Pipecat Voice Pipeline
Pipecat is a Python framework that orchestrates STT, TTS, and LLM services. It supports Ollama, Whisper, and Piper as first-class services.
What Pipecat Gives You
Section titled “What Pipecat Gives You”- A pipeline framework: audio capture → STT → LLM → TTS → playback
- Service adapters for Ollama, Whisper, and Piper — minimal custom glue needed
- Quickstart examples to test a full voice loop in minutes
Mapping to the Homelab
Section titled “Mapping to the Homelab”- Ollama — Keep running bare-metal. Point Pipecat to
http://localhost:11434. - Whisper (Docker) — Expose STT endpoint to Pipecat via host networking or published port.
- Piper (Docker) — Wire container so Pipecat can post text and receive audio stream.
- Pipecat server — Install on same host, configure service URIs, run quickstart to validate.
Minimal Wiring Checklist
Section titled “Minimal Wiring Checklist”pip install "pipecat-ai[ollama,whisper,piper]"- Confirm Ollama is reachable:
ollama ps - Start Whisper and Piper containers; verify health endpoints
- Configure
services: { llm: ollama://..., stt: whisper://..., tts: piper://... }and run quickstart
Practical Notes
Section titled “Practical Notes”- Networking — Run Pipecat and Ollama on the same host; use host networking for containers.
- GPU — Verify NVIDIA drivers and Ollama GPU config for Ollama and Faster-Whisper.
- Audio formats — Ensure Whisper container accepts Pipecat’s sample rate/format.
- Security — Bind services to localhost; firewall external access.