Skip to content

Commit 6203876

Browse files
author
SEO7077
committed
Update README.md
1 parent cd8822b commit 6203876

File tree

1 file changed

+258
-1
lines changed

1 file changed

+258
-1
lines changed

‎README.md‎

Lines changed: 258 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,2 +1,259 @@
11
# GraphRAG4OpenWebUI
2-
GraphRAG4OpenWebUI integrates Microsoft's GraphRAG technology into Open WebUI, providing a versatile information retrieval API. It combines local, global, and web searches for advanced Q&A systems and search engines. This tool simplifies graph-based retrieval integration in open web environments.
2+
3+
GraphRAG4OpenWebUI 是一个专为 Open WebUI 设计的 API 接口，旨在集成微软研究院的 GraphRAG（Graph-based Retrieval-Augmented Generation）技术。该项目提供了一个强大的信息检索系统，支持多种搜索模型，特别适合在开放式 Web 用户界面中使用。
4+
5+
## 项目概述
6+
7+
本项目的主要目标是为 Open WebUI 提供一个便捷的接口，以利用 GraphRAG 的强大功能。它集成了三种主要的检索方法，并提供了一个综合搜索选项，使用户能够获得全面而精确的搜索结果。
8+
9+
### 主要检索功能
10+
11+
1. **本地搜索（Local Search）**
12+
- 利用 GraphRAG 技术在本地知识库中进行高效检索
13+
- 适用于快速访问预先定义的结构化信息
14+
- 利用图结构提高检索的准确性和相关性
15+
16+
2. **全局搜索（Global Search）**
17+
- 在更广泛的范围内搜索信息，超越本地知识库的限制
18+
- 适用于需要更全面信息的查询
19+
- 利用 GraphRAG 的全局上下文理解能力，提供更丰富的搜索结果
20+
21+
3. **Tavily 搜索**
22+
- 集成外部 Tavily 搜索 API
23+
- 提供额外的互联网搜索能力，扩展信息源
24+
- 适用于需要最新或广泛网络信息的查询
25+
26+
4. **全模型搜索（Full Model Search）**
27+
- 综合上述三种搜索方法
28+
- 提供最全面的搜索结果，满足复杂的信息需求
29+
- 自动整合和排序来自不同来源的信息
30+
31+
## 安装
32+
33+
确保你的系统中已安装 Python 3.8 或更高版本。然后，按照以下步骤安装：
34+
35+
1. 克隆仓库：
36+
37+
```
38+
git clone https://github.com/your-username/GraphRAG4OpenWebUI.git
39+
cd GraphRAG4OpenWebUI
40+
```
41+
42+
2. 创建并激活虚拟环境：
43+
44+
```
45+
python -m venv venv
46+
source venv/bin/activate # 在 Windows 上使用 venv\Scripts\activate
47+
```
48+
49+
3. 安装依赖：
50+
51+
```
52+
pip install fastapi uvicorn pandas tiktoken graphrag tavily-python pydantic python-dotenv asyncio aiohttp numpy scikit-learn matplotlib seaborn nltk spacy transformers torch torchvision torchaudio
53+
```
54+
55+
注意：`graphrag` 包可能需要从特定的源安装。如果上述命令无法安装 `graphrag`，请参考微软研究院的具体说明或联系维护者获取正确的安装方法。
56+
57+
## 配置
58+
59+
在运行 API 之前，需要设置以下环境变量。你可以通过创建 `.env` 文件或直接在终端中导出这些变量：
60+
61+
```bash
62+
export GRAPHRAG_API_KEY="your_graphrag_api_key"
63+
export TAVILY_API_KEY="your_tavily_api_key"
64+
export GRAPHRAG_LLM_MODEL="gpt-3.5-turbo"
65+
export GRAPHRAG_EMBEDDING_MODEL="text-embedding-3-small"
66+
export INPUT_DIR="/path/to/your/input/directory"
67+
```
68+
69+
确保将上述命令中的占位符替换为实际的 API 密钥和路径。
70+
71+
## 使用方法
72+
73+
1. 启动服务器：
74+
75+
```
76+
python main.py
77+
```
78+
79+
服务器将在 `http://localhost:8012` 上运行。
80+
81+
2. API 端点：
82+
83+
- `/v1/chat/completions`: POST 请求，用于执行搜索
84+
- `/v1/models`: GET 请求，获取可用模型列表
85+
86+
3. 在 Open WebUI 中集成：
87+
88+
在 Open WebUI 的配置中，将 API 端点设置为 `http://localhost:8012/v1/chat/completions`。这将允许 Open WebUI 使用 GraphRAG4OpenWebUI 的搜索功能。
89+
90+
4. 发送搜索请求示例：
91+
92+
```python
93+
import requests
94+
import json
95+
96+
url = "http://localhost:8012/v1/chat/completions"
97+
headers = {"Content-Type": "application/json"}
98+
data = {
99+
"model": "full-model:latest",
100+
"messages": [{"role": "user", "content": "你的搜索查询"}],
101+
"temperature": 0.7
102+
}
103+
104+
response = requests.post(url, headers=headers, data=json.dumps(data))
105+
print(response.json())
106+
```
107+
108+
## 可用模型
109+
110+
- `graphrag-local-search:latest`: 本地搜索
111+
- `graphrag-global-search:latest`: 全局搜索
112+
- `tavily-search:latest`: Tavily 搜索
113+
- `full-model:latest`: 综合搜索（包含上述所有搜索方法）
114+
115+
## 注意事项
116+
117+
- 确保在 `INPUT_DIR` 目录中有正确的输入文件（如 Parquet 文件）。
118+
- API 使用异步编程，确保你的环境支持异步操作。
119+
- 对于大规模部署，考虑使用生产级的 ASGI 服务器。
120+
- 本项目专为 Open WebUI 设计，可以轻松集成到各种基于 Web 的应用中。
121+
122+
## 贡献
123+
124+
欢迎提交 Pull Requests 来改进这个项目。对于重大变更，请先开 issue 讨论你想要改变的内容。
125+
126+
## 许可证
127+
128+
[MIT License](LICENSE)
129+
130+
---
131+
132+
# GraphRAG4OpenWebUI
133+
134+
GraphRAG4OpenWebUI is an API interface specifically designed for Open WebUI, aiming to integrate Microsoft Research's GraphRAG (Graph-based Retrieval-Augmented Generation) technology. This project provides a powerful information retrieval system that supports multiple search models, particularly suitable for use in open web user interfaces.
135+
136+
## Project Overview
137+
138+
The main goal of this project is to provide a convenient interface for Open WebUI to leverage the powerful features of GraphRAG. It integrates three main retrieval methods and offers a comprehensive search option, allowing users to obtain thorough and precise search results.
139+
140+
### Key Retrieval Features
141+
142+
1. **Local Search**
143+
- Utilizes GraphRAG technology for efficient retrieval in local knowledge bases
144+
- Suitable for quick access to pre-defined structured information
145+
- Leverages graph structures to improve retrieval accuracy and relevance
146+
147+
2. **Global Search**
148+
- Searches for information in a broader scope, beyond local knowledge bases
149+
- Suitable for queries requiring more comprehensive information
150+
- Utilizes GraphRAG's global context understanding capabilities to provide richer search results
151+
152+
3. **Tavily Search**
153+
- Integrates external Tavily search API
154+
- Provides additional internet search capabilities, expanding information sources
155+
- Suitable for queries requiring the latest or extensive web information
156+
157+
4. **Full Model Search**
158+
- Combines all three search methods above
159+
- Provides the most comprehensive search results, meeting complex information needs
160+
- Automatically integrates and ranks information from different sources
161+
162+
## Installation
163+
164+
Ensure that you have Python 3.8 or higher installed on your system. Then, follow these steps to install:
165+
166+
1. Clone the repository:
167+
168+
```
169+
git clone https://github.com/your-username/GraphRAG4OpenWebUI.git
170+
cd GraphRAG4OpenWebUI
171+
```
172+
173+
2. Create and activate a virtual environment:
174+
175+
```
176+
python -m venv venv
177+
source venv/bin/activate # On Windows use venv\Scripts\activate
178+
```
179+
180+
3. Install dependencies:
181+
182+
```
183+
pip install fastapi uvicorn pandas tiktoken graphrag tavily-python pydantic python-dotenv asyncio aiohttp numpy scikit-learn matplotlib seaborn nltk spacy transformers torch torchvision torchaudio
184+
```
185+
186+
Note: The `graphrag` package might need to be installed from a specific source. If the above command fails to install `graphrag`, please refer to Microsoft Research's specific instructions or contact the maintainer for the correct installation method.
187+
188+
## Configuration
189+
190+
Before running the API, you need to set the following environment variables. You can do this by creating a `.env` file or exporting them directly in your terminal:
191+
192+
```bash
193+
export GRAPHRAG_API_KEY="your_graphrag_api_key"
194+
export TAVILY_API_KEY="your_tavily_api_key"
195+
export GRAPHRAG_LLM_MODEL="gpt-3.5-turbo"
196+
export GRAPHRAG_EMBEDDING_MODEL="text-embedding-3-small"
197+
export INPUT_DIR="/path/to/your/input/directory"
198+
```
199+
200+
Make sure to replace the placeholders in the above commands with your actual API keys and paths.
201+
202+
## Usage
203+
204+
1. Start the server:
205+
206+
```
207+
python main.py
208+
```
209+
210+
The server will run on `http://localhost:8012`.
211+
212+
2. API Endpoints:
213+
214+
- `/v1/chat/completions`: POST request for performing searches
215+
- `/v1/models`: GET request to retrieve the list of available models
216+
217+
3. Integration with Open WebUI:
218+
219+
In the Open WebUI configuration, set the API endpoint to `http://localhost:8012/v1/chat/completions`. This will allow Open WebUI to use the search functionality of GraphRAG4OpenWebUI.
220+
221+
4. Example search request:
222+
223+
```python
224+
import requests
225+
import json
226+
227+
url = "http://localhost:8012/v1/chat/completions"
228+
headers = {"Content-Type": "application/json"}
229+
data = {
230+
"model": "full-model:latest",
231+
"messages": [{"role": "user", "content": "Your search query"}],
232+
"temperature": 0.7
233+
}
234+
235+
response = requests.post(url, headers=headers, data=json.dumps(data))
236+
print(response.json())
237+
```
238+
239+
## Available Models
240+
241+
- `graphrag-local-search:latest`: Local search
242+
- `graphrag-global-search:latest`: Global search
243+
- `tavily-search:latest`: Tavily search
244+
- `full-model:latest`: Comprehensive search (includes all search methods above)
245+
246+
## Notes
247+
248+
- Ensure that you have the correct input files (such as Parquet files) in the `INPUT_DIR` directory.
249+
- The API uses asynchronous programming, make sure your environment supports async operations.
250+
- For large-scale deployment, consider using a production-grade ASGI server.
251+
- This project is specifically designed for Open WebUI and can be easily integrated into various web-based applications.
252+
253+
## Contributing
254+
255+
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.
256+
257+
## License
258+
259+
[MIT License](LICENSE)

0 commit comments

Comments
 (0)