The Problem

Existing note-taking tools send your data to the cloud. I wanted a way to ask natural language questions about my Markdown notes and get AI-powered answers without leaving my machine.

The Solution

Built a local-first RAG system in Python that indexes Markdown notes and answers questions using Ollama or Hugging Face Transformers. Uses ChromaDB for vector search and SQLite for metadata tracking. Only changed files are re-indexed, making it fast and efficient.

Key features:

• Runs entirely locally, no data leaves your machine
• Supports 100+ languages via dictionary API aggregation
• Incremental sync with SQLite change detection
• Works with Ollama or built-in Transformers models
• Package with Poetry, CLI with Typer, linting with Ruff