Emome beta 102 — Voice-first (ElevenLabs + Gemini)

This is a drop-in Next.js app you can drag into GitHub and deploy on Vercel.

What’s inside

• Talk to Emome page (/voice) using ElevenLabs Agents via @elevenlabs/react
• After speaking, click Convert to interaction event
• A server route calls Gemini to turn the transcript into:
  • structured event fields (person, intensity, duration, valence, tags)
  • a short reflection
• The Dashboard updates the Relationship Pulse (time-weighted emotional index)

Setup

1. Install deps

npm install

2. Add env vars (copy .env.example → .env.local)

cp .env.example .env.local

3. Create an ElevenLabs Agent

• In ElevenLabs Agents dashboard, create an agent for “Emome voice logging”.
• Recommended: set it as a private agent and use signed URLs (more secure).
• Put your ELEVENLABS_AGENT_ID and ELEVENLABS_API_KEY into .env.local.

Signed URL route used:

• POST /api/elevenlabs/signed-url

Docs reference:

• Signed URLs are the recommended way to secure client-side connections.

4. Get a Gemini API key

• Put it in GOOGLE_GEMINI_API_KEY.

5. Run

npm run dev

Design note (logo + colors)

• This project includes a placeholder public/logo.svg.
• Replace it with your original Emome logo file if you already have one.
• Colors are defined in tailwind.config.ts + app/globals.css.

Deployment (Vercel)

• Add env vars in Vercel project settings:
  • GOOGLE_GEMINI_API_KEY
  • ELEVENLABS_API_KEY
  • ELEVENLABS_AGENT_ID
• Deploy.

Hackathon note

Keep wording non-medical:

• Emome is a “clarity tool” / “relationship tracker” (not therapy, not diagnosis).

Good luck 😈