OBP-GPT (Opey)

A helpful chatbot trained on Open Bank Project API documentation

Running with Docker (reccommended)

1. Set up Environment

• Install docker compose if you haven't already

• Copy .env-example to .env

  • Copy over your OpenAI API Key

• Make sure you have a ./certs/public_key.pem file. Make the directory and copy over the key from API Explorer II if not.

Change the .env file to have

ENDPOINT_METADATA_PATH=./vector-database/endpoint_metadata.json
GLOSSAY_METADATA_PATH=./vector-database/glossary_metadata.json

ENDPOINT_VECTOR_DATABASE_PATH=./vector-database/endpoint_index.faiss
GLOSSARY_VECTOR_DATABASE_PATH=./vector-database/glossary_index.faiss

2. Build App

sudo docker compose build

3. Run

sudo docker compose up

Note: If you are running this on your local docker engine and you already have an instance of redis running, you may need to change the REDIS_PORT in the env file to avoid clashing.

4. Chat

The chat endpoint should now be running at http://127.0.0.1:5000/chat

The best way to interact with Opey locally is by running a version of Api Explorer II Locally and using the chat widget.

Else you can chat with the bot using curl (or whatever http client you like):

curl -XPOST -H "Content-type: application/json" -d '{
  "session_id": "123456789", "obp_api_host": "https://test.openbankproject.com", "message": "Which endpoint would I use to create a new customer at a bank?"
}' 'http://127.0.0.1:5000/chat'

Running Locally

1. Install dependencies

pip install -r requirements.txt

2. Set up Environment

• Copy .env-example to .env
  • Copy over your OpenAI API Key
  • For running locally, set REDIS_HOST=localhost
  • You will also need redis set up and running locally, find instructions for this here

3. Create vector index

We need to register the OBP API documentation in a vector index, run:

python create_vector_index.py

4. Run

For development:

python app.py

For production we use gunicorn:

gunicorn --bind 0.0.0.0:5000 aop:app

5. Chat

Same as for docker (see above)