korvus

Why we included this project

If your stack already runs on Postgres, Korvus keeps RAG inside that same database. A single SQL query handles embedding generation, vector search, reranking, and summarization, so you don't need to wire together a separate vector store, embedding service, and reranker. It ships official SDKs for Python, JavaScript, Rust, and C, which makes it easy to drop into most existing codebases. Because the pipeline lives in Postgres, you avoid adding a whole extra infrastructure tier for vector search and keep chunking and retrieval close to your data. For teams that want a self-hosted, database-centric approach to retrieval-augmented generation, this is a pragmatic starting point.