mirror of
https://github.com/JasonYANG170/CodeGeeX4.git
synced 2024-11-27 14:16:35 +00:00
46 lines
984 B
Markdown
46 lines
984 B
Markdown
|
|
|||
|
|
|
|||
|
|
[English](README.md) | [中文](README_zh.md)
|
|||
|
|
|
|||
|
|
## RAG功能
|
|||
|
|
|
|||
|
|
CodeGeeX4支持RAG检索增强，并兼容LlamaIndex框架，实现项目级检索问答。
|
|||
|
|
|
|||
|
|
## 使用教程
|
|||
|
|
|
|||
|
|
### 1. 安装依赖项
|
|||
|
|
|
|||
|
|
```bash
|
|||
|
|
cd llamaindex_demo
|
|||
|
|
pip install -r requirements.txt
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
注：此项目使用到tree-sitter-language，其与python3.10兼容的有问题，因此请使用python3.8或python3.9运行该项目。
|
|||
|
|
|
|||
|
|
### 2. 配置Embedding API Key
|
|||
|
|
|
|||
|
|
本项目使用智谱开放平台的Embedding API实现向量化功能，请先注册并获取API Key。
|
|||
|
|
|
|||
|
|
并在`models/embedding.py`中配置API Key。
|
|||
|
|
|
|||
|
|
详情可参考 https://open.bigmodel.cn/dev/api#text_embedding
|
|||
|
|
|
|||
|
|
### 3. 生成向量数据
|
|||
|
|
|
|||
|
|
```bash
|
|||
|
|
python vectorize.py --workspace . --output_path vectors
|
|||
|
|
|
|||
|
|
>>> 文件向量化完成,已保存至vectors
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 4. 运行问答脚本
|
|||
|
|
|
|||
|
|
```bash
|
|||
|
|
python chat.py --vector_path vectors
|
|||
|
|
|
|||
|
|
>>> Running on local URL: http://127.0.0.1:8080
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
## Demo
|
|||
|
|
|
|||
|
|
|