VoiceParticipant (participant.py)

Voice participant builders for live calls. A VoiceParticipant produces a LiveKit AgentSession for one side of a call. Cascade (STT -> LLM -> TTS) is the only implementation today. The interface is shaped to admit a future RealtimeParticipant that wraps a native speech-to-speech m

# Source: https://github.com/voicetestdev/voicetest | License: Apache-2.0 | Copyright You must retain, in the Source form of any Derivative Works
# Provided via justclaw.it.com; keep this notice with the code.
"""Voice participant builders for live calls.

A VoiceParticipant produces a LiveKit AgentSession for one side of a call.
Cascade (STT -> LLM -> TTS) is the only implementation today.

The interface is shaped to admit a future RealtimeParticipant that wraps a
native speech-to-speech model in the llm slot: it does not assume a separable
LLM, and it treats intended text as an optional capability (a cascade exposes
the LLM's intended output; an S2S model has no separable text brain, so it
exposes only what the observer hears). A RealtimeParticipant is not built yet
and is expected to re

... (truncated -- full source via MCP)