Personal Notes Assistant is a Retrieval-Augmented Generation (RAG) MCP server designed to index your Obsidian vault into a Milvus vector store and answer questions over your notes. It supports querying via either a local LLM deployed with Ollama or the OpenAI API, enabling you to ask complex questions and receive point-in-time accurate responses drawn from your notes. The server continuously watches your vault and keeps the knowledge base synchronized in real-time, so your queries reflect the latest updates to your notes.

Prerequisites:

• Python 3.9+
• uv (for Python package management)
• Docker and Docker Compose (for Milvus)
• Obsidian vault
• Ollama (optional, for local models)

Setup steps:

1. Clone the repository git clone https://github.com/your/repo.git cd repo

2. Start Milvus with Docker docker-compose up -d

3. Create and activate a Python virtual environment using uv uv venv .venv\Scripts\activate # Windows

   On Linux/macOS: source .venv/bin/activate

4. Install dependencies in editable mode uv pip install -e .

5. Configure environment variables Copy the sample env and edit it as needed: cp .env.sample .env

   Set OBSIDIAN_VAULT_PATH, LLM_PROVIDER, MILVUS_HOST/PORT, and API keys as appropriate

6. Run the server python main.py

Notes and tips:

• Ensure Milvus is reachable at MILVUS_HOST:MILVUS_PORT; adjust docker-compose if you run Milvus differently.
• Choose LLM_PROVIDER in the environment (.env) to match your setup: ollama for local models or openai for API access.
• If you switch to CUDA-enabled PyTorch, follow the PyTorch CUDA installation steps and reinstall Torch accordingly.
• For local models with Ollama, ensure Ollama is installed and the model specified in LLM_MODEL is downloaded.
• The server watches the Obsidian vault in real-time; changes will be reflected in search results after indexing completes.
• If you encounter authentication issues with OpenAI, verify the API key and ensure it has the required permissions for the selected model.