GitHub – Modular, real-time AI voice assistant platform

Summary: Stimm is an open-source platform for orchestrating ultra-low latency AI voice pipelines using Python (FastAPI) and Next.js. It enables flexible creation and management of voice agents by integrating various LLM, TTS, and STT providers, supporting real-time conversations over WebRTC.

What it does

Stimm provides a modular infrastructure that handles voice agent orchestration, including interruptions, voice activity detection, and WebRTC streaming. It supports swapping providers like OpenAI, Mistral, Deepgram, and ElevenLabs, and is deployable via Docker.

Who it's for

Developers building AI voice assistants who require low-latency, customizable pipelines and prefer an open-source, self-hosted solution.

Why it matters

It addresses the latency issues common in proprietary APIs by offering a fast, open-source orchestration layer for real-time voice interactions.