Voice Agents in Sayvy AI
Voice agents are the central orchestration entity in Sayvy AI. Each agent configures:- Prompt Intelligence: System instructions, situational knowledge, guardrails, and role definitions.
- Speech-to-Text (STT): High-speed acoustic transcription engine (Deepgram Nova-2 or Whisper Realtime).
- Reasoning Engine (LLM): Fast conversational models (Claude 3.5 Sonnet, GPT-4o, or Gemini 1.5 Flash).
- Voice Synthesis (TTS): Low-latency neural voices (Cartesia Sonic, ElevenLabs Turbo v2, or PlayHT).
Life Cycle of a Voice Call
Agent Configuration Parameters
string
required
Internal name used to identify the agent in reports and the console.
string
required
The system instruction provided to the language model during inference.
string
default:"cartesia/sonic-english"
Identifier of the neural voice model used to generate speech.
number
default:"350"
Millisecond duration of user voice activity before halting current agent playback.