Skip to content

🔁 Event Flow Diagrams

This section visualizes the communication between Wyoming components using Mermaid sequence diagrams.

🎙️ Speech-to-Text (ASR)

sequenceDiagram participant Client participant ASR Client->>ASR: transcribe Client->>ASR: audio-start loop while speaking Client->>ASR: audio-chunk end Client->>ASR: audio-stop ASR-->>Client: transcript

🗣️ Text-to-Speech (TTS)

sequenceDiagram participant Client participant TTS Client->>TTS: synthesize TTS-->>Client: audio-start loop while speaking TTS-->>Client: audio-chunk end TTS-->>Client: audio-stop

🚨 Wake Word Detection

sequenceDiagram participant Client participant Wakeword Client->>Wakeword: detect Client->>Wakeword: audio-start loop stream Client->>Wakeword: audio-chunk alt detected Wakeword-->>Client: detection end end Client->>Wakeword: audio-stop Wakeword-->>Client: not-detected

🎧 Voice Activity Detection

sequenceDiagram participant Client participant VAD loop stream Client->>VAD: audio-chunk alt speech starts VAD-->>Client: voice-started end alt speech ends VAD-->>Client: voice-stopped end end

⏱️ Timer Pipeline

sequenceDiagram participant User participant ASR participant Intent participant Handler participant Timer User->>ASR: "Set a timer for 5 minutes" ASR->>Intent: transcribe Intent->>Handler: parsed intent Handler->>Timer: timer-started Timer-->>Handler: timer-confirmation

🔁 Full Pipeline Flow

sequenceDiagram participant Mic participant Wake participant VAD participant ASR participant Intent participant Handler participant TTS participant Speaker Mic->>Wake: audio-start + audio-chunk Wake-->>VAD: detection VAD->>ASR: audio-start + chunk VAD->>ASR: audio-stop ASR-->>Intent: transcript Intent-->>Handler: intent Handler-->>TTS: synthesize TTS-->>Speaker: audio-start + chunk + stop

🧠 Notes

  • Event streams are asynchronous and bidirectional
  • All diagrams assume persistent socket connections (no HTTP overhead)
  • Services must send describe/info to negotiate capabilities