A communication system for aphasia patients that speaks in their own voice, learns who they are through every interaction, and gets smarter every time their caregiver answers one question.

1 in 3 stroke survivors develops aphasia — the inability to reliably form words despite fully intact understanding. The standard hardware solution costs $15,000. Speech therapy runs $200 per session. Most patients have nothing.

Clarivo runs on any laptop or tablet, for free.

The patient taps large icon-based buttons to express what they need. The system converts their selection into a natural sentence and speaks it aloud — in their own cloned voice. A caregiver panel sits alongside, giving full visibility and control to the person caring for them.

• Voice cloning — patient records 60 seconds of their voice during onboarding. ElevenLabs clones it. Every sentence the system speaks sounds like them.
• Icon-based communication — two input modes: a navigable decision tree and a free-form icon composer for more expressive combinations
• Personalization engine — caregiver fills in medical context, preferences, and daily routine. OpenAI uses all of it on every generation.
• Post-session learning — after every session, the system asks the caregiver one targeted question to fill a knowledge gap. The system gets more accurate over time.
• Knowledge score — a visible 0-100% bar showing how well the system knows the patient, broken down by profile, medical, preferences, and conversation history
• Adaptive shortcuts — frequently used paths surface automatically as one-tap shortcuts
• Phrase predictions — time-of-day aware suggestions before the patient taps anything
• Urgency alerts — red alert fires if patient expresses distress 3+ times within 2 hours
• Two-voice system — patient speaks in their cloned voice, caregiver-to-patient communication uses a distinct neutral voice
• Caregiver dashboard — 30-day analytics, daily AI digest, exportable PDF clinical report
• Offline resilience — common paths cached locally, app survives wifi failure

Clarivo/
├── shared/
│   ├── api-contract.ts      # API types and contracts
│   └── mock-data.ts         # Mock responses for frontend dev
├── frontend/                # Next.js app (Engineer 1)
├── backend-ai/              # OpenAI + ElevenLabs service (Engineer 2)
└── backend-data/            # MongoDB data service (Engineer 3)

To get Clarivo running on your machine, follow these steps. If you don't have API keys (OpenAI/ElevenLabs) or MongoDB installed, you can still run the app in Offline Mode.

Each service needs its own configuration file. Follow these steps to set them up:

AI Backend (E2):

cd backend-ai
cp .env.example .env
# Edit .env and set USE_MOCK=true if you don't have API keys

Data Backend (E3):

cd backend-data
cp .env.example .env
# Edit .env and set USE_MOCK_DB=true to run without MongoDB

Frontend:

cd frontend
cp .env.example .env.local

AI Backend (Port 8001):

cd backend-ai
pip install -r requirements.txt
python main.py

Data Backend (Port 8002):

cd backend-data
pip install -r requirements.txt
python main.py

Frontend (Port 3000):

cd frontend
npm install
npm run dev

If you or your collaborators want to test the UI without setting up OpenAI, ElevenLabs, or MongoDB:

1. In backend-ai/.env, set USE_MOCK=true.
2. In backend-data/.env, set USE_MOCK_DB=true.
3. Start all three services as described above.

The app will use pre-defined mock responses for all AI and database operations, allowing you to navigate the UI and test the buttons immediately.

Open the app and click "Load Demo Profile" on the onboarding screen. This seeds the database with a complete patient profile (Yuki) and 30 days of session history, then drops you straight into the patient view.

# Full suite
python tests/runner.py --all

# Pre-demo check only
python tests/runner.py --demo

Built for Los Altos Hacks X targeting:

• Best Use of ElevenLabs
• Best Use of Gen AI
• Best Use of MongoDB Atlas
• Best Social Impact
• Best UI/UX
• Best .Tech Domain — Clarivo.tech

Built at Los Altos Hacks X · April 11–12, 2026