pgvecto-rs

What it is and what it does

pgvecto-rs is a Python client library for PGVecto.rs, a PostgreSQL extension that adds native vector data types and approximate nearest-neighbor search capabilities. It bridges Python applications to PostgreSQL by providing type mappings and helper functions for four vector types: standard vectors, sparse vectors, half-precision vectors, and binary vectors. The library integrates with three major Python database frameworks—SQLAlchemy, Psycopg3, and Django—allowing developers to store, index, and query embeddings through their ORM or query interface.

The package handles vector insertion, approximate indexing via HNSW and IVF algorithms, and distance-based queries. It requires a running PGVecto.rs PostgreSQL instance and supports Python 3.8 through 3.12. Because the package is dormant with no recent updates, users should verify compatibility with their target PostgreSQL and PGVecto.rs versions before adopting it in production.

Use it for

• Store and query embeddings from machine learning models in PostgreSQL using ORM without writing raw SQL.
• Perform approximate nearest-neighbor searches on vector embeddings with HNSW or IVF indexing for semantic search or recommendation systems.
• Bulk-load large sets of vectors into PostgreSQL via Psycopg3's COPY interface for efficient data ingestion.
• Define vector fields in Django models and run similarity queries through the ORM without custom database functions.
• Support multiple vector representations (full-precision, sparse, half-precision, binary) in a single PostgreSQL database for different use cases.

Worth the install?

AI-flagged interpretation of the facts on this page. Verify before relying on it.

Install

pgvecto-rs on PyPI

Before you install

Low install friction with only two runtime dependencies (numpy and toml). However, the package is dormant—last release was 675 days ago—so maintenance updates and bug fixes are unlikely.

Requires a running PGVecto.rs PostgreSQL instance (e.g., via Docker) and the vectors extension enabled in the database.

License in practice

Licensed under Apache-2.0 (permissive), allowing commercial and private use with minimal restrictions.

Quickstart

pip install pgvecto_rs

from pgvecto_rs.sqlalchemy import Vector

# Define a model with a vector column
class Item:
    embedding = Vector(3)

# Insert vectors and query by distance
# session.execute(insert(Item).values(embedding=[1, 2, 3]))
# session.scalars(select(Item).filter(Item.embedding.l2_distance([3, 1, 2]) < 5))

Package facts

Evidence: pgvecto_rs-0.2.2-py3-none-any.whl

Tags

More Database packages

See also pgvector · nano-vectordb · pyobvector · vecs · sqlite-vec · pg0-embedded · llama-index-vector-stores-postgres · pgserver · psycopg · voyager