Classes

Concrete class implementations — providers, orchestrators, and core SDK components.

• AgentStateMachine — Orchestrator state machine that derives a high-level AgentState from three underlying sub-machines: capture, playback, and processing.
• AnthropicLLM — Anthropic LLM provider for Claude models.
• AssemblyAISTT — AssemblyAI real-time STT provider using a raw WebSocket connection.
• AudioBufferQueue — Bounded FIFO queue that buffers audio chunks between pipeline stages.
• AudioCapture — Manages microphone audio capture using the Web Audio API.
• AudioHeaderCache — Caches the audio container header from a stream for re-injection on reconnect.
• AudioPlayer — Manages audio playback using the Web Audio API with support for both complete and streaming playback modes.
• AudioRouter — Manages audio buffer queues and header cache wiring for the pipeline.
• AzureSTT — Microsoft Azure Speech real-time STT provider using a raw WebSocket.
• AzureTTS — Microsoft Azure Speech TTS provider using native fetch.
• BrowserAudioOutput — Browser audio output provider that plays audio through the Web Audio API.
• BufferInput — Server-side audio input provider that accepts pushed audio buffers.
• CartesiaTTS — Cartesia TTS provider for low-latency real-time streaming text-to-speech via WebSocket.
• CompositeVoice— The primary class of the CompositeVoice SDK, orchestrating a complete 5-role audio pipeline from input capture through speech recognition, language model…
• ConversationManager — Manages multi-turn conversation history for the LLM pipeline.
• DeepgramAgent — Deepgram Agent API provider — covers STT + LLM + TTS in a single WebSocket.
• DeepgramFlux — Deepgram Flux (V2) real-time STT provider using native WebSocket.
• DeepgramSTT — Deepgram real-time STT provider using native WebSocket (no SDK required).
• DeepgramTTS — Deepgram TTS provider for real-time streaming text-to-speech via native WebSocket.
• DiscordVoice — Duplex Discord voice-channel provider (‘input’ + ‘output’).
• EagerLLMController — Manages the eager/speculative LLM pipeline.
• ElevenLabsAgent — ElevenLabs Conversational AI agent — covers STT + LLM + TTS in a single WebSocket.
• ElevenLabsSTT — ElevenLabs STT provider for real-time streaming speech-to-text via WebSocket.
• ElevenLabsTTS — ElevenLabs TTS provider for real-time streaming text-to-speech via WebSocket.
• EventEmitter — A type-safe event emitter with support for wildcard listeners and both synchronous and asynchronous event dispatch.
• FallbackSTT — A meta STT provider that chains multiple live STT providers with automatic failover.
• FishAudioTTS — Fish Audio TTS provider using native fetch with msgpack-encoded requests.
• GeminiLiveAgent — Gemini Live speech-to-speech agent — covers STT + LLM + TTS in a single WebSocket.
• GeminiLLM — Google Gemini LLM provider.
• GladiaSTT — Gladia real-time STT provider using the v2 live API.
• GoogleMeetInput — Browser audio input provider that joins a Google Meet conference via the Meet Media API (Developer Preview) and emits the meeting’s mixed audio.
• GoogleSTT — Google Cloud STT provider using native fetch for batch (per-utterance) speech-to-text.
• GoogleTTS — Google Cloud TTS provider using native fetch for text-to-speech synthesis.
• GroqLLM — Groq LLM provider for ultra-fast inference.
• GuardrailPipeline — Runs a Guardrail chain over LLM text on its way to TTS.
• GuardrailStream — Stateful guardrail filter for one streamed LLM response.
• LMNTTTS — LMNT TTS provider using native fetch for text-to-speech synthesis.
• Logger — Structured logger with context-aware formatting and configurable levels.
• MicrophoneInput — Browser audio input provider that captures audio from the microphone.
• MiniMaxTTS — MiniMax TTS provider using native fetch for text-to-speech synthesis.
• MistralLLM — Mistral LLM provider.
• MurfTTS — Murf AI TTS provider using native fetch for text-to-speech synthesis.
• NativeSTT — Native browser STT provider backed by the Web Speech API (SpeechRecognition).
• NativeTTS — Native browser TTS provider using the Web Speech API (SpeechSynthesis).
• NullOutput — No-op audio output provider that discards all audio.
• OpenAICompatibleLLM — Base LLM provider for any service that speaks the OpenAI chat completions format.
• OpenAILLM — OpenAI LLM provider for GPT models.
• OpenAIRealtimeAgent — OpenAI Realtime speech-to-speech agent — covers STT + LLM + TTS in a single WebSocket.
• OpenAIRealtimeSTT — OpenAI real-time STT provider using the Realtime API transcription intent over a raw WebSocket connection.
• OpenAITTS — OpenAI TTS provider using native fetch for text-to-speech synthesis.
• PollyTTS — Amazon Polly TTS provider using native fetch with SigV4-signed requests.
• ProviderEventAdapter — Optional adapter that bridges provider callbacks to a typed EventEmitter.
• RecoveryOrchestrator — Pipeline-level error recovery orchestrator that coordinates recovery across providers with exponential backoff.
• RevAISTT — Rev AI streaming STT provider using a raw WebSocket connection.
• RimeTTS — Rime TTS provider using native fetch for text-to-speech synthesis.
• SileroVAD — Local Silero voice-activity-detection engine.
• SmallestTTS — Smallest.ai Waves TTS provider using native fetch for text-to-speech synthesis.
• SonioxSTT — Soniox real-time STT provider using a raw WebSocket connection.
• SpeechifyTTS — Speechify TTS provider using native fetch for text-to-speech synthesis.
• SpeechmaticsSTT — Speechmatics real-time STT provider using a raw WebSocket connection.
• TeamsCall — Duplex provider that joins a Microsoft Teams meeting via Azure Communication Services and bridges its audio to the pipeline.
• TranscribeSTT — Amazon Transcribe real-time STT provider using the streaming WebSocket API.
• TTSBackpressure — A counting semaphore that provides backpressure between the LLM and TTS pipeline stages.
• TurnMetricsCollector — Collects per-turn timing marks and emits a TurnMetricsSummary when each turn finishes.
• TurnTakingController — Manages pause/resume of audio capture during TTS playback.
• TwilioMediaStream — Duplex input/output provider for Twilio Media Streams phone calls.
• VADProcessor — Converts pipeline audio into VAD frames and tracks speech segments.
• VonageAudioSocket — Duplex audio provider for the Vonage Voice API WebSocket endpoint.
• WebLLMLLM — WebLLM in-browser LLM provider.
• WebRTCInput — Browser audio input provider that extracts PCM from a WebRTC track.
• WebRTCOutput — Browser audio output provider that renders TTS audio into a WebRTC track.
• WebSocketManager — Managed WebSocket connection with automatic reconnection and exponential backoff.
• ZoomRtmsInput — Zoom Realtime Media Streams (RTMS) input provider — streams live meeting audio into the CompositeVoice pipeline.