Personal RAG

You can run a Personal RAG MCP Server that stores your notes and documents locally, builds semantic search indexes, and serves questions with retrieval-augmented generation across multiple clients. It supports both stdio and HTTP streaming transports, works with SQLite for documents and Qdrant for vectors, and connects to Ollama and LiteLLM-backed LLMs for embeddings and generation.

How to use

You interact with the MCP server through an MCP client. Use the store tools to add notes and documents, search them with semantic queries, and ask questions that are answered with retrieved context plus generated text. The server exposes two transport modes: stdio for local desktop integrations and HTTP streaming for web-based interfaces. When you issue questions, the server fetches relevant context chunks and then generates an answer along with sources.

Key workflows you will perform regularly: 1) add content to your personal knowledge base, 2) run semantic searches to retrieve relevant notes, and 3) ask questions that return a composed answer with supporting contexts.

How to install

Prerequisites you need before installation: ensure you have Python installed and a modern shell. You will also need Docker if you choose the container route and basic familiarity with environment variables.

# Create a Python virtual environment
python -m venv venv
source venv/bin/activate  # on Windows use `venv\Scripts\activate`

# Install dependencies
pip install -r requirements.txt

# Optional: verify Python version
python --version
# Optional: verify pip version
pip --version

# Prepare environment (examples shown; adapt to your setup)
export SQLITE_PATH=./data/documents.db
export QDRANT_URL=http://localhost:6333
export OLLAMA_URL=http://localhost:11434
export LITELLM_URL=http://localhost:4000

# Start the MCP server in stdio mode
python -m personal_rag_mcp.server

# Or start in HTTP mode
export TRANSPORT=http
export PORT=8765
python -m personal_rag_mcp.server
"} ,{

Additional configuration and deployment notes

The server uses a hybrid storage setup with SQLite for full-text documents and Qdrant for semantic search. It is designed to run in containers or on a local machine, and it can connect to Ollama for embeddings and LiteLLM as the LLM proxy. You can customize the configuration through environment variables or compose files to suit your infrastructure.

services:
  # Required: Qdrant vector database
  qdrant:
    image: qdrant/qdrant:latest
    container_name: qdrant
    ports:
      - "6333:6333"
    volumes:
      - qdrant-data:/qdrant/storage
    restart: unless-stopped

  # Required: Ollama for embeddings
  ollama:
    image: ollama/ollama:latest
    container_name: ollama
    ports:
      - "11434:11434"
    volumes:
      - ollama-data:/root/.ollama
    restart: unless-stopped

  # LiteLLM proxy for LLM access
  litellm-proxy:
    image: ghcr.io/berriai/litellm:main-latest
    container_name: litellm-proxy
    ports:
      - "4080:8000"
    volumes:
      - ./litellm_config.yaml:/app/config.yaml
    environment:
      - AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID}
      - AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY}
      - AWS_REGION=${AWS_REGION}
      - OLLAMA_API_BASE=http://ollama:11434
    entrypoint: ["litellm", "--config", "/app/config.yaml", "--port", "8000"]
    depends_on:
      - ollama
    restart: unless-stopped

  # Personal RAG MCP Server
  personal-rag-mcp:
    build: ./personal-rag-mcp
    container_name: personal-rag-mcp
    ports:
      - "8765:8765"
    environment:
      - TRANSPORT=http
      - PORT=8765
      - QDRANT_URL=http://qdrant:6333
      - OLLAMA_URL=http://ollama:11434
      - LITELLM_URL=http://litellm-proxy:8000
      - LITELLM_API_KEY=${LITELLM_API_KEY}
      - SQLITE_PATH=/app/data/documents.db
    volumes:
      - personal-rag-data:/app/data
      - ./config/personal-rag:/app/config:ro
    depends_on:
      - qdrant
      - ollama
      - litellm-proxy
    restart: unless-stopped
volumes:
  qdrant-data:
  ollama-data:
  personal-rag-data:

"} ,{

## Troubleshooting and tips

If you encounter connectivity issues, verify that Qdrant, Ollama, and LiteLLM services are reachable at their configured URLs. Confirm the MCP server is running and listening on the expected port by attempting to connect from your MCP client. Check logs for any missing environment variables or startup errors.

## Available tools

### store\_memory

Store notes, documents, or snippets in the knowledge base with namespace and metadata.

### search\_memory

Perform semantic search across stored content with optional content\_type filtering.

### ask\_with\_context

Ask questions with retrieved context to generate informed answers.