Signallama is a bridge between Signal messenger and LLM models. It allows you to chat with AI models through Signal messages. While primarily designed for use with Ollama, it supports multiple LLM providers through LiteLLM.

Requirements

Python 3.8+
signal-cli-rest-api - Required for Signal messaging interface
Ollama or alternative LLM provider
Whisper ASR Webservice (onerahmet/openai-whisper-asr-webservice) - Required for voice message transcription (see below)

Setup

Deploy signal-cli-rest-api and set it up with your Signal account
Clone this repository
Copy example_settings.py to settings.py and configure your settings:
- Set your Signal phone number
- Configure your preferred LLM provider (Ollama by default)
- Set API details for your LLM
- (Optional) Set up Whisper ASR for voice message transcription

Whisper ASR Webservice (Voice Transcription)

To enable transcription of voice messages, you need to run the onerahmet/openai-whisper-asr-webservice Docker image:

docker run -d --name whisper-asr -p 9000:9000 onerahmet/openai-whisper-asr-webservice:latest

This will start the ASR service on port 9000 by default.
You can access the interactive API docs at http://localhost:9000/docs

In your settings.py, set the Whisper ASR API URL:

WHISPER_URL = 'http://localhost:9000'  # Whisper ASR webservice endpoint

If WHISPER_URL is set, voice messages sent to the bot will be transcribed and the transcript will be sent as a reply.

Configuration

The settings.py file contains all necessary configuration:

# Signal API Configuration
SIGNAL_URL = 'http://localhost:8080'  # signal-cli-rest-api endpoint
SIGNAL_NUMBER = '+123456789'  # Your Signal number

# LLM Configuration
LLM_PROVIDER = 'ollama'  # or 'openai', 'anthropic', etc.
LLM_MODEL = 'ollama/model-name'  # Format: provider/model
LLM_API_BASE = 'http://localhost:11434'  # Ollama API endpoint
LLM_API_KEY = None  # API key if required

Running

python signallama.py

The bot will listen for messages through Signal and respond using your configured LLM.