Barabasz Signal Bot

What is Barabasz?

This is a bot that uses the SIGNAL REST API (specifically, websocket JSON-RPC) to receive and analyze messages from Signal messenger and send queries to OpenAI language models. Once configured, the bot redirects messages to the selected language model and returns a response from the network. At this point, OpenAI does not provide a GPT Chat API, so the bot does not understand and store the context of the utterance and the entire query must be contained in a single prompt.

Features

• Support for direct messages
• Support for group messages, (you need to add a Signal user to a group and then mention him in the message)
• Open to modification- you can easily add other logic to process messages
• Multiple OpenAI language models to choose

Configuration

Configure your own Signal Backend:

Important note: Configure your API in JSON-RPC mode

More info about SIGNAL REST CLI

Clone the repository

git clone git@github.com:jezonek/Barabasz.git

Configure your own env variables

cd barabasz

cp .env_example .env

Complete the .env file with the data. Add your OpenAI API key, your Signal account number, and the endpoints of your Signal Rest API

Install pipenv dependencies

pipenv install

Start the bot

pipenv run python src/barabasz.py

Bot will connect to your configured secure websocket (wss) endpoint SIGNAL_WEBSOCKET_RX_ENDPOINT and start to listen for incoming messages.

Sending a response is done by executing a post request to the URI specified in SIGNAL_HTTPS_ENDPOINT

Deployment

I deploy my bot as a service on linux, using systemd. The sample .service file is in the deploy folder. In place of USERNAME, insert the name of the user who is running the service. I upload the files to the server to the user's root directory via scp.