RAG Document Server

# Quick Start Guide - Secure MCP Setup ## TL;DR - Add to Claude Code Copy this configuration (no API keys needed!): ```json { "mcpServers": { "rag-dagster-docs": { "command": "uv", "args": [ "--directory", "/home/ubuntu/work/rag", "run", "python", "-m", "mcp_server.server" ] } } } ``` ## How It Works Securely 1. **API Key Location:** `/home/ubuntu/work/rag/.env` (gitignored ✅) 2. **MCP Config:** No secrets, just command to run 3. **Auto-Loading:** Settings.py loads `.env` automatically via Pydantic ## Query Only Dagster Docs Always use tags to filter for Dagster documentation: ``` Use query_rag to ask "How do I define assets?" with tags: ["dagster"] ``` ## Available Tags - `dagster` - Main Dagster documentation (use this!) - `docs` - General documentation tag - `getting-started` - Getting started section - `tutorial` - Tutorial content ## System Status ✅ **543 documents** ingested ✅ **4,566 chunks** ready to search ✅ **API key** securely in `.env` ✅ **Hierarchical paths** preserved (e.g., `getting-started > Concepts > Asset`) ## MCP Tools | Tool | Use | |------|-----| | `query_rag` | Search docs with tags: ["dagster"] | | `get_rag_stats` | Check system status | | `get_tags` | List available tags | | `list_documents` | View all documents | ## Security Checklist - ✅ `.env` file in `.gitignore` - ✅ No API keys in MCP config JSON - ✅ Settings auto-load from `.env` - ✅ Google API key working ## Test It After adding to Claude Code MCP settings, try: ``` Use get_rag_stats to show system information ``` Should return: 543 documents, 4566 chunks Then: ``` Use query_rag with question "What is an asset?" and tags: ["dagster"] ``` Should return: Dagster asset definition with hierarchical source paths.