yahorbarkouski/semantic-mail

Generates embeddings locally via Ollama or OpenAI and stores them in ChromaDB for vector similarity search, with support for swappable embedding models and incremental syncing. The `ask` command uses retrieved emails as context for LLM-powered question answering, with configurable search depth and token limits across multiple LLM providers. Designed for privacy-first local operation with optional cloud embedding providers, requiring only Gmail OAuth credentials and minimal disk overhead (~1GB per 10,000 emails).

No commits in the last 6 months.