Skip to content

README.md

Latest commit

 

History

History
35 lines (24 loc) · 1.02 KB

README.md

File metadata and controls

35 lines (24 loc) · 1.02 KB
Raw

Reviews MCP Server

Model Context Protocol (MCP) server exposing review retrieval as tools for AI agents.

Purpose

Separate MCP server for reviews (vs items). Agents can call tools from multiple servers—items for products, reviews for customer feedback. Used by notebooks/week4/04-MCP.ipynb.

Architecture

reviews_mcp_server/
├── src/reviews_mcp_server/
│   ├── main.py      # FastMCP app, @mcp.tool() get_formatted_reviews_context
│   ├── utils.py     # retrieve_reviews_data, process_reviews_context (Qdrant)
│   └── core/config.py
├── Dockerfile
└── pyproject.toml

Tool

  • get_formatted_reviews_context(query, item_list, top_k=15): Returns formatted reviews from Amazon-items-collection-01-reviews, prefiltered by item IDs. Same logic as api.agents.tools.get_formatted_reviews_context.

Running

# Via Docker Compose
docker compose up reviews_mcp_server

# Standalone
uv run python -m reviews_mcp_server.main

Port: 8002 (host) → 8000 (container)