AI导哈

鸡尾酒食谱RAG MCP服务器

一个通过RAG从本地数据集检索鸡尾酒配方并生成推荐，为AI助手提供鸡尾酒知识查询的MCP服务器。

  • by 00200200
  • Jupyter Notebook
  • 本地部署
  • 6 星标
  • 更新于
在 GitHub 查看 00200200/cocktails-rag-mcp

使用说明

Cocktails RAG MCP Server

MCP server for looking up cocktails and suggesting recipes from a local dataset. It retrieves recipes with FAISS, reranks the matches with a Hugging Face model, and sends the selected context to Groq to generate an answer.

Requirements

  • Python 3.11+
  • uv package manager - https://docs.astral.sh/uv/getting-started/installation/

Quick Start

  1. Get a Groq API key: https://console.groq.com/keys

  2. Setup:

    # Clone the repository
git clone https://github.com/00200200/cocktails-rag-mcp.git
cd cocktails-rag-mcp

# Copy environment template
cp .env.example .env

# Edit .env and add your GROQ_API_KEY
nano .env

# Install dependencies
uv sync

  3. Initialize the pipeline:

    This loads the embedding and reranker models and creates or loads the FAISS index. Model weights are downloaded on first use; the Groq API key must be set.

    uv run python -c "from src.rag.rag import RAG; RAG(); print('Pipeline initialized')"

  4. Install for Claude Desktop:

    Automatic (Recommended)

    uv run fastmcp install claude-desktop fastmcp.json --name cocktails --env-file .env

    Manual

    Edit config file:

    • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
    • Windows: %APPDATA%\Claude\claude_desktop_config.json
    {
	"mcpServers": {
		"cocktails": {
			"command": "uv",
			"args": [
				"run",
				"--with","faiss-cpu",
				"--with","fastmcp",
				"--with","jq",
				"--with","langchain",
				"--with","langchain-community",
				"--with","langchain-groq",
				"--with","langchain-huggingface",
				"--with","pandas",
				"--with","python-dotenv",
				"--with","sentence-transformers",
				"fastmcp",
				"run",
				"/ABSOLUTE/PATH/TO/src/mcp/server.py:mcp"
			],
			"env": {
				"GROQ_API_KEY": "your_groq_api_key_here"
			}
		}
	}
}

    Replace /ABSOLUTE/PATH/TO/ with your project path and GROQ_API_KEY with your API key.

Example Usage

Local Testing

# Test RAG pipeline directly
uv run python -m src.rag.rag

# Test MCP server locally
uv run python src/mcp/server.py

Development

Code Formatting

# Format code with black
uv tool run black .

# Sort imports with isort
uv tool run isort .

Project Structure

RAG/
├── src/
│   ├── mcp/         # MCP server implementation (FastMCP)
│   ├── rag/         # RAG pipeline (retrieve, rerank, generate)
│   ├── db/          # FAISS vector database handler
│   └── data/        # Data loading utilities
├── data/            # Cocktail dataset
├── faiss_index/     # Generated FAISS index (auto-created on first run)
├── notebooks/       # EDA notebook
├── fastmcp.json     # FastMCP configuration
├── pyproject.toml   # Project dependencies
└── .env.example     # Environment template

Tech Stack

  • MCP Framework: FastMCP
  • RAG Framework: LangChain
  • Embeddings: BAAI/bge-m3 (local via HuggingFace)
  • Vector DB: FAISS (local)
  • Reranker: BAAI/bge-reranker-v2-m3 (local via HuggingFace)
  • LLM: Groq API (llama-3.1-8b-instant)
  • Package Manager: uv

相关 MCP