Verbi is a modular voice assistant designed for research and development in voice technology. It integrates multiple state-of-the-art APIs and local models to handle the full voice interaction pipeline: transcription via speech-to-text, response generation using large language models, and speech synthesis via text-to-speech. The modular architecture allows developers to swap different models and providers, facilitating easy comparison and experimentation across transcription, response generation, and TTS components.