★★★★☆ 3.9 CE

pgvector Framework Integrations & SDKs 2026

Q: How do I use pgvector with LangChain?

Install langchain-postgres and psycopg with the binary extra, then import PGVector from langchain_postgres. Connect using a psycopg3 connection string, with the postgresql+psycopg prefix, set collection_name and embeddings, then call add_documents to index and similarity_search to retrieve. Note that langchain-postgres uses psycopg3, not psycopg2, and the old langchain-community PGVector class is deprecated.

Q: How do I use pgvector with LlamaIndex?

Install llama-index-vector-stores-postgres, at v0.8.1, which requires Python 3.10 or newer, and initialize PGVectorStore through PGVectorStore.from_params, passing host, port, database, table_name and embed_dim. Supported query modes include vector similarity, hybrid dense-plus-sparse, BM25, full-text and MMR for diverse results.

Q: Which Python drivers does pgvector support?

The pgvector Python package v0.4.2 supports psycopg3, psycopg2, asyncpg, pg8000, Django, SQLAlchemy, SQLModel and Peewee. langchain-postgres requires psycopg3 specifically. asyncpg is available for high-throughput async workloads.

Q: Which Node.js drivers does pgvector support?

The pgvector npm package v0.3.0 supports 13 drivers: node-postgres, Prisma, TypeORM, Drizzle ORM, Knex.js, Objection.js, Kysely, Sequelize, pg-promise, Postgres.js, Slonik, MikroORM and Bun SQL. It also supports Deno and WebAssembly environments.

Q: Is there a pgvector client for Go or Rust?

No dedicated pgvector client exists for Go or Rust. Teams using those languages access pgvector through the standard Postgres wire protocol and the SQL distance operators, directly through existing Postgres drivers such as pgx for Go or tokio-postgres for Rust.

pgvector integrates natively with LangChain and LlamaIndex, plus Django, SQLAlchemy and 13 Node.js drivers. Teams on Postgres skip a separate vector service.

Vector DatabasesVerified today

Full review Pricing Official site ↗

pgvector Framework Integrations verdict

Verified today·7 sources checked

pgvector integrates natively with LangChain, through langchain-postgres v0.0.17, and LlamaIndex.

Through llama-index-vector-stores-postgres v0.8.1, for Python RAG pipelines. Django, SQLAlchemy and psycopg3 or asyncpg cover direct access.

How it fits your stack

If your stack is Python on LangChain or LlamaIndex, pgvector setup takes a pip install and a connection string. For Node.js, the pgvector npm package covers all major ORMs. Teams already running Postgres avoid adding a separate vector database service entirely.

Honest limits

langchain-postgres replaces the deprecated langchain-community PGVector class, so existing code using langchain-community.vectorstores.PGVector must migrate to langchain-postgres.PGVector.
langchain-postgres requires psycopg3, not psycopg2, and the connection string prefix is postgresql+psycopg, not postgresql+psycopg2.
llama-index-vector-stores-postgres requires Python 3.10 or newer, while the base pgvector Python package only requires 3.9 or newer.
There is no official Go, Rust, Java or PHP pgvector client, so those ecosystems access vectors through the standard Postgres wire protocol and SQL operators.

LangChain package: langchain-postgres v0.0.17
LlamaIndex package: llama-index-vector-stores-postgres v0.8.1
Node.js drivers supported: 13
Python ORMs supported: Django, SQLAlchemy, Peewee, SQLModel
Official Go / Rust client: No

View sources

pgvector framework and orchestration integrations

Integration	Type	Capabilities	Setup
LangChain	Native	RAG orchestration · Hybrid search · Metadata filtering	Low
LlamaIndex	Native	Vector index · Hybrid search · MMR retrieval · Metadata filtering	Low
Django (ORM)	Native	Vector field · ORM query · Nearest neighbor	Low
SQLAlchemy	Native	Vector type · ORM · Distance methods	Low
psycopg 3 / psycopg 2	Driver	Raw SQL · Vector serialization · Async support	Low
asyncpg	Driver	Async queries · High-throughput inserts	Low
node-postgres (pg)	Driver	Node.js queries · Vector insert/query	Low
Prisma / TypeORM / Drizzle	ORM	TypeScript ORM · Vector column · Nearest neighbor	Medium

LangChain + pgvector quickstart

pip install langchain-postgres langchain-openai psycopg[binary]python

from langchain_postgres import PGVector
from langchain_openai import OpenAIEmbeddings

# Connect to an existing Postgres database with pgvector installed
connection = "postgresql+psycopg://user:pass@localhost:5432/mydb"

vector_store = PGVector(
    embeddings=OpenAIEmbeddings(model="text-embedding-3-large"),
    collection_name="my_docs",
    connection=connection,
    use_jsonb=True,
)

# Index documents
vector_store.add_documents(docs)

# Retrieve with metadata filter
hits = vector_store.similarity_search("how does billing work?", k=4, filter={"source": "docs"})

# Or get results with similarity scores
scored = vector_store.similarity_search_with_score("query text", k=5)

Python client packages and versions

Package	Version	Released	Status
pgvector (PyPI)	0.4.2	Dec 5, 2025	Official
langchain-postgres	0.0.17	Feb 17, 2026	Official
llama-index-vector-stores-postgres	0.8.1	Mar 13, 2026	Official
Python requirement	>=3.9 (>=3.10 for LI)	—	Requirement

pgvector LangChain integration detail

Installed via pip install langchain-postgres; requires psycopg3 (not psycopg2)
Primary class is PGVector (imported from langchain_postgres); connect via a psycopg3 connection string
Supports add_documents(), delete(), similarity_search(), similarity_search_with_score(), and as_retriever()
Metadata filter operators: $eq, $ne, $lt, $lte, $gt, $gte, $in, $nin, $between, $like, $ilike, $and, $or
Connection string format: postgresql+psycopg://user:pass@host/db (psycopg3 driver, not psycopg2)
Supports hybrid search combining vector similarity and keyword-based search via reciprocal rank fusion

pgvector LlamaIndex integration detail

Installed via pip install llama-index-vector-stores-postgres (v0.8.1, Mar 2026); requires Python >=3.10
Primary class is PGVectorStore, initialized via PGVectorStore.from_params() with host/port/db/embed_dim
Supports multiple query modes: vector similarity, hybrid (dense + sparse), BM25 text search, and MMR for diverse results
HNSW parameters (m, ef_construction, ef_search, dist_method) are configurable at PGVectorStore creation via hnsw_kwargs
Metadata filter operators include >=, <=, in, nin, contains, and nested AND/OR logic
Custom SQL query building is supported via customize_query_fn for table joins and advanced retrieval

pgvector Framework Integrations FAQ

How do I use pgvector with LangChain?

Install langchain-postgres and psycopg with the binary extra, then import PGVector from langchain_postgres. Connect using a psycopg3 connection string, with the postgresql+psycopg prefix, set collection_name and embeddings, then call add_documents to index and similarity_search to retrieve. Note that langchain-postgres uses psycopg3, not psycopg2, and the old langchain-community PGVector class is deprecated.

How do I use pgvector with LlamaIndex?

Install llama-index-vector-stores-postgres, at v0.8.1, which requires Python 3.10 or newer, and initialize PGVectorStore through PGVectorStore.from_params, passing host, port, database, table_name and embed_dim. Supported query modes include vector similarity, hybrid dense-plus-sparse, BM25, full-text and MMR for diverse results.

Which Python drivers does pgvector support?

The pgvector Python package v0.4.2 supports psycopg3, psycopg2, asyncpg, pg8000, Django, SQLAlchemy, SQLModel and Peewee. langchain-postgres requires psycopg3 specifically. asyncpg is available for high-throughput async workloads.

Which Node.js drivers does pgvector support?

The pgvector npm package v0.3.0 supports 13 drivers: node-postgres, Prisma, TypeORM, Drizzle ORM, Knex.js, Objection.js, Kysely, Sequelize, pg-promise, Postgres.js, Slonik, MikroORM and Bun SQL. It also supports Deno and WebAssembly environments.

Is there a pgvector client for Go or Rust?

No dedicated pgvector client exists for Go or Rust. Teams using those languages access pgvector through the standard Postgres wire protocol and the SQL distance operators, directly through existing Postgres drivers such as pgx for Go or tokio-postgres for Rust.

Sources & verification

Verified by ComparEdgeMethod: Vendor docs, official pages, and selected independent sources

Source	What was checked	Last checked
GitHub Pgvector Pgvector	Official product page	July 10, 2026
GitHub Pgvector Pgvector Node	Pgvector Pgvector Node	July 10, 2026
GitHub Pgvector Pgvector Python	Pgvector Pgvector Python	July 10, 2026
Langchain Developer docs	Independent reference	July 10, 2026
Llamaindex Vector Stores Postgres	Independent reference	July 10, 2026
Pypi Project Langchain Postgres	Project Langchain Postgres	July 10, 2026
Pypi Project Llama Index Vector Stores Postgres	Project Llama Index Vector Stores Postgres	July 10, 2026

Every fact on this pgvector page is tied to a named source and a verification date. Freshness-sensitive figures trace to the sources above; verify against the vendor before relying on them.