pipecat

Why we included this project

Pipecat is a Python framework for teams that want real-time voice and multimodal agents without hand-wiring the infrastructure. Instead of gluing speech recognition, an LLM call, text-to-speech, and audio or video transport together from scratch, you assemble them into pipelines that run over WebRTC or WebSockets with round trips between 500 and 800ms, the latency range that keeps streaming conversation feeling natural. That fits voice-first products like phone support bots, in-browser assistants, and AI companions. The framework also steps up from a single agent to a multi-agent system where specialist workers hand off control or run in parallel over a shared bus, either in-process or distributed across machines. Developers who already know which AI services they want to plug in get the orchestration and transport handled for them, leaving room to focus on conversation logic.