InstiGPT

The answer to all of insti's questions.

Developing

The application has 4 parts: backend, frontend a vector database and a relational database.

Backend

1. Install the dependencies using pipenv install.
2. Run the app using pipenv run python main.py.

Frontend

1. Install the dependencies using pnpm i.
2. Run the app using pnpm dev.

Vector DB

This app uses chromadb as its vector store. To run an instance of chromadb in docker, run: docker run -p 8000:8000 --env IS_PERSIST=true --env ALLOW_RESET=true --name chromadb chromadb/chroma.

Also make sure you set the required environment variables in the .env file namely VECTOR_DB_HOST to localhost and VECTOR_DB_PORT to 8000

To generate embeddings, run pipenv run python load_data_in_db.py in scripts folder. Do the commented changes accordingly to use resobin data or ugrulebook data in load_data_in_db.py and config.py.

NOTE: You may choose to recreate all the embeddings from scratch or reuse the ones provided here. To reuse the ones provided in the link, first download that folder to a folder named embeddings and then run docker exec -it chromadb rm -rf /chroma/chroma; docker cp ./embeddings chromadb:/chroma/chroma

Relational DB

This app uses postgres as its relational database. To run an instance of postgres in docker, run: docker run -p 5432:5432 --env POSTGRES_PASSWORD=<password> --name postgres postgres.

Running in Production

This app can easily be run in production through docker by running docker compose up -d