AI system that automatically calls elderly patients every morning to check if they took their medication. If something's wrong — they missed doses, they're feeling sick, they're not answering — it texts their family and doctor immediately. Everything is tracked in a dashboard the family can see.
This repo now includes the Person 2 foundation:
- InsForge-ready backend API for patients, call results, escalation logic, and weekly summaries.
- TinyFish FDA monitor endpoint that checks FDA recall/safety RSS and matches against patient medications.
- Live dashboard at
/showing patient call status and FDA match checks.
This project uses the TypeScript 7 native preview toolchain (tsgo) for type-check and build scripts.
For InsForge integration, this project uses the @insforge/sdk client pattern (INSFORGE_URL + INSFORGE_ANON_KEY) with email alerts enabled by default. SMS is deferred as a future enhancement.
- Install dependencies:
npm install- Create environment file:
cp .env.example .env- Start local server:
npm run devOptional validation/build:
npm run check
npm run build- Open:
http://localhost:8080
GET /api/healthGET /api/patientsGET /api/call-resultsPOST /api/call-resultsPOST /api/vapi-outboundPOST /api/vapi-webhookPOST /api/pharmacology/queryGET /api/tinyfish/fda-alerts/:patientIdGET /api/reports/weekly/:patientIdPOST /api/doctor-briefs/generate/:callIdGET /api/doctor-briefs/latest/:patientId
POST /api/call-results accepts:
{
"patient_id": "11111111-1111-4111-8111-111111111111",
"timestamp": "2026-04-24T18:00:00.000Z",
"status": "took_meds",
"transcript": "Yes, I took my pills this morning.",
"flags": [],
"fda_alerts": []
}Alerting behavior:
- escalates on
concern - escalates on 2+ missed doses in recent history
- escalates on 3 consecutive
no_answer