SimCity AI — Autonomous Digital Twin for Policy Testing

An autonomous multi-agent system that creates a living digital twin of Kochi, India to simulate, test, and optimize public policy decisions before real-world implementation.

Architecture

User Input (Policy Description)
        │
        ▼
┌─────────────────────────────────────┐
│     LangGraph Multi-Agent Pipeline  │
│                                     │
│  1. Data Ingestion Agent            │  ← OpenStreetMap via OSMnx
│  2. Simulation Engine Agent         │  ← NetworkX traffic model
│  3. Citizen Proxy Agent             │  ← Claude LLM (5 demographics)
│  4. Policy Testing Agent            │  ← Claude LLM + graph mutation
│  5. Impact Analysis Agent           │  ← Quantified before/after metrics
│  6. Recommendation Agent            │  ← Claude LLM policy report
└─────────────────────────────────────┘
        │
        ▼
FastAPI + SSE → Next.js Dashboard

Tech Stack

Component	Technology
Multi-agent orchestration	LangGraph
LLM	Anthropic Claude (claude-sonnet-4-6)
City data	OSMnx (OpenStreetMap)
Traffic simulation	NetworkX
Backend API	FastAPI + Server-Sent Events
Frontend	Next.js 14 + Tailwind CSS
Map	Leaflet.js

Quick Start

1. Clone & configure environment

cp .env.example backend/.env
# Edit backend/.env and add your ANTHROPIC_API_KEY

2. Start the backend

cd backend
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
uvicorn main:app --reload --port 8000

The first run will fetch Kochi's road network from OpenStreetMap (~30 seconds). Subsequent runs use the local cache.

3. Start the frontend

cd frontend
npm install
npm run dev

Open http://localhost:3000

Policy Scenarios (Demo Presets)

Preset	Type	Description
Pedestrianize MG Road	Road Closure	Convert Mahatma Gandhi Road to car-free pedestrian zone
Add BRT on NH-66	New Route	Bus Rapid Transit corridor from Edappally to Tripunithura
AI Signal Optimization	Signal Timing	Adaptive signals at 15 key intersections
Expand Water Taxi	Transit Add	8 new ferry routes across Vembanad backwaters

How It Works

1. Enter a policy — select a preset or describe any traffic/mobility policy in plain English
2. Watch agents run — 6 agents execute sequentially with real-time status updates via SSE
3. See the impact — color-coded map shows before/after congestion levels
4. Read the report — Claude generates a structured Go/No-Go policy recommendation

Application Images

API Endpoints

Method	Endpoint	Description
GET	/api/city/kochi	Kochi road network as GeoJSON
GET	/api/city/kochi/metrics	Baseline traffic metrics
GET	/api/presets	Pre-built policy presets
POST	/api/simulate	Start simulation, returns job_id
GET	/api/simulate/{id}/stream	SSE stream of agent updates
GET	/api/simulate/{id}/result	Final simulation result

Project Structure

simcity-ai/
├── backend/
│   ├── agents/          # 6 LangGraph agent nodes
│   ├── core/            # City model, state schema
│   ├── workflow/        # LangGraph graph definition
│   ├── api/             # FastAPI routes + SSE
│   ├── data/            # Cached OSM graph (auto-generated)
│   └── main.py
├── frontend/
│   ├── app/             # Next.js App Router pages
│   ├── components/      # UI components
│   └── lib/             # API client + TypeScript types
└── .env.example