LiveKit Inference

Status: implemented as adapter support.

Use the livekit provider when you want rtp-agent to reach LiveKit Inference through the repository's adapter package.

The adapter supports LLM, STT, and TTS. It is not a separate model catalog page; model names should come from source examples, tests, or LiveKit Inference configuration you control.

App configuration

The basic agent configures LiveKit Inference by setting:

• LLMProvider = "livekit"
• STTProvider = "livekit"
• TTSProvider = "livekit"

Credentials are resolved from LIVEKIT_API_KEY and LIVEKIT_API_SECRET for the inference adapter paths.

Direct constructors

Use direct constructors when you are outside app.Init:

• livekit.NewLiveKitInferenceLLM
• livekit.NewSTT
• livekit.NewTTS

Evidence:

• adapter/livekit/llm.go
• adapter/livekit/stt.go
• adapter/livekit/tts.go
• adapter/livekit/models.go
• app/app.go
• examples/voice_agents/basic_agent/basicagent/basic_agent.go