Releif.AI

``` ██████╗ ███████╗██╗ ██╗███████╗███████╗ █████╗ ██╗ ██╔══██╗██╔════╝██║ ██║██╔════╝██╔════╝██╔══██╗██║ ██████╔╝█████╗ ██║ ██║█████╗ █████╗ ███████║██║ ██╔══██╗██╔══╝ ██║ ██║██╔══╝ ██╔══╝ ██╔══██║██║ ██║ ██║███████╗███████╗██║██║ ███████╗██║ ██║██║ ╚═╝ ╚═╝╚══════╝╚══════╝╚═╝╚═╝ ╚══════╝╚═╝ ╚═╝╚═╝ ``` ### When disaster strikes, every second matters. **ReliefAI** is a full-stack disaster response platform — built so citizens can get help fast, and responders can see the full picture.
[](https://nextjs.org/) [](https://www.postgresql.org/) [](https://fastapi.tiangolo.com/) [](https://langchain-ai.github.io/langgraph/) [](LICENSE)
[Try the demo](#-see-it-in-action) · [Quick start](#-get-running-in-5-minutes) · [Architecture](#-how-it-all-fits-together) · [API](#-api-at-a-glance)

---

Why we built this

Disasters don't wait for paperwork. Someone trapped in a flood needs to report it now. A coordinator needs to see every open request on one map. A volunteer needs to know where to go.

ReliefAI brings those pieces into one place:

Citizens report and track emergencies without logging in.
Ops teams dispatch volunteers, monitor risk, and run predictions from a live dashboard.
AI helps with flood forecasting and first-aid guidance when every minute counts.

No silos. No "check three different tools." Just one system that actually talks to itself.

---

See it in action

Spin it up locally, then open these in your browser:

What	URL	What you'll see
Landing	localhost:3000	The public face — calm, clear, ready
Get Help	localhost:3000/get-help	3-step emergency report wizard
Track a request	localhost:3000/tracking?code=RF-7K29X4P1	Live status for a seeded demo request
Ops dashboard	localhost:3000/dashboard	Map, triage queue, activity feed
Health check	localhost:3000/api/health	JSON status for all services
AI Engine docs	localhost:8000/docs	OpenAPI for prediction API
AI Centre	localhost:7860	Disaster chat + first-aid API

Demo tracking code: RF-7K29X4P1 — try it after running ./scripts/setup.sh

---

Two worlds, one platform

Citizens and responders use different interfaces — on purpose. People in crisis shouldn't need an ops login. Teams on the ground shouldn't wade through a marketing page.

flowchart LR
    subgraph CITIZEN["Citizen portal — no login"]
        direction TB
        c1[Landing]
        c2[Get Help wizard]
        c3[Tracking page]
        c1 --> c2 --> c3
    end

    db[(PostgreSQL)]

    subgraph OPS["Operations console"]
        direction TB
        o1[Dashboard]
        o2[Live map]
        o3[Emergency queue]
        o4[Volunteers]
        o5[AI prediction]
        o1 --> o2 & o3 & o4 & o5
    end

    c2 -->|creates request| db
    db -->|status sync| c3
    o3 -->|manage requests| db
    o4 -->|assign volunteer| db

---

How it all fits together

Three services, one monorepo. The webapp is the hub — everything else plugs in.

flowchart TB
    subgraph WEB["webapp — Next.js 16"]
        direction TB
        pub[Public UI]
        ops[Ops UI]
        api[REST API]
        pub --> api
        ops --> api
    end

    db[(PostgreSQL)]
    engine[ai-engine]
    centre[ai-centre]

    api --> db
    api --> engine
    api --> centre

Plain-text version (for terminals & copy-paste) ``` ┌─────────────────────────────────────────────────────────────┐ │ webapp (Next.js 16) — http://localhost:3000 │ │ Landing · Get Help · Tracking · Ops Dashboard · Maps │ │ │ │ /api/emergency-requests ──► PostgreSQL (:5432) │ │ /api/volunteers ──► PostgreSQL │ │ /api/predictions/run ──► AI Engine (:8000) │ │ /api/first-aid ──► AI Centre (:7860) │ └─────────────────────────────────────────────────────────────┘ │ │ ▼ ▼ ┌─────────────────────┐ ┌─────────────────────────┐ │ ai-engine/ │ │ ai-centre/ │ │ LangGraph + FAISS │ │ FastAPI + Groq API │ │ Flood prediction │ │ Disaster chat assistant│ └─────────────────────┘ └─────────────────────────┘ ```

---

What's inside

### For people in crisis - **Get Help** — guided wizard: type, urgency, location, contact - **Tracking** — look up status with a single code, no account - **Mobile-friendly** — works when networks are shaky	### For response teams - **Live dashboard** — triage queue, stats, activity feed - **Map view** — emergencies + volunteers on one canvas - **Volunteer dispatch** — assign, track, complete missions - **AI prediction** — flood/disaster risk with LangGraph + RAG
### Under the hood - **PostgreSQL + Prisma** — real persistence, not mock JSON files - **Seed data included** — 6 emergencies, 4 volunteers, demo predictions - **Graceful fallbacks** — if AI services are down, the app keeps working with mock providers

---

Get running in 5 minutes

You'll need Node 20+, Python 3.11+, and either Docker (easiest for Postgres) or a local PostgreSQL 16 install.

1 · Clone & one-shot setup

git clone <your-repo-url> reliefai
cd reliefai
chmod +x scripts/*.sh
./scripts/setup.sh

This installs dependencies, copies env templates, starts Postgres, runs migrations, and seeds demo data. Grab a coffee — it does the boring stuff for you.

2 · Add your keys (optional but recommended)

cp .env.example .env
cp webapp/.env.local.example webapp/.env.local   # if present

Open .env and set:

GROQ_API_KEY=gsk_...   # powers AI prediction + disaster chat

No key? The app still runs — AI features fall back to mock responses so you can explore the UI.

3 · Start everything

./scripts/start-all.sh
# or: npm run dev

4 · Sanity check

npm run health
# or: curl http://localhost:3000/api/health

You should see all services reporting ok (or degraded if AI keys aren't set — that's fine for local dev).

Manual Postgres setup (no Docker) ```bash psql -U postgres -c "CREATE USER reliefai WITH PASSWORD 'reliefai';" psql -U postgres -c "CREATE DATABASE reliefai OWNER reliefai;" cd webapp && npx prisma db push && npm run db:seed ``` Default connection string: ``` postgresql://reliefai:reliefai@localhost:5432/reliefai ``` > **Heads up:** Prisma reads `webapp/.env`, not `.env.local`. Make sure `DATABASE_URL` is there.

---

Project layout

reliefai/
├── webapp/                 Next.js 16 — UI + REST API + Prisma
│   ├── prisma/             Schema, seed data
│   └── src/
│       ├── app/(citizen)/  Public routes: get-help, tracking
│       ├── app/(app)/      Ops console: dashboard, map, admin
│       └── components/     Landing, wizard, maps, design system
│
├── ai-engine/              LangGraph flood prediction (FastAPI :8000)
├── ai-centre/              Groq disaster assistant (FastAPI :7860)
├── scripts/                setup.sh · start-all.sh · stop-all.sh
├── docker-compose.yml      Postgres (+ optional full stack)
└── .env.example            Environment template

---

API at a glance

Endpoint	Method	Does what
/api/emergency-requests	GET, POST	List or create emergencies
/api/emergency-requests/track?code=	GET	Citizen lookup by tracking code
/api/emergency-requests/[id]	PATCH	Update status, dispatch
/api/volunteers	GET, POST	Volunteer registry
/api/volunteers/[id]/assign	POST	Assign to a request
/api/predictions/run	POST	Run disaster prediction (AI Engine)
/api/first-aid	POST	AI first-aid guidance (AI Centre)
/api/activity	GET	Ops activity feed
/api/health	GET	Ping all services

---

Environment variables

Variable	Where	Default	Notes
DATABASE_URL	webapp	postgresql://reliefai:reliefai@localhost:5432/reliefai	Required for persistence
AI_ENGINE_URL	webapp	http://localhost:8000	Prediction engine
AI_CENTRE_URL	webapp	http://localhost:7860	Disaster chat
GROQ_API_KEY	ai-engine, ai-centre	—	Enables real LLM responses
GOOGLE_API_KEY	ai-engine	—	Optional embeddings fallback

---

Database commands

cd webapp
npx prisma db push      # apply schema
npm run db:seed         # load demo emergencies + volunteers
npx prisma studio       # browse data visually (optional)

---

Docker

# Just Postgres (recommended for local dev)
docker compose up -d postgres

# Full stack
docker compose up --build

---

Production build

cd webapp
npm run build
npm start

Set DATABASE_URL, AI_ENGINE_URL, and AI_CENTRE_URL in your deployment environment.

---

Something broke?

We've hit these too. Here's what usually fixes it:

Symptom	Likely fix
DATABASE_URL not found	Create webapp/.env with your Postgres URL — Prisma ignores .env.local
Port 3000 / 8000 / 7860 in use	./scripts/stop-all.sh, then restart
AI Engine shows offline	Add GROQ_API_KEY to .env, or ignore it — mocks kick in automatically
CSS build error after pull	Restart the dev server (Turbopack cache quirk)
Empty dashboard	Run cd webapp && npm run db:seed

Still stuck? Open an issue — we'd rather fix the docs than leave you guessing.

---

Built with

Layer	Stack
Frontend	Next.js 16, React 19, Tailwind CSS 4, Motion, Leaflet
Backend	Next.js API routes, Prisma ORM, PostgreSQL 16
AI Engine	Python, FastAPI, LangGraph, FAISS, Groq
AI Centre	Python, FastAPI, Groq (openai/gpt-oss-120b)

---

Standing on the shoulders of

ReliefAI started as three projects that deserved to work together:

Origin	Now lives in	Contribution
reliefai_v3	webapp/	Landing, citizen flow, ops dashboard
Releif.ai	ai-engine/	LangGraph agentic flood analysis
AI-centre	ai-centre/	Disaster management chat

---

License

MIT — see LICENSE. Use it, fork it, improve it. If ReliefAI helps even one person get help faster, it was worth building.

---

**Predict. Respond. Recover.** *Built for the moments when software actually matters.*

Built With

• css
• dockerfile
• javascript
• python
• shell
• typescript