学习mcp—in-action

基于MCP的rag系统

https://github.com/FlyAIBox/mcp-in-action/tree/rag_0.1.1/mcp-rag 基于MCP的RAG系统主要包含三个核心部分:

  1. 1. 知识库服务(MCP Server):基于Milvus向量数据库实现的后端服务,负责文档存储和检索

  2. 2. 客户端工具(MCP Client):与MCP Server通信的客户端,实现知识库的构建和检索功能

  3. 3. 大模型集成:通过LLM实现文档切分、FAQ提取、问题拆解和回答生成等核心功能 启动步骤如下所示 CodeSnap_2025-05-29_at_110956.png 拷贝.env.template 为.env文件

1
2
3
4
5
6
MILVUS_HOST=localhost
MILVUS_PORT=19530
EMBEDDING_MODEL=all-MiniLM-L6-v2
KNOWLEDGE_COLLECTION=knowledge_store
FAQ_COLLECTION=faq_store
VECTOR_DIMENSION=384

接下来就是启动这个项目了

1
2
export HF_ENDPOINT=https://hf-mirror.com
python -m app.main

如果一切正常,你会得到下面的输出

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
INFO:     Started server process [76521]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:8080 (Press CTRL+C to quit)
INFO:     127.0.0.1:42300 - "GET /sse HTTP/1.1" 200 OK
INFO:     127.0.0.1:57392 - "GET /api/v1 HTTP/1.1" 404 Not Found
INFO:     127.0.0.1:57392 - "GET /api/v1 HTTP/1.1" 404 Not Found
INFO:     127.0.0.1:44854 - "GET /api/v1 HTTP/1.1" 404 Not Found
INFO:     127.0.0.1:44854 - "GET /api/v1 HTTP/1.1" 404 Not Found
INFO:     127.0.0.1:38692 - "GET /sse HTTP/1.1" 200 OK

验证服务器是否启动成功: http://localhost:8080/sse image.png 接下来本地启动sse的调试模式

1
2
  # 启动 MCP Inspector
  npx @modelcontextprotocol/inspector node build/index.js

image.png 打开·http://localhost:6277· image.png 提问,可以参考https://github.com/FlyAIBox/mcp-in-action/blob/main/mcp-rag/milvus-mcp-server/MCP-Tools-%E6%B5%8B%E8%AF%95%E6%96%87%E6%A1%A3.md 其中注意,json中的内容,需要分开对应到输入框中,不然会报错的 image.png image.png

Licensed under CC BY-NC-SA 4.0
最后更新于 May 29, 2025 06:06 UTC
comments powered by Disqus
Built with Hugo
主题 StackJimmy 设计
Caret Up