doc = Document(page_content="init", metadata={})
vector_store = FAISS.from_documents([doc], embeddings, normalize_L2=True,distance_strategy="METRIC_INNER_PRODUCT") vector_store = FAISS.from_documents([doc], embeddings, normalize_L2=True,distance_strategy="METRIC_INNER_PRODUCT")

2
server/knowledge_base/kb_doc_api.py
View File

@ -95,7 +95,6 @@ def _save_files_in_thread(files: List[UploadFile],
and not override and not override
and os.path.getsize(file_path) == len(file_content) and os.path.getsize(file_path) == len(file_content)
): ):
# TODO: filesize 不同后的处理
file_status = f"文件 {filename} 已存在。" file_status = f"文件 {filename} 已存在。"
logger.warn(file_status) logger.warn(file_status)
return dict(code=404, msg=file_status, data=data) return dict(code=404, msg=file_status, data=data)
@ -116,7 +115,6 @@ def _save_files_in_thread(files: List[UploadFile],
yield result yield result
# TODO: 等langchain.document_loaders支持内存文件的时候再开通
# def files2docs(files: List[UploadFile] = File(..., description="上传文件，支持多文件"), # def files2docs(files: List[UploadFile] = File(..., description="上传文件，支持多文件"),
# knowledge_base_name: str = Form(..., description="知识库名称", examples=["samples"]), # knowledge_base_name: str = Form(..., description="知识库名称", examples=["samples"]),
# override: bool = Form(False, description="覆盖已有文件"), # override: bool = Form(False, description="覆盖已有文件"),

1
server/knowledge_base/kb_service/base.py
View File

@ -191,7 +191,6 @@ class KBService(ABC):
''' '''
传入参数为： {doc_id: Document, ...} 传入参数为： {doc_id: Document, ...}
如果对应 doc_id 的值为 None，或其 page_content 为空，则删除该文档 如果对应 doc_id 的值为 None，或其 page_content 为空，则删除该文档
TODO：是否要支持新增 docs ？
''' '''
self.del_doc_by_ids(list(docs.keys())) self.del_doc_by_ids(list(docs.keys()))
docs = [] docs = []

1
server/knowledge_base/kb_service/milvus_kb_service.py
View File

@ -70,7 +70,6 @@ class MilvusKBService(KBService):
return score_threshold_process(score_threshold, top_k, docs) return score_threshold_process(score_threshold, top_k, docs)
def do_add_doc(self, docs: List[Document], **kwargs) -> List[Dict]: def do_add_doc(self, docs: List[Document], **kwargs) -> List[Dict]:
# TODO: workaround for bug #10492 in langchain
for doc in docs: for doc in docs:
for k, v in doc.metadata.items(): for k, v in doc.metadata.items():
doc.metadata[k] = str(v) doc.metadata[k] = str(v)

2
server/knowledge_base/kb_service/pg_kb_service.py
View File

@ -32,8 +32,6 @@ class PGKBService(KBService):
results = [Document(page_content=row[0], metadata=row[1]) for row in results = [Document(page_content=row[0], metadata=row[1]) for row in
session.execute(stmt, {'ids': ids}).fetchall()] session.execute(stmt, {'ids': ids}).fetchall()]
return results return results
# TODO:
def del_doc_by_ids(self, ids: List[str]) -> bool: def del_doc_by_ids(self, ids: List[str]) -> bool:
return super().del_doc_by_ids(ids) return super().del_doc_by_ids(ids)

1
server/knowledge_base/kb_summary/base.py
View File

@ -13,7 +13,6 @@ from server.db.repository.knowledge_metadata_repository import add_summary_to_db
from langchain.docstore.document import Document from langchain.docstore.document import Document
# TODO 暂不考虑文件更新，需要重新删除相关文档，再重新添加
class KBSummaryService(ABC): class KBSummaryService(ABC):
kb_name: str kb_name: str
embed_model: str embed_model: str

6
server/knowledge_base/kb_summary/summary_chunk.py
View File

@ -112,12 +112,6 @@ class SummaryAdapter:
docs: List[DocumentWithVSId] = []) -> List[Document]: docs: List[DocumentWithVSId] = []) -> List[Document]:
logger.info("start summary") logger.info("start summary")
# TODO 暂不处理文档中涉及语义重复、上下文缺失、document was longer than the context length 的问题
# merge_docs = self._drop_overlap(docs)
# # 将merge_docs中的句子合并成一个文档
# text = self._join_docs(merge_docs)
# 根据段落于句子的分隔符，将文档分成chunk，每个chunk长度小于token_max长度
""" """
这个过程分成两个部分： 这个过程分成两个部分：
1. 对每个文档进行处理，得到每个文档的摘要 1. 对每个文档进行处理，得到每个文档的摘要

1
server/knowledge_base/utils.py
View File

@ -174,7 +174,6 @@ def get_loader(loader_name: str, file_path: str, loader_kwargs: Dict = None):
if encode_detect is None: if encode_detect is None:
encode_detect = {"encoding": "utf-8"} encode_detect = {"encoding": "utf-8"}
loader_kwargs["encoding"] = encode_detect["encoding"] loader_kwargs["encoding"] = encode_detect["encoding"]
## TODO：支持更多的自定义CSV读取逻辑
elif loader_name == "JSONLoader": elif loader_name == "JSONLoader":
loader_kwargs.setdefault("jq_schema", ".") loader_kwargs.setdefault("jq_schema", ".")

2
server/model_workers/azure.py
View File

@ -67,12 +67,10 @@ class AzureWorker(ApiModelWorker):
self.logger.error(f"请求 Azure API 时发生错误：{resp}") self.logger.error(f"请求 Azure API 时发生错误：{resp}")
def get_embeddings(self, params): def get_embeddings(self, params):
# TODO: 支持embeddings
print("embedding") print("embedding")
print(params) print(params)
def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation: def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation:
# TODO: 确认模板是否需要修改
return conv.Conversation( return conv.Conversation(
name=self.model_names[0], name=self.model_names[0],
system_message="You are a helpful, respectful and honest assistant.", system_message="You are a helpful, respectful and honest assistant.",

2
server/model_workers/baichuan.py
View File

@ -88,12 +88,10 @@ class BaiChuanWorker(ApiModelWorker):
yield data yield data
def get_embeddings(self, params): def get_embeddings(self, params):
# TODO: 支持embeddings
print("embedding") print("embedding")
print(params) print(params)
def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation: def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation:
# TODO: 确认模板是否需要修改
return conv.Conversation( return conv.Conversation(
name=self.model_names[0], name=self.model_names[0],
system_message="", system_message="",

2
server/model_workers/base.py
View File

@ -125,8 +125,6 @@ class ApiModelWorker(BaseModelWorker):
def count_token(self, params): def count_token(self, params):
# TODO：需要完善
# print("count token")
prompt = params["prompt"] prompt = params["prompt"]
return {"count": len(str(prompt)), "error_code": 0} return {"count": len(str(prompt)), "error_code": 0}

33
server/model_workers/fangzhou.py
View File

@ -12,16 +12,16 @@ class FangZhouWorker(ApiModelWorker):
""" """
def __init__( def __init__(
self, self,
*, *,
model_names: List[str] = ["fangzhou-api"], model_names: List[str] = ["fangzhou-api"],
controller_addr: str = None, controller_addr: str = None,
worker_addr: str = None, worker_addr: str = None,
version: Literal["chatglm-6b-model"] = "chatglm-6b-model", version: Literal["chatglm-6b-model"] = "chatglm-6b-model",
**kwargs, **kwargs,
): ):
kwargs.update(model_names=model_names, controller_addr=controller_addr, worker_addr=worker_addr) kwargs.update(model_names=model_names, controller_addr=controller_addr, worker_addr=worker_addr)
kwargs.setdefault("context_len", 16384) # TODO: 不同的模型有不同的大小 kwargs.setdefault("context_len", 16384)
super().__init__(**kwargs) super().__init__(**kwargs)
self.version = version self.version = version
@ -53,15 +53,15 @@ class FangZhouWorker(ApiModelWorker):
if error := resp.error: if error := resp.error:
if error.code_n > 0: if error.code_n > 0:
data = { data = {
"error_code": error.code_n, "error_code": error.code_n,
"text": error.message, "text": error.message,
"error": { "error": {
"message": error.message, "message": error.message,
"type": "invalid_request_error", "type": "invalid_request_error",
"param": None, "param": None,
"code": None, "code": None,
}
} }
}
self.logger.error(f"请求方舟 API 时发生错误：{data}") self.logger.error(f"请求方舟 API 时发生错误：{data}")
yield data yield data
elif chunk := resp.choice.message.content: elif chunk := resp.choice.message.content:
@ -77,7 +77,6 @@ class FangZhouWorker(ApiModelWorker):
break break
def get_embeddings(self, params): def get_embeddings(self, params):
# TODO: 支持embeddings
print("embedding") print("embedding")
print(params) print(params)

13
server/model_workers/gemini.py
View File

@ -3,7 +3,7 @@ from fastchat.conversation import Conversation
from server.model_workers.base import * from server.model_workers.base import *
from server.utils import get_httpx_client from server.utils import get_httpx_client
from fastchat import conversation as conv from fastchat import conversation as conv
import json,httpx import json, httpx
from typing import List, Dict from typing import List, Dict
from configs import logger, log_verbose from configs import logger, log_verbose
@ -14,14 +14,14 @@ class GeminiWorker(ApiModelWorker):
*, *,
controller_addr: str = None, controller_addr: str = None,
worker_addr: str = None, worker_addr: str = None,
model_names: List[str] = ["Gemini-api"], model_names: List[str] = ["gemini-api"],
**kwargs, **kwargs,
): ):
kwargs.update(model_names=model_names, controller_addr=controller_addr, worker_addr=worker_addr) kwargs.update(model_names=model_names, controller_addr=controller_addr, worker_addr=worker_addr)
kwargs.setdefault("context_len", 4096) kwargs.setdefault("context_len", 4096)
super().__init__(**kwargs) super().__init__(**kwargs)
def create_gemini_messages(self,messages) -> json: def create_gemini_messages(self, messages) -> json:
has_history = any(msg['role'] == 'assistant' for msg in messages) has_history = any(msg['role'] == 'assistant' for msg in messages)
gemini_msg = [] gemini_msg = []
@ -46,7 +46,7 @@ class GeminiWorker(ApiModelWorker):
def do_chat(self, params: ApiChatParams) -> Dict: def do_chat(self, params: ApiChatParams) -> Dict:
params.load_config(self.model_names[0]) params.load_config(self.model_names[0])
data = self.create_gemini_messages(messages=params.messages) data = self.create_gemini_messages(messages=params.messages)
generationConfig=dict( generationConfig = dict(
temperature=params.temperature, temperature=params.temperature,
topK=1, topK=1,
topP=1, topP=1,
@ -55,7 +55,7 @@ class GeminiWorker(ApiModelWorker):
) )
data['generationConfig'] = generationConfig data['generationConfig'] = generationConfig
url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-pro:generateContent"+ '?key=' + params.api_key url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-pro:generateContent" + '?key=' + params.api_key
headers = { headers = {
'Content-Type': 'application/json', 'Content-Type': 'application/json',
} }
@ -67,7 +67,7 @@ class GeminiWorker(ApiModelWorker):
text = "" text = ""
json_string = "" json_string = ""
timeout = httpx.Timeout(60.0) timeout = httpx.Timeout(60.0)
client=get_httpx_client(timeout=timeout) client = get_httpx_client(timeout=timeout)
with client.stream("POST", url, headers=headers, json=data) as response: with client.stream("POST", url, headers=headers, json=data) as response:
for line in response.iter_lines(): for line in response.iter_lines():
line = line.strip() line = line.strip()
@ -95,7 +95,6 @@ class GeminiWorker(ApiModelWorker):
print("Invalid JSON string:", json_string) print("Invalid JSON string:", json_string)
def get_embeddings(self, params): def get_embeddings(self, params):
# TODO: 支持embeddings
print("embedding") print("embedding")
print(params) print(params)

5
server/model_workers/minimax.py
View File

@ -37,7 +37,6 @@ class MiniMaxWorker(ApiModelWorker):
def do_chat(self, params: ApiChatParams) -> Dict: def do_chat(self, params: ApiChatParams) -> Dict:
# 按照官网推荐，直接调用abab 5.5模型 # 按照官网推荐，直接调用abab 5.5模型
# TODO: 支持指定回复要求，支持指定用户名称、AI名称
params.load_config(self.model_names[0]) params.load_config(self.model_names[0])
url = 'https://api.minimax.chat/v1/text/chatcompletion{pro}?GroupId={group_id}' url = 'https://api.minimax.chat/v1/text/chatcompletion{pro}?GroupId={group_id}'
@ -55,7 +54,7 @@ class MiniMaxWorker(ApiModelWorker):
"temperature": params.temperature, "temperature": params.temperature,
"top_p": params.top_p, "top_p": params.top_p,
"tokens_to_generate": params.max_tokens or 1024, "tokens_to_generate": params.max_tokens or 1024,
# TODO: 以下参数为minimax特有，传入空值会出错。 # 以下参数为minimax特有，传入空值会出错。
# "prompt": params.system_message or self.conv.system_message, # "prompt": params.system_message or self.conv.system_message,
# "bot_setting": [], # "bot_setting": [],
# "role_meta": params.role_meta, # "role_meta": params.role_meta,
@ -143,12 +142,10 @@ class MiniMaxWorker(ApiModelWorker):
return {"code": 200, "data": result} return {"code": 200, "data": result}
def get_embeddings(self, params): def get_embeddings(self, params):
# TODO: 支持embeddings
print("embedding") print("embedding")
print(params) print(params)
def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation: def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation:
# TODO: 确认模板是否需要修改
return conv.Conversation( return conv.Conversation(
name=self.model_names[0], name=self.model_names[0],
system_message="你是MiniMax自主研发的大型语言模型，回答问题简洁有条理。", system_message="你是MiniMax自主研发的大型语言模型，回答问题简洁有条理。",

3
server/model_workers/qianfan.py
View File

@ -187,14 +187,11 @@ class QianFanWorker(ApiModelWorker):
i += batch_size i += batch_size
return {"code": 200, "data": result} return {"code": 200, "data": result}
# TODO: qianfan支持续写模型
def get_embeddings(self, params): def get_embeddings(self, params):
# TODO: 支持embeddings
print("embedding") print("embedding")
print(params) print(params)
def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation: def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation:
# TODO: 确认模板是否需要修改
return conv.Conversation( return conv.Conversation(
name=self.model_names[0], name=self.model_names[0],
system_message="你是一个聪明的助手，请根据用户的提示来完成任务", system_message="你是一个聪明的助手，请根据用户的提示来完成任务",

2
server/model_workers/qwen.py
View File

@ -100,12 +100,10 @@ class QwenWorker(ApiModelWorker):
return {"code": 200, "data": result} return {"code": 200, "data": result}
def get_embeddings(self, params): def get_embeddings(self, params):
# TODO: 支持embeddings
print("embedding") print("embedding")
print(params) print(params)
def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation: def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation:
# TODO: 确认模板是否需要修改
return conv.Conversation( return conv.Conversation(
name=self.model_names[0], name=self.model_names[0],
system_message="你是一个聪明、对人类有帮助的人工智能，你可以对人类提出的问题给出有用、详细、礼貌的回答。", system_message="你是一个聪明、对人类有帮助的人工智能，你可以对人类提出的问题给出有用、详细、礼貌的回答。",

2
server/model_workers/tiangong.py
View File

@ -70,12 +70,10 @@ class TianGongWorker(ApiModelWorker):
yield data yield data
def get_embeddings(self, params): def get_embeddings(self, params):
# TODO: 支持embeddings
print("embedding") print("embedding")
print(params) print(params)
def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation: def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation:
# TODO: 确认模板是否需要修改
return conv.Conversation( return conv.Conversation(
name=self.model_names[0], name=self.model_names[0],
system_message="", system_message="",

3
server/model_workers/xinghuo.py
View File

@ -42,7 +42,6 @@ class XingHuoWorker(ApiModelWorker):
self.version = version self.version = version
def do_chat(self, params: ApiChatParams) -> Dict: def do_chat(self, params: ApiChatParams) -> Dict:
# TODO: 当前每次对话都要重新连接websocket，确认是否可以保持连接
params.load_config(self.model_names[0]) params.load_config(self.model_names[0])
version_mapping = { version_mapping = {
@ -73,12 +72,10 @@ class XingHuoWorker(ApiModelWorker):
yield {"error_code": 0, "text": text} yield {"error_code": 0, "text": text}
def get_embeddings(self, params): def get_embeddings(self, params):
# TODO: 支持embeddings
print("embedding") print("embedding")
print(params) print(params)
def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation: def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation:
# TODO: 确认模板是否需要修改
return conv.Conversation( return conv.Conversation(
name=self.model_names[0], name=self.model_names[0],
system_message="你是一个聪明的助手，请根据用户的提示来完成任务", system_message="你是一个聪明的助手，请根据用户的提示来完成任务",

7
server/utils.py
View File

@ -36,7 +36,6 @@ async def wrap_done(fn: Awaitable, event: asyncio.Event):
await fn await fn
except Exception as e: except Exception as e:
logging.exception(e) logging.exception(e)
# TODO: handle exception
msg = f"Caught exception: {e}" msg = f"Caught exception: {e}"
logger.error(f'{e.__class__.__name__}: {msg}', logger.error(f'{e.__class__.__name__}: {msg}',
exc_info=e if log_verbose else None) exc_info=e if log_verbose else None)
@ -404,7 +403,7 @@ def fschat_controller_address() -> str:
def fschat_model_worker_address(model_name: str = LLM_MODELS[0]) -> str: def fschat_model_worker_address(model_name: str = LLM_MODELS[0]) -> str:
if model := get_model_worker_config(model_name): # TODO: depends fastchat if model := get_model_worker_config(model_name):
host = model["host"] host = model["host"]
if host == "0.0.0.0": if host == "0.0.0.0":
host = "127.0.0.1" host = "127.0.0.1"
@ -449,7 +448,7 @@ def get_prompt_template(type: str, name: str) -> Optional[str]:
from configs import prompt_config from configs import prompt_config
import importlib import importlib
importlib.reload(prompt_config) # TODO: 检查configs/prompt_config.py文件有修改再重新加载 importlib.reload(prompt_config)
return prompt_config.PROMPT_TEMPLATES[type].get(name) return prompt_config.PROMPT_TEMPLATES[type].get(name)
@ -550,7 +549,7 @@ def run_in_thread_pool(
thread = pool.submit(func, **kwargs) thread = pool.submit(func, **kwargs)
tasks.append(thread) tasks.append(thread)
for obj in as_completed(tasks): # TODO: Ctrl+c无法停止 for obj in as_completed(tasks):
yield obj.result() yield obj.result()

2
startup.py
View File

@ -418,7 +418,7 @@ def run_openai_api(log_level: str = "INFO", started_event: mp.Event = None):
set_httpx_config() set_httpx_config()
controller_addr = fschat_controller_address() controller_addr = fschat_controller_address()
app = create_openai_api_app(controller_addr, log_level=log_level) # TODO: not support keys yet. app = create_openai_api_app(controller_addr, log_level=log_level)
_set_app_event(app, started_event) _set_app_event(app, started_event)
host = FSCHAT_OPENAI_API["host"] host = FSCHAT_OPENAI_API["host"]

78
webui_pages/dialogue/dialogue.py
View File

@ -12,7 +12,6 @@ from server.knowledge_base.utils import LOADER_DICT
import uuid import uuid
from typing import List, Dict from typing import List, Dict
chat_box = ChatBox( chat_box = ChatBox(
assistant_avatar=os.path.join( assistant_avatar=os.path.join(
"img", "img",
@ -127,7 +126,6 @@ def dialogue_page(api: ApiRequest, is_lite: bool = False):
chat_box.use_chat_name(conversation_name) chat_box.use_chat_name(conversation_name)
conversation_id = st.session_state["conversation_ids"][conversation_name] conversation_id = st.session_state["conversation_ids"][conversation_name]
# TODO: 对话模型与会话绑定
def on_mode_change(): def on_mode_change():
mode = st.session_state.dialogue_mode mode = st.session_state.dialogue_mode
text = f"已切换到 {mode} 模式。" text = f"已切换到 {mode} 模式。"
@ -138,11 +136,11 @@ def dialogue_page(api: ApiRequest, is_lite: bool = False):
st.toast(text) st.toast(text)
dialogue_modes = ["LLM 对话", dialogue_modes = ["LLM 对话",
"知识库问答", "知识库问答",
"文件对话", "文件对话",
"搜索引擎问答", "搜索引擎问答",
"自定义Agent问答", "自定义Agent问答",
] ]
dialogue_mode = st.selectbox("请选择对话模式：", dialogue_mode = st.selectbox("请选择对话模式：",
dialogue_modes, dialogue_modes,
index=0, index=0,
@ -166,9 +164,9 @@ def dialogue_page(api: ApiRequest, is_lite: bool = False):
available_models = [] available_models = []
config_models = api.list_config_models() config_models = api.list_config_models()
if not is_lite: if not is_lite:
for k, v in config_models.get("local", {}).items(): # 列出配置了有效本地路径的模型 for k, v in config_models.get("local", {}).items(): # 列出配置了有效本地路径的模型
if (v.get("model_path_exists") if (v.get("model_path_exists")
and k not in running_models): and k not in running_models):
available_models.append(k) available_models.append(k)
for k, v in config_models.get("online", {}).items(): # 列出ONLINE_MODELS中直接访问的模型 for k, v in config_models.get("online", {}).items(): # 列出ONLINE_MODELS中直接访问的模型
if not v.get("provider") and k not in running_models: if not v.get("provider") and k not in running_models:
@ -250,14 +248,14 @@ def dialogue_page(api: ApiRequest, is_lite: bool = False):
elif dialogue_mode == "文件对话": elif dialogue_mode == "文件对话":
with st.expander("文件对话配置", True): with st.expander("文件对话配置", True):
files = st.file_uploader("上传知识文件：", files = st.file_uploader("上传知识文件：",
[i for ls in LOADER_DICT.values() for i in ls], [i for ls in LOADER_DICT.values() for i in ls],
accept_multiple_files=True, accept_multiple_files=True,
) )
kb_top_k = st.number_input("匹配知识条数：", 1, 20, VECTOR_SEARCH_TOP_K) kb_top_k = st.number_input("匹配知识条数：", 1, 20, VECTOR_SEARCH_TOP_K)
## Bge 模型会超过1 ## Bge 模型会超过1
score_threshold = st.slider("知识匹配分数阈值：", 0.0, 2.0, float(SCORE_THRESHOLD), 0.01) score_threshold = st.slider("知识匹配分数阈值：", 0.0, 2.0, float(SCORE_THRESHOLD), 0.01)
if st.button("开始上传", disabled=len(files)==0): if st.button("开始上传", disabled=len(files) == 0):
st.session_state["file_chat_id"] = upload_temp_docs(files, api) st.session_state["file_chat_id"] = upload_temp_docs(files, api)
elif dialogue_mode == "搜索引擎问答": elif dialogue_mode == "搜索引擎问答":
search_engine_list = api.list_search_engines() search_engine_list = api.list_search_engines()
@ -279,9 +277,9 @@ def dialogue_page(api: ApiRequest, is_lite: bool = False):
chat_input_placeholder = "请输入对话内容，换行请使用Shift+Enter。输入/help查看自定义命令 " chat_input_placeholder = "请输入对话内容，换行请使用Shift+Enter。输入/help查看自定义命令 "
def on_feedback( def on_feedback(
feedback, feedback,
message_id: str = "", message_id: str = "",
history_index: int = -1, history_index: int = -1,
): ):
reason = feedback["text"] reason = feedback["text"]
score_int = chat_box.set_feedback(feedback=feedback, history_index=history_index) score_int = chat_box.set_feedback(feedback=feedback, history_index=history_index)
@ -296,7 +294,7 @@ def dialogue_page(api: ApiRequest, is_lite: bool = False):
} }
if prompt := st.chat_input(chat_input_placeholder, key="prompt"): if prompt := st.chat_input(chat_input_placeholder, key="prompt"):
if parse_command(text=prompt, modal=modal): # 用户输入自定义命令 if parse_command(text=prompt, modal=modal): # 用户输入自定义命令
st.rerun() st.rerun()
else: else:
history = get_messages_history(history_len) history = get_messages_history(history_len)
@ -306,11 +304,11 @@ def dialogue_page(api: ApiRequest, is_lite: bool = False):
text = "" text = ""
message_id = "" message_id = ""
r = api.chat_chat(prompt, r = api.chat_chat(prompt,
history=history, history=history,
conversation_id=conversation_id, conversation_id=conversation_id,
model=llm_model, model=llm_model,
prompt_name=prompt_template_name, prompt_name=prompt_template_name,
temperature=temperature) temperature=temperature)
for t in r: for t in r:
if error_msg := check_error_msg(t): # check whether error occured if error_msg := check_error_msg(t): # check whether error occured
st.error(error_msg) st.error(error_msg)
@ -321,12 +319,12 @@ def dialogue_page(api: ApiRequest, is_lite: bool = False):
metadata = { metadata = {
"message_id": message_id, "message_id": message_id,
} }
chat_box.update_msg(text, streaming=False, metadata=metadata) # 更新最终的字符串，去除光标 chat_box.update_msg(text, streaming=False, metadata=metadata) # 更新最终的字符串，去除光标
chat_box.show_feedback(**feedback_kwargs, chat_box.show_feedback(**feedback_kwargs,
key=message_id, key=message_id,
on_submit=on_feedback, on_submit=on_feedback,
kwargs={"message_id": message_id, "history_index": len(chat_box.history) - 1}) kwargs={"message_id": message_id, "history_index": len(chat_box.history) - 1})
elif dialogue_mode == "自定义Agent问答": elif dialogue_mode == "自定义Agent问答":
if not any(agent in llm_model for agent in SUPPORT_AGENT_MODEL): if not any(agent in llm_model for agent in SUPPORT_AGENT_MODEL):
@ -373,13 +371,13 @@ def dialogue_page(api: ApiRequest, is_lite: bool = False):
]) ])
text = "" text = ""
for d in api.knowledge_base_chat(prompt, for d in api.knowledge_base_chat(prompt,
knowledge_base_name=selected_kb, knowledge_base_name=selected_kb,
top_k=kb_top_k, top_k=kb_top_k,
score_threshold=score_threshold, score_threshold=score_threshold,
history=history, history=history,
model=llm_model, model=llm_model,
prompt_name=prompt_template_name, prompt_name=prompt_template_name,
temperature=temperature): temperature=temperature):
if error_msg := check_error_msg(d): # check whether error occured if error_msg := check_error_msg(d): # check whether error occured
st.error(error_msg) st.error(error_msg)
elif chunk := d.get("answer"): elif chunk := d.get("answer"):
@ -397,13 +395,13 @@ def dialogue_page(api: ApiRequest, is_lite: bool = False):
]) ])
text = "" text = ""
for d in api.file_chat(prompt, for d in api.file_chat(prompt,
knowledge_id=st.session_state["file_chat_id"], knowledge_id=st.session_state["file_chat_id"],
top_k=kb_top_k, top_k=kb_top_k,
score_threshold=score_threshold, score_threshold=score_threshold,
history=history, history=history,
model=llm_model, model=llm_model,
prompt_name=prompt_template_name, prompt_name=prompt_template_name,
temperature=temperature): temperature=temperature):
if error_msg := check_error_msg(d): # check whether error occured if error_msg := check_error_msg(d): # check whether error occured
st.error(error_msg) st.error(error_msg)
elif chunk := d.get("answer"): elif chunk := d.get("answer"):

117
webui_pages/knowledge_base/knowledge_base.py
View File

@ -7,15 +7,12 @@ from server.knowledge_base.utils import get_file_path, LOADER_DICT
from server.knowledge_base.kb_service.base import get_kb_details, get_kb_file_details from server.knowledge_base.kb_service.base import get_kb_details, get_kb_file_details
from typing import Literal, Dict, Tuple from typing import Literal, Dict, Tuple
from configs import (kbs_config, from configs import (kbs_config,
EMBEDDING_MODEL, DEFAULT_VS_TYPE, EMBEDDING_MODEL, DEFAULT_VS_TYPE,
CHUNK_SIZE, OVERLAP_SIZE, ZH_TITLE_ENHANCE) CHUNK_SIZE, OVERLAP_SIZE, ZH_TITLE_ENHANCE)
from server.utils import list_embed_models, list_online_embed_models from server.utils import list_embed_models, list_online_embed_models
import os import os
import time import time
# SENTENCE_SIZE = 100
cell_renderer = JsCode("""function(params) {if(params.value==true){return '✓'}else{return '×'}}""") cell_renderer = JsCode("""function(params) {if(params.value==true){return '✓'}else{return '×'}}""")
@ -32,7 +29,7 @@ def config_aggrid(
gb.configure_selection( gb.configure_selection(
selection_mode=selection_mode, selection_mode=selection_mode,
use_checkbox=use_checkbox, use_checkbox=use_checkbox,
# pre_selected_rows=st.session_state.get("selected_rows", [0]), pre_selected_rows=st.session_state.get("selected_rows", [0]),
) )
gb.configure_pagination( gb.configure_pagination(
enabled=True, enabled=True,
@ -59,7 +56,8 @@ def knowledge_base_page(api: ApiRequest, is_lite: bool = None):
try: try:
kb_list = {x["kb_name"]: x for x in get_kb_details()} kb_list = {x["kb_name"]: x for x in get_kb_details()}
except Exception as e: except Exception as e:
st.error("获取知识库信息错误，请检查是否已按照 `README.md` 中 `4 知识库初始化与迁移` 步骤完成初始化或迁移，或是否为数据库连接错误。") st.error(
"获取知识库信息错误，请检查是否已按照 `README.md` 中 `4 知识库初始化与迁移` 步骤完成初始化或迁移，或是否为数据库连接错误。")
st.stop() st.stop()
kb_names = list(kb_list.keys()) kb_names = list(kb_list.keys())
@ -150,7 +148,8 @@ def knowledge_base_page(api: ApiRequest, is_lite: bool = None):
[i for ls in LOADER_DICT.values() for i in ls], [i for ls in LOADER_DICT.values() for i in ls],
accept_multiple_files=True, accept_multiple_files=True,
) )
kb_info = st.text_area("请输入知识库介绍:", value=st.session_state["selected_kb_info"], max_chars=None, key=None, kb_info = st.text_area("请输入知识库介绍:", value=st.session_state["selected_kb_info"], max_chars=None,
key=None,
help=None, on_change=None, args=None, kwargs=None) help=None, on_change=None, args=None, kwargs=None)
if kb_info != st.session_state["selected_kb_info"]: if kb_info != st.session_state["selected_kb_info"]:
@ -200,8 +199,8 @@ def knowledge_base_page(api: ApiRequest, is_lite: bool = None):
doc_details = doc_details[[ doc_details = doc_details[[
"No", "file_name", "document_loader", "text_splitter", "docs_count", "in_folder", "in_db", "No", "file_name", "document_loader", "text_splitter", "docs_count", "in_folder", "in_db",
]] ]]
# doc_details["in_folder"] = doc_details["in_folder"].replace(True, "✓").replace(False, "×") doc_details["in_folder"] = doc_details["in_folder"].replace(True, "✓").replace(False, "×")
# doc_details["in_db"] = doc_details["in_db"].replace(True, "✓").replace(False, "×") doc_details["in_db"] = doc_details["in_db"].replace(True, "✓").replace(False, "×")
gb = config_aggrid( gb = config_aggrid(
doc_details, doc_details,
{ {
@ -252,7 +251,8 @@ def knowledge_base_page(api: ApiRequest, is_lite: bool = None):
st.write() st.write()
# 将文件分词并加载到向量库中 # 将文件分词并加载到向量库中
if cols[1].button( if cols[1].button(
"重新添加至向量库" if selected_rows and (pd.DataFrame(selected_rows)["in_db"]).any() else "添加至向量库", "重新添加至向量库" if selected_rows and (
pd.DataFrame(selected_rows)["in_db"]).any() else "添加至向量库",
disabled=not file_exists(kb, selected_rows)[0], disabled=not file_exists(kb, selected_rows)[0],
use_container_width=True, use_container_width=True,
): ):
@ -285,39 +285,39 @@ def knowledge_base_page(api: ApiRequest, is_lite: bool = None):
st.divider() st.divider()
# cols = st.columns(3) cols = st.columns(3)
# if cols[0].button( if cols[0].button(
# "依据源文件重建向量库", "依据源文件重建向量库",
# # help="无需上传文件，通过其它方式将文档拷贝到对应知识库content目录下，点击本按钮即可重建知识库。", help="无需上传文件，通过其它方式将文档拷贝到对应知识库content目录下，点击本按钮即可重建知识库。",
# use_container_width=True, use_container_width=True,
# type="primary", type="primary",
# ): ):
# with st.spinner("向量库重构中，请耐心等待，勿刷新或关闭页面。"): with st.spinner("向量库重构中，请耐心等待，勿刷新或关闭页面。"):
# empty = st.empty() empty = st.empty()
# empty.progress(0.0, "") empty.progress(0.0, "")
# for d in api.recreate_vector_store(kb, for d in api.recreate_vector_store(kb,
# chunk_size=chunk_size, chunk_size=chunk_size,
# chunk_overlap=chunk_overlap, chunk_overlap=chunk_overlap,
# zh_title_enhance=zh_title_enhance): zh_title_enhance=zh_title_enhance):
# if msg := check_error_msg(d): if msg := check_error_msg(d):
# st.toast(msg) st.toast(msg)
# else: else:
# empty.progress(d["finished"] / d["total"], d["msg"]) empty.progress(d["finished"] / d["total"], d["msg"])
# st.rerun() st.rerun()
# if cols[2].button( if cols[2].button(
# "删除知识库", "删除知识库",
# use_container_width=True, use_container_width=True,
# ): ):
# ret = api.delete_knowledge_base(kb) ret = api.delete_knowledge_base(kb)
# st.toast(ret.get("msg", " ")) st.toast(ret.get("msg", " "))
# time.sleep(1) time.sleep(1)
# st.rerun() st.rerun()
# with st.sidebar: with st.sidebar:
# keyword = st.text_input("查询关键字") keyword = st.text_input("查询关键字")
# top_k = st.slider("匹配条数", 1, 100, 3) top_k = st.slider("匹配条数", 1, 100, 3)
st.write("文件内文档列表。双击进行修改，在删除列填入 Y 可删除对应行。") st.write("文件内文档列表。双击进行修改，在删除列填入 Y 可删除对应行。")
docs = [] docs = []
@ -325,11 +325,12 @@ def knowledge_base_page(api: ApiRequest, is_lite: bool = None):
if selected_rows: if selected_rows:
file_name = selected_rows[0]["file_name"] file_name = selected_rows[0]["file_name"]
docs = api.search_kb_docs(knowledge_base_name=selected_kb, file_name=file_name) docs = api.search_kb_docs(knowledge_base_name=selected_kb, file_name=file_name)
data = [{"seq": i+1, "id": x["id"], "page_content": x["page_content"], "source": x["metadata"].get("source"), data = [
"type": x["type"], {"seq": i + 1, "id": x["id"], "page_content": x["page_content"], "source": x["metadata"].get("source"),
"metadata": json.dumps(x["metadata"], ensure_ascii=False), "type": x["type"],
"to_del": "", "metadata": json.dumps(x["metadata"], ensure_ascii=False),
} for i, x in enumerate(docs)] "to_del": "",
} for i, x in enumerate(docs)]
df = pd.DataFrame(data) df = pd.DataFrame(data)
gb = GridOptionsBuilder.from_dataframe(df) gb = GridOptionsBuilder.from_dataframe(df)
@ -343,22 +344,24 @@ def knowledge_base_page(api: ApiRequest, is_lite: bool = None):
edit_docs = AgGrid(df, gb.build()) edit_docs = AgGrid(df, gb.build())
if st.button("保存更改"): if st.button("保存更改"):
# origin_docs = {x["id"]: {"page_content": x["page_content"], "type": x["type"], "metadata": x["metadata"]} for x in docs} origin_docs = {
x["id"]: {"page_content": x["page_content"], "type": x["type"], "metadata": x["metadata"]} for x in
docs}
changed_docs = [] changed_docs = []
for index, row in edit_docs.data.iterrows(): for index, row in edit_docs.data.iterrows():
# origin_doc = origin_docs[row["id"]] origin_doc = origin_docs[row["id"]]
# if row["page_content"] != origin_doc["page_content"]: if row["page_content"] != origin_doc["page_content"]:
if row["to_del"] not in ["Y", "y", 1]: if row["to_del"] not in ["Y", "y", 1]:
changed_docs.append({ changed_docs.append({
"page_content": row["page_content"], "page_content": row["page_content"],
"type": row["type"], "type": row["type"],
"metadata": json.loads(row["metadata"]), "metadata": json.loads(row["metadata"]),
}) })
if changed_docs: if changed_docs:
if api.update_kb_docs(knowledge_base_name=selected_kb, if api.update_kb_docs(knowledge_base_name=selected_kb,
file_names=[file_name], file_names=[file_name],
docs={file_name: changed_docs}): docs={file_name: changed_docs}):
st.toast("更新文档成功") st.toast("更新文档成功")
else: else:
st.toast("更新文档失败") st.toast("更新文档失败")