Vector DB Benchmark

Supported databases/extensions:

• pgvecto.rs
• pgvector
• qdrant

Supported datasets:

• random generated
• GIST 960
• GLOVE
• Deep Image
• LAION

For more information, check the source.py.

Installation

pip install vector_bench

Run

Server

Run the docker compose file under server folder.

docker compose -f docker/compose.${DB_NAME}.yaml up -d

Client

# help
vector_bench --help
# only insert the data
vector_bench --insert --url postgresql://postgres:[email protected]:5432/postgres -s gist_960_l2
# only query the data (make sure the data is already inserted)
vector_bench --query --url postgresql://postgres:password@localhost:5432/postgres -s gist_960_l2
# insert and query the data
vector_bench --insert --query --url postgresql://postgres:password@localhost:5432/postgres -s gist_960_l2

How to contribute

# install all the necessary dependencies:
make dev
# format code
make format
# lint
make lint

Add more datasets

• Add new DatasetConfig to vector_bench/dataset/source.py

Add more clients

• Inherit and implement the BaseClient class in vector_bench/client/base.py