Voice agents

ProsodyAI streams committed speaker lanes, transcript deltas, physical voice measurements, and speaker-relative change. Your agent runtime decides how those facts affect an LLM or speech synthesizer.

import { VoiceFrame } from '@prosodyai/sdk';

const stream = prosody.stream.connect({
  onDirective(event) {
    const frame = VoiceFrame.fromDirective(event);

    agentContext.update({
      speaker: frame.speakerId,
      pitch: frame.state.intonation.pitch,
      loudness: frame.state.stress.loudness,
      change: frame.change,
      vad: frame.vad,
    });
  },
});

The WebSocket audio clock advances continuously. Speaker attribution and speaker-relative state come from the model's recurrent lanes. Product events carry measurements and committed identity facts.

Agent policy

Build response policy in your application from the measurements your use case needs. For example, compare change.stress.loudness or change.intonation.pitch with the same speaker's prior state, then record the policy decision in your own agent trace.

Event inputs

Speech synthesis remains with your selected provider.