The memory bridge for pediatric care
An AI-powered platform that helps parents of neurodivergent children log daily behavioral episodes effortlessly — and gives doctors the clinical data they actually need.
Every day, millions of parents in India care for children with ADHD and Autism. When a meltdown happens — screaming, crying, throwing things — the parent manages it, calms things down, and by morning, the details are gone.
Weeks later at the doctor's office, when asked "How has he been?", the parent says:
"He had a few rough days... I think the medicine is helping a little, not sure."
The doctor, working with only that, adjusts treatment based on guesswork. This cycle repeats every 4-8 weeks.
No accurate data. No patterns. No real picture.
That gap is exactly what Nena is built to close.
Nena is a web application that makes daily behavior tracking effortless. Every day, the parent talks to Nena for 90 seconds — through a voice call, chat, or quick form — and the AI quietly turns that conversation into structured clinical notes using the ABC method:
| Step | What It Means | Example |
|---|---|---|
| A — Antecedent | What happened before the episode? | "Someone touched his bag on the bus" |
| B — Behavior | What did the child do? | "Cried and threw things for 20 minutes" |
| C — Consequence | What happened after? What helped? | "Calmed down with a sensory ball" |
The doctor sees the full picture. The parent doesn't have to remember a thing.
- 🗣️ Voice Check-in — Talk to Nena via VAPI-powered voice calls. No typing needed
- 💬 AI Chat — Chat naturally in English, Hindi, or Hinglish. Nena extracts the clinical data automatically
- 📝 Quick Form — Fill a simple ABC form when you prefer structured input
- 📊 Timeline — See your child's full behavioral history, severity trends, and patterns
- 🎨 ABC Visualizations — AI-generated 3D clay renders that bring each episode to life
- 📲 WhatsApp Reminders — Gentle nudges if you miss a day
- 🧠 AI Thinking Mode — Ask Nena deep questions about your child's patterns
- 📋 Patient Dashboard — All assigned children at a glance with status indicators
- 🔍 RAG-Powered AI Chat — Ask questions like "Show me sensory triggers in the last 30 days" and get data-backed answers
- 📄 PDF Reports — Generate hospital-grade clinical reports with one click
- 📈 Severity Trends — Visualize behavioral patterns over time
- 👤 User Management — Create parent and doctor accounts
- 🔗 Relationship Linking — Connect parents ↔ children ↔ doctors
- 📊 Platform Statistics — Overview of all users, children, and assignments
graph TD
subgraph Frontend["🌐 Next.js 16 + Tailwind v4 + Shadcn/UI"]
AdminUI["Admin Dashboard"]
ParentUI["Parent Portal"]
DoctorUI["Doctor Dashboard"]
end
subgraph Backend["⚡ FastAPI — Python 3.12+"]
AuthAPI["Auth — JWT httpOnly Cookies"]
BehaviorAPI["Behavior Logs + ABC Extraction"]
ChatAPI["AI Chat — 4 Modes"]
ReportsAPI["PDF Reports — ReportLab"]
WebhooksAPI["Voice Webhooks"]
end
subgraph AI["🧠 AI Services"]
Groq["Groq LLM — Llama 3.1/3.3"]
Gemini["Gemini 2.5 Flash — Prompt Enhancement"]
Flux["Cloudflare FLUX-1-schnell — ABC Images"]
Sarvam["Sarvam AI — Hinglish STT"]
end
subgraph Data["🗄️ Supabase — Mumbai Region"]
Postgres["PostgreSQL + pgvector"]
Storage["File Storage"]
end
Frontend -->|"REST + httpOnly cookies"| Backend
Backend --> AI
Backend --> Data
ParentUI -->|"@vapi-ai/web"| WebhooksAPI
| Layer | Technology | Purpose |
|---|---|---|
| Frontend | Next.js 16.2, React 19, Tailwind v4, Shadcn/UI | Responsive web app with role-based dashboards |
| Backend | FastAPI, Python 3.12+, Uvicorn | REST API, business logic, AI orchestration |
| Database | Supabase PostgreSQL + pgvector (Mumbai) | Data storage, vector embeddings, RLS |
| AI/LLM | Groq (Llama 3.1 8B + 3.3 70B) | ABC extraction, RAG chat, summaries |
| Image Gen | Cloudflare FLUX-1-schnell + Gemini 2.5 Flash | 3D clay-style ABC scene visualizations |
| Voice | VAPI AI | In-browser voice calls for daily check-ins |
| STT | Sarvam AI Saaras V3 | Hinglish-optimized speech-to-text |
| Notifications | Twilio WhatsApp Business API | Daily behavior logging reminders |
| ReportLab | Hospital-grade clinical PDF reports | |
| Auth | Custom JWT (PyJWT + bcrypt) | httpOnly cookies, role-based access control |
| Package Mgmt | uv (backend), npm (frontend) | Fast, reliable dependency management |
Nena/
├── assets/ # Logo and brand assets
├── backend/ # FastAPI backend
│ ├── app/
│ │ ├── main.py # App entry + admin seed + scheduler
│ │ ├── config.py # Environment configuration
│ │ ├── dependencies.py # FastAPI dependency injection
│ │ ├── api/v1/ # All REST endpoints (9 routers)
│ │ ├── core/ # Permissions and exception handling
│ │ ├── models/ # Pydantic schemas + enums
│ │ └── services/ # AI, auth, reports, notifications, STT
│ └── pyproject.toml # Python dependencies (uv)
├── frontend/ # Next.js frontend
│ ├── src/
│ │ ├── app/
│ │ │ ├── (admin)/ # Admin dashboard pages
│ │ │ ├── (parent)/ # Parent portal pages
│ │ │ ├── (doctor)/ # Doctor dashboard pages
│ │ │ └── login/ # Authentication
│ │ ├── components/ui/ # Shadcn/UI components
│ │ └── lib/ # Auth context, API helpers, types
│ └── package.json
├── .env # Unified environment variables
├── ARCHITECTURE.md # System design document
├── DATABASE_SCHEMA.md # Full SQL schema + RLS policies
├── FEATURE_SPEC.md # Feature requirements
├── TECH_STACK.md # Technology decisions
├── ROADMAP.md # Development roadmap
├── PROMPT_LIBRARY.md # All AI system prompts
└── SETUP_FOR_TEAM.md # Team onboarding guide
- Node.js v20+ — nodejs.org
- Python 3.12+ — python.org
- uv — Fast Python package manager (install guide)
git clone https://github.com/your-username/nena.git
cd nenaCreate a .env file in the project root:
# ── Supabase ──
SUPABASE_URL=https://<YOUR_PROJECT>.supabase.co
SUPABASE_ANON_KEY=<YOUR_ANON_KEY>
SUPABASE_SERVICE_ROLE_KEY=<YOUR_SERVICE_ROLE_KEY>
# ── Auth ──
JWT_SECRET_KEY=your-secret-key-change-in-production
# ── AI ──
GROQ_API_KEY=<YOUR_GROQ_KEY>
GEMINI_API_KEY=<YOUR_GEMINI_KEY>
# ── Voice (VAPI) ──
VAPI_PUBLIC_KEY=<YOUR_VAPI_PUBLIC_KEY>
VAPI_PRIVATE_KEY=<YOUR_VAPI_PRIVATE_KEY>
VAPI_ASSISTANT_ID=<YOUR_VAPI_ASSISTANT_ID>
# ── WhatsApp (Twilio) ──
TWILIO_ACCOUNT_SID=<YOUR_SID>
TWILIO_AUTH_TOKEN=<YOUR_TOKEN>
TWILIO_WHATSAPP_NUMBER=<YOUR_NUMBER>
# ── Image Generation (Cloudflare) ──
CLOUDFLARE_ACCOUNT_ID=<YOUR_ACCOUNT_ID>
CLOUDFLARE_API_TOKEN=<YOUR_TOKEN>
# ── Frontend ──
FRONTEND_URL=http://localhost:3000
NEXT_PUBLIC_API_URL=http://localhost:8000
NEXT_PUBLIC_SUPABASE_URL=https://<YOUR_PROJECT>.supabase.co
NEXT_PUBLIC_SUPABASE_ANON_KEY=<YOUR_ANON_KEY>cd backend
uv sync
uv run uvicorn app.main:app --reload --host 0.0.0.0 --port 8000cd frontend
npm install
npm run devOpen http://localhost:3000/login and use the default admin credentials:
| Field | Value |
|---|---|
admin@gmail.com |
|
| Password | admin123 |
The admin account is automatically created on first backend startup.
Parent speaks naturally Structured clinical data
│ │
▼ ▼
┌─────────────────┐ ┌──────────────┐ ┌─────────────────┐
│ "Aaj Arjun ka │───▶│ Groq LLM │───▶│ antecedent: │
│ school se aate │ │ Llama 3.1 │ │ "Touched bag" │
│ hi meltdown │ │ 8B Instant │ │ behavior: │
│ hua..." │ │ │ │ "Crying 20m" │
└─────────────────┘ └──────────────┘ │ severity: 4 │
│ ai_summary: │
│ "Arjun had a │
│ meltdown..." │
└─────────────────┘
Key design decisions:
- Plain text enforcement — AI responses never include markdown, ensuring clean rendering across all UI surfaces
[LOG_SESSION_NOW]trigger — When the parent chat AI collects all three ABC pieces, it appends this hidden signal. The frontend detects it and auto-saves the behavior log- Hinglish-first — All prompts handle English, Hindi, and Hinglish seamlessly
- No hallucination — Missing data uses clinical placeholders like "Not specifically identified" instead of
null
| Document | What's Inside |
|---|---|
| ARCHITECTURE.md | System design, data flows, module responsibilities |
| DATABASE_SCHEMA.md | All 10 tables, RLS policies, indexes, SQL migrations |
| FEATURE_SPEC.md | Detailed feature requirements for all 12 modules |
| TECH_STACK.md | Technology choices and rationale |
| ROADMAP.md | 6-phase development roadmap with progress tracking |
| PROMPT_LIBRARY.md | Every AI system prompt used in the application |
| SETUP_FOR_TEAM.md | Team onboarding and local setup guide |
10 tables powering the platform:
| Table | Purpose |
|---|---|
profiles |
All users — admin, parent, doctor (with bcrypt passwords) |
children |
Child profiles — diagnosis, medications, triggers |
parent_child_links |
Parent ↔ Child relationships |
doctor_child_assignments |
Doctor ↔ Child assignments |
behavior_logs |
Core ABC data with severity, categories, AI analysis |
consent_records |
DPDP Act compliance — consent tracking |
call_logs |
Voice call metadata and recordings |
ai_summaries |
Weekly/monthly AI-generated behavior summaries |
generated_reports |
PDF report tracking and storage |
notification_logs |
WhatsApp reminder delivery tracking |
All tables are protected by Row-Level Security (RLS) — parents see only their children, doctors see only assigned patients, admins see everything.
| Phase | Feature | Status |
|---|---|---|
| 0 | Project Setup & Infrastructure | ✅ Complete |
| 1A | Database + Auth | ✅ Complete |
| 1B | Admin Dashboard (CRUD) | ✅ Complete |
| 2 | Parent Quick Form + Timeline | ✅ Complete |
| 3A | AI Pipeline (Groq ABC Extraction) | ✅ Complete |
| 3B | Parent AI Chat | ✅ Complete |
| 4A | Doctor Dashboard + Patient List | ✅ Complete |
| 4B | Doctor RAG Chat | ✅ Complete |
| 5A | PDF Reports (ReportLab) | ✅ Complete |
| 5B | Voice Integration (VAPI AI) | ✅ Complete |
| 6A | WhatsApp Nudges (Twilio) | ✅ Complete |
| 6B | Polish & Testing | 🟡 In Progress |
- httpOnly cookies — Tokens are never exposed to JavaScript
- bcrypt password hashing — Industry-standard password security
- Role-based access control —
require_admin,require_parent,require_doctordecorators - Row-Level Security — Database-level data isolation per user
- DPDP Act compliance — Consent tracking tables, Mumbai-region data residency
- Soft deletes — No data is permanently deleted;
is_activeflags used throughout
- Fork the repository
- Create your feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'Add amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
Please read SETUP_FOR_TEAM.md for development setup instructions.
This project is part of an academic/hackathon initiative for pediatric behavioral care in India.
Built with 💜 for every parent who's doing their best