postgres-ts-vectors/README.md

TimescaleDB + VectorChord + pgvector

Docker image combining PostgreSQL with three powerful extensions:

• TimescaleDB - Time-series database extension
• VectorChord - High-performance vector similarity search
• pgvector - Open-source vector similarity search

Supported PostgreSQL Versions

Tag	PostgreSQL	TimescaleDB	VectorChord	pgvector
pg15	15	2.28.3	1.1.1	0.8.1
pg16	16	2.28.3	1.1.1	0.8.1
pg17	17	2.28.3	1.1.1	0.8.1
pg18, latest	18	2.28.3	1.1.1	0.8.1

Quick Start

docker run -d \
  --name my-postgres \
  -e POSTGRES_PASSWORD=mysecretpassword \
  -v postgres_data:/var/lib/postgresql \
  -p 5432:5432 \
  gitea.killinger.fr/maxime.killinger/postgres-ts-vectors:latest

All extensions are automatically created on first startup.

Data Persistence

Important

Starting with PostgreSQL 18, mount volumes at /var/lib/postgresql (not /var/lib/postgresql/data). Data will be stored in a versioned subdirectory (e.g., /var/lib/postgresql/18/main), which facilitates future upgrades with pg_upgrade.

Environment Variables

Variable	Description	Default
POSTGRES_PASSWORD	PostgreSQL password (required)	-
POSTGRES_USER	PostgreSQL user	postgres
POSTGRES_DB	Default database	postgres

Extensions Usage

TimescaleDB

-- Create a hypertable
CREATE TABLE conditions (
  time TIMESTAMPTZ NOT NULL,
  device_id INTEGER,
  temperature DOUBLE PRECISION
);

SELECT create_hypertable('conditions', 'time');

VectorChord / pgvector

-- Create a table with vector embeddings
CREATE TABLE documents (
  id SERIAL PRIMARY KEY,
  content TEXT,
  embedding vector(1536)
);

-- Create an index for fast similarity search
CREATE INDEX ON documents USING vchordrq (embedding vector_l2_ops);

-- Query similar documents
SELECT * FROM documents
ORDER BY embedding <-> '[0.1, 0.2, ...]'
LIMIT 10;

Building Locally

# Build for PG18
docker build --build-arg PG_VERSION=18 -t my-image:pg18 .

# Build for PG16
docker build --build-arg PG_VERSION=16 -t my-image:pg16 .

License

This project is licensed under the Apache 2.0 License.

• TimescaleDB: Timescale License
• VectorChord: Apache 2.0
• pgvector: PostgreSQL License