Moeabdelaziz007 SKILL: Voice Control (Gemini Live)
Description
Implements a simulation of the Gemini Live bidirectional streaming protocol. It handles InputAudio events from the user and generates OutputAudio events using ADK patterns.
Architecture
Based on google.adk.models.gemini_llm_connection:
- Audio In: User speech is captured as chunks.
- Transcode: Audio is converted to text (STT) for the context window (if needed) or processed directly by multimodal models.
- Audio Out: The agent response is synthesized (TTS).
Capabilities
process_audio_stream(stream_id):- Simulates connecting to a WebSocket.
- Listens for interrupt signals.
speak(text_or_ssml):- Synthesizes speech.
- Supports SSML for tone/pitch control (e.g.,
<speak>Hello</speak>).
Usage Prompt
"Agent, activate voice mode. Listen for my command 'Deploy' and confirm verbally."