PostgreSQL-V

Open-Source Vector Similarity Search for PostgreSQL

Introduction

PostgreSQL-V is a new integrated vector database system developed by Purdue Database Group that enables fast vector search in PostgreSQL. It achieves performance on par with specialized vector databases.

Unlike prior systems (e.g., pgvector) that inherits legacy overhead by reusing PostgreSQL’s page-oriented structure, PostgreSQL-V adopts a novel architectural design that decouples vector indexes from PostgreSQL’s core engine. This decoupling offers many benefits, such as directly leveraging native vector index libraries for high performance. However, it also introduces the challenge of index inconsistency, which we address with a lightweight consistency mechanism.

PostgreSQL-V is fully compatible with pgvector (by implementing the same APIs in IndexAmRoutine), making it a viable drop-in replacement.

Reference:
Jiayi Liu, Yunan Zhang, Chenzhe Jin, Aditya Gupta, Shige Liu, and Jianguo Wang. Fast Vector Search in PostgreSQL: A Decoupled Approach. Conference on Innovative Data Systems Research (CIDR), 2026.

Installation

1. Install PostgreSQL

Follow the official PostgreSQL installation guide for your operating system.

2. Build and Install the Extension

cd postgresqlv
vim ./install_pgvector.sh

Locate the line starting with:

PG_CONFIG=

and update it to point to your local pg_config path. Then run:

./install_pgvector.sh

Getting Started

PostgreSQL-V is fully compatible with pgvector. You can follow pgvector's user tutorial to get started with basic usage and examples.

Architecture Overview

The following figure illustrates the high-level architecture of PostgreSQL-V.

(Details of the design will be provided in our upcoming CIDR paper; the link will be added here once available.)

Experiment Overview

SIFT10M DEEP10M

Note

⚠️The main branch is currently under active development.

For an executable prototype, please switch to the cidr branch, which contains the version used in our CIDR submission to validate the proposed design.

Citation

If you find our system useful for your research, please cite the following paper:

@inproceedings{PostgreSQLV-CIDR26,
  title={Fast Vector Search in PostgreSQL: A Decoupled Approach},
  author={Jiayi Liu and Yunan Zhang and Chenzhe Jin and Aditya Gupta and Shige Liu and and Jianguo Wang},
  booktitle={Conference on Innovative Data Systems Research (CIDR)},
  year={2026}
}

Contact

If you have any questions, you are more than welcome to contact Dr. Jianguo Wang