Skip to content

Thymia Sentinel

Voice biomarker platform for real-time conversation intelligence.

Sentinel streams voice conversations to Thymia's Lyra server for real-time extraction of speech biomarkers, combined with policy-based reasoning to surface insights that text-only systems miss.

  • Voice Biomarkers

    Extract clinical-grade and wellness biomarkers from speech in real-time.

  • Custom Policies

    Define policies that combine biomarkers with conversation context to drive actions.

  • Framework Agnostic

    Works with LiveKit, Pipecat, VAPI, Gemini, and any voice AI stack that produces audio.

  • Any Domain

    Mental health safety, contact centers, education, coaching, employee wellness, and more.

How It Works

  1. Stream audio from your voice application to Sentinel
  2. Biomarkers are extracted in real-time by Thymia's Lyra server
  3. Policies analyze biomarkers + conversation context
  4. Actions are returned for your application to act on

Policies are configured with Thymia based on your use case. In code, you simply reference which policies to run:

sentinel = SentinelClient(
    user_label="user-123",
    policies=["demo_wellbeing_awareness"],  # Your configured policies
)

Use Cases

Domain Example Policies What You Learn
Mental Health Safety Risk classification, crisis detection Distress levels, minimization detection, recommended interventions
Contact Centers Caller frustration, agent burnout Escalation risk, when to transfer, agent wellness alerts
Education Student anxiety, tutor wellbeing Learning barriers, engagement drops, burnout risk
Coaching Client engagement, session effectiveness Emotional state, receptiveness, session quality
Employee Wellness Stress monitoring, burnout detection Team health trends, early intervention signals

Quick Start

pip install thymia-sentinel

from thymia_sentinel import SentinelClient

sentinel = SentinelClient(
    user_label="user-123",
    policies=["your-policy"],
)

@sentinel.on_policy_result
async def handle_result(result):
    # Policy results contain actions for your application
    actions = result["result"].get("recommended_actions", {})
    if actions.get("for_agent"):
        print(f"Agent guidance: {actions['for_agent']}")

await sentinel.connect()

# Stream audio and transcripts from your voice app
await sentinel.send_user_audio(audio_bytes)
await sentinel.send_user_transcript("transcript text")

await sentinel.close()

Framework Integrations

Sentinel provides plug-and-play integrations for popular voice AI frameworks:

  • LiveKit — Automatic audio capture from LiveKit rooms
  • Pipecat — FrameProcessor integration for Pipecat pipelines
  • VAPI — WebSocket transport integration
  • Gemini Live — Google Gemini Live API integration

Getting Access

Sentinel requires an API key from Thymia. Contact us to get started.

License

MIT License — see LICENSE for details.