2023-04-17 16:25:49 +08:00
|
|
|
|
# Chinese-LangChain
|
2023-04-17 16:20:32 +08:00
|
|
|
|
|
2023-04-17 16:25:49 +08:00
|
|
|
|
> Chinese-LangChain:中文langchain项目,基于ChatGLM-6b+langchain实现本地化知识库检索与智能答案生成
|
2023-04-17 16:20:32 +08:00
|
|
|
|
|
2023-04-19 02:25:15 +08:00
|
|
|
|
https://github.com/yanqiangmiffy/Chinese-LangChain
|
|
|
|
|
|
2023-04-19 01:15:22 +08:00
|
|
|
|
俗称:小必应,Q.Talk,强聊,QiangTalk
|
|
|
|
|
|
2023-04-17 16:27:43 +08:00
|
|
|
|
## 🔥 效果演示
|
2023-04-17 18:03:26 +08:00
|
|
|
|
|
2023-04-18 17:44:03 +08:00
|
|
|
|
![](https://github.com/yanqiangmiffy/Chinese-LangChain/blob/master/images/web_demo.png)
|
2023-04-19 01:48:10 +08:00
|
|
|
|
![](https://github.com/yanqiangmiffy/Chinese-LangChain/blob/master/images/web_demo_new.png)
|
2023-04-17 16:27:43 +08:00
|
|
|
|
|
2023-04-19 13:14:42 +08:00
|
|
|
|
## 🚋 使用教程
|
|
|
|
|
|
|
|
|
|
- 选择知识库询问相关领域的问题
|
|
|
|
|
|
|
|
|
|
## 🏗️ 部署教程
|
|
|
|
|
|
2023-04-17 16:25:49 +08:00
|
|
|
|
## 🚀 特性
|
2023-04-18 23:45:16 +08:00
|
|
|
|
|
2023-04-19 13:14:42 +08:00
|
|
|
|
- 📝 2023/04/19 发布45万Wikipedia的文本预处理语料以及FAISS索引向量
|
2023-04-19 02:09:00 +08:00
|
|
|
|
- 🐯 2023/04/19 引入ChuanhuChatGPT皮肤
|
2023-04-19 10:58:10 +08:00
|
|
|
|
- 📱 2023/04/19 增加web search功能,需要确保网络畅通!(感谢[@wanghao07456](https://github.com/wanghao07456),提供的idea)
|
2023-04-19 02:09:00 +08:00
|
|
|
|
- 📚 2023/04/18 webui增加知识库选择功能
|
2023-04-18 21:29:34 +08:00
|
|
|
|
- 🚀 2023/04/18 修复推理预测超时5s报错问题
|
2023-04-18 23:45:16 +08:00
|
|
|
|
- 🎉 2023/04/17 支持多种文档上传与内容解析:pdf、docx,ppt等
|
2023-04-18 21:29:34 +08:00
|
|
|
|
- 🎉 2023/04/17 支持知识增量更新
|
2023-04-17 16:20:32 +08:00
|
|
|
|
|
|
|
|
|
[//]: # (- 支持检索结果与LLM生成结果对比)
|
|
|
|
|
|
2023-04-18 23:45:16 +08:00
|
|
|
|
## 🧰 知识库
|
|
|
|
|
|
2023-04-19 13:14:42 +08:00
|
|
|
|
### 构建知识库
|
|
|
|
|
|
|
|
|
|
- Wikipedia-zh
|
|
|
|
|
|
|
|
|
|
> 详情见:corpus/zh_wikipedia/README.md
|
|
|
|
|
|
|
|
|
|
### 知识库向量索引
|
|
|
|
|
|
2023-04-18 23:45:16 +08:00
|
|
|
|
| 知识库数据 |FAISS向量|
|
|
|
|
|
|--------|----|
|
|
|
|
|
|💹 [大规模金融研报知识图谱](http://openkg.cn/dataset/fr2kg)|链接:https://pan.baidu.com/s/1FcIH5Fi3EfpS346DnDu51Q?pwd=ujjv 提取码:ujjv |
|
|
|
|
|
|
2023-04-17 16:25:49 +08:00
|
|
|
|
## 🔨 TODO
|
2023-04-17 18:03:26 +08:00
|
|
|
|
|
2023-04-17 16:25:49 +08:00
|
|
|
|
* [x] 支持上下文
|
|
|
|
|
* [x] 支持知识增量更新
|
2023-04-18 23:57:05 +08:00
|
|
|
|
* [x] 支持加载不同知识库
|
2023-04-17 17:31:54 +08:00
|
|
|
|
* [x] 支持检索结果与LLM生成结果对比
|
|
|
|
|
* [ ] 支持检索生成结果与原始LLM生成结果对比
|
2023-04-17 18:03:26 +08:00
|
|
|
|
* [ ] 检索结果过滤与排序
|
2023-04-19 01:15:22 +08:00
|
|
|
|
* [x] 互联网检索结果接入
|
2023-04-17 18:03:26 +08:00
|
|
|
|
* [ ] 模型初始化有问题
|
2023-04-18 17:44:03 +08:00
|
|
|
|
* [ ] 增加非LangChain策略
|
2023-04-17 18:03:26 +08:00
|
|
|
|
|
2023-04-17 17:46:50 +08:00
|
|
|
|
## 交流
|
2023-04-18 23:45:16 +08:00
|
|
|
|
|
2023-04-19 13:18:01 +08:00
|
|
|
|
欢迎多提建议、Bad cases,目前尚不完善,欢迎进群及时交流,也欢迎大家多提PR</br>
|
2023-04-19 13:24:30 +08:00
|
|
|
|
|
2023-04-19 13:24:06 +08:00
|
|
|
|
<figure class="third">
|
|
|
|
|
<img src="https://raw.githubusercontent.com/yanqiangmiffy/Chinese-LangChain/master/images/ch.jpg" width="200px">
|
|
|
|
|
<img src="https://raw.githubusercontent.com/yanqiangmiffy/Chinese-LangChain/master/images/chatgroup.jpg" width="200px" height="270px">
|
|
|
|
|
<img src="https://raw.githubusercontent.com/yanqiangmiffy/Chinese-LangChain/master/images/personal.jpg" width="200px">
|
|
|
|
|
</figure>
|
2023-04-19 13:18:01 +08:00
|
|
|
|
|
|
|
|
|
|
2023-04-17 17:46:50 +08:00
|
|
|
|
|
2023-04-17 16:25:49 +08:00
|
|
|
|
## ❤️引用
|
2023-04-17 16:20:32 +08:00
|
|
|
|
|
|
|
|
|
- webui参考:https://github.com/thomas-yanxin/LangChain-ChatGLM-Webui
|
2023-04-17 16:30:54 +08:00
|
|
|
|
- knowledge问答参考:https://github.com/imClumsyPanda/langchain-ChatGLM
|
2023-04-17 16:28:13 +08:00
|
|
|
|
- LLM模型:https://github.com/THUDM/ChatGLM-6B
|
2023-04-19 02:25:15 +08:00
|
|
|
|
- CSS:https://huggingface.co/spaces/JohnSmith9982/ChuanhuChatGPT
|