patricktrainer/duckdb-embedding-search

Leverages OpenAI embeddings stored in DuckDB's columnar format for fast vector similarity comparisons, enabling semantic search across large datasets like Hacker News comments. The modular architecture separates concerns across connection management, embedding generation, and vector operations, with lazy embedding creation that calls the OpenAI API on-demand for new queries. Targets Python developers building semantic search applications who need lightweight, embedded similarity lookup without managing separate vector databases.

146 stars. No commits in the last 6 months.