postgresql multimodal retrieval
1.0.0
Multimodal retrieval using Vision Language Model with PostgreSQL database - A full stack implementation.
transformers implementation)Features:
conda create -n postgresql-multimodal python=3.10
conda activate postgresql-multimodalconda install -c conda-forge postgresql
psql --versionconda install -c conda-forge pgvectorinitdb -D mylocal_db
pg_ctl -D mylocal_db -l logfile startcreateuser retrieval_user
createdb retrieval_db -O retrieval_userpip install -r requirements.txtInstall the pgmmr package
pip install -e .
Run
python compute_embeddings.py
This will compute embeddings for all the images in the database.
Run hybrid search (vector search and keyword search). The results are ranked using RRF.
python query.py "a cat and flower" --num_results 12
Run
python gradio_app.py
flet run flet_app.py
