pgvector-python

Why we included this project

Python teams that already run PostgreSQL can add vector search without standing up a separate database service, because this package layers pgvector's vector type and distance operators onto the drivers and ORMs they probably use already: Django, SQLAlchemy, SQLModel, Psycopg 2 and 3, asyncpg, pg8000, and Peewee. Nearest-neighbor queries fit into an existing schema with a few lines, so there is no need to rework the data layer. The included examples reach beyond basic similarity search into hybrid retrieval with reciprocal rank fusion, sparse and late-interaction search, and image or document lookup, which makes it handy for teams prototyping RAG on infrastructure they already maintain. Since it is a client binding rather than a standalone engine, you get vector capabilities inside your current Postgres deployment instead of committing to a new database.