Pipecat
used in
0
stacks
Open-source Python framework from Daily for building realtime voice and multimodal conversational agents.
Orchestrates audio, video, AI services, and conversation logic as composable pipeline components. Offers pluggable integrations across many STT, TTS, and LLM providers, plus transports including Daily WebRTC, LiveKit, and WebSockets. Supports multi-agent patterns such as handoff and parallel fan-out, and agents can run locally or be deployed across machines. Licensed under BSD-2-Clause, with a managed hosting option from Daily.
Strengths
- Vendor-neutral pipeline with pluggable STT, TTS, and LLM providers
- Composable frame/pipeline architecture for custom voice flows
- Multiple transport options including Daily WebRTC, LiveKit, and WebSockets
- Permissive BSD-2-Clause license
Tradeoffs
- Pipeline and frame abstractions take time to learn before simple bots feel simple
- Latency and reliability depend heavily on the chosen providers and transport, which you must tune yourself