LegalGPT

ChatGPT of legal domain

Running Instructions

Backend

The backend instance uses PostgreSQL for vector DB. Thus, you need to add pg_vector extension beforehand.

postgres=# CREATE EXTENSION vector;

After set up the DB, you need to add the .env file.

cd legalgpt/backend

# copy the .env.template
cp .env.template .env

Open the created file, and fill in the blanks (OpenAI API key, and postgresql url):

# model could be gpt-4, gpt-3.5-turbo, gpt-4-vision-preview, etc.
MODEL=gpt-4
OPENAI_API_KEY=
# For generating a connection URI, see https://docs.timescale.com/use-timescale/latest/services/create-a-service
PG_CONNECTION_STRING=

For dependency management, LegalGPT uses poetry:

cd legalgpt/backend

# install dependencies into poetry's venv
poetry install

Before start the inference service, we need to generate the DB first:

# run the ingestion service
poetry run python3 generate.py

Now, run the inference service:

# run app with poetry's venv
poetry run python3 main.py

We currently use python3.10 and python3.11, thus, please make sure that you use the correct python version.

Frontend

Follow the instructions below:

cd legalgpt/frontend

# make the .env file
cp .env.template .env

# install the dependencies
npm install

# to run demo, use npm run dev
npm run dev