🛡️ AARAKSHAKA — Crime Intelligence Platform

AI-driven crime intelligence platform for Karnataka State Police

Transforms raw FIR data (~433K records across 27 relational tables) into actionable, visual, and bilingual insights — powered by Zoho Catalyst.

Live Demo: aarakshaka-kar.onslate.in

🎯 Problem Statement

Karnataka State Police manages crime data across 30 districts and 240 police stations. Today, that data lives in fragmented spreadsheets, disconnected databases, and manual SCRB reports. Officers cannot:

See crime patterns across districts in real time
Identify repeat offenders and their criminal networks
Get AI-powered explanations for why crimes spike in certain areas
Access insights in Kannada for field-level officers
Correlate crime with socio-economic factors like literacy and urbanization

The result: reactive policing. Crimes happen, reports get filed, and patterns are only visible weeks later — if at all.

🛡️ Why AARAKSHAKA

AARAKSHAKA (Protector) flips this to proactive, intelligence-driven policing. It takes raw FIR data and transforms it into visual, interactive, and AI-generated insights that any officer — from a constable in Raichur to the DGP in Bangalore — can understand and act on.

Real-world Impact

User	Benefit
🚔 Constable / IO	Instant hotspot maps, case status, offender profiles
📍 Station SHO	Resource allocation, crime trend awareness
🏛️ SP / DCP	District comparison, resource allocation, strategy
🏛️ DGP / HQ	State-wide overview, policy decisions, resource planning
🌾 Rural Officers	Full Kannada support — no language barrier

📸 Application Screenshots

Login

Dashboard — Overview

Desktop (EN)	Desktop (KN)

Dashboard — Analysis

Desktop (EN)

Dashboard — Operations

Desktop (EN)

Crime Map

Desktop (EN)	District Panel

Network Graph

Desktop (EN)

Offenders

Desktop (EN)	Person Detail

Predictions

Desktop (EN)	Risk Map Popup

Predictions

Scrolled (Predictions)	Scrolled (AI Generated Insights)

AI Insights (Kannada)

📱 Tablet View

Login	Dashboard	Crime Map	Kannada

✨ Key Features

🎛️ Dashboard & Analytics

📊 KPI Cards — Total cases, active investigations, chargesheet rate, conviction rate
📈 Monthly Trends — Year-over-year comparison with multi-year overlay
🗺️ District Breakdown — Cases by district with interactive filtering
🥧 Crime Categories — Theft, assault, murder, cybercrime breakdown
🍩 Case Status — Under investigation vs. chargesheeted vs. closed
📅 Day-of-Week — Which days see the most crime
⚖️ Gravity Chart — Serious vs. non-serious crime ratio over time
📊 Socio-Economic Correlation — Census 2011 data (literacy, urbanization, sex ratio) vs. crime rates

🗺️ Interactive Crime Map

🗺️ 30 District Polygons — Karnataka GeoJSON boundaries
📍 Crime Markers — Individual FIR locations with lat/lng coordinates
🎯 Risk-Level Color Coding — Green (low) → Yellow (moderate) → Red (high)
🎯 District Filtering — Click a district to zoom and filter all data
📍 Point-in-Polygon Resolution — Automatic district assignment from coordinates

🕸️ Criminal Network Graph (3D)

🕸️ 3D Force-Directed Graph — Three.js + react-force-graph-3d
🎮 Orbit Controls — Rotate, zoom, pan the 3D graph
🏷️ 3D Text Labels — Names rendered directly on nodes
🎯 Focus-on-Node — Click any node to center the graph on that person
👥 Offender Sidebar — Ranked list of most frequent accused
👤 Person Detail Panel — Cases, charges, associates, timeline

🔮 Predictions & Risk Scoring

📈 30-Day Forecast — Weighted moving average with seasonal multipliers
🎯 District Risk Scores — Composite: Volume (40%) + Severity (30%) + Trend (30%)
🚨 Anomaly Detection — Z-score based alerts for unusual spikes
📝 Risk Explanation — AI-generated narrative with actionable recommendations

🤖 AI-Powered Insights

🧠 Natural Language Analysis — GLM-4.7-Flash (30B MoE) via QuickML
🗣️ Bilingual Output — English + Kannada
📝 Crime Insights — "Violent crimes in Bangalore Urban spiked 42% in December..."
📋 Risk Explanations — "Raichur has high risk due to 60% increase in theft..."
⚡ Statistical Fallback — Data-driven insights when LLM times out

🌐 Bilingual Support

🌐 425+ Translation Keys per language
🌐 Language Toggle — EN | ಕನ್ನಡ in header
💾 localStorage Persistence — Language choice survives page refresh
🔄 Fallback — Missing keys fall back to English

🏗️ Architecture

flowchart TD
    subgraph Client["Frontend — React 18 + Vite + Tailwind CSS"]
        direction LR
        A1["Dashboard<br/>Recharts"]
        A2["Crime Map<br/>Leaflet"]
        A3["3D Network<br/>Three.js"]
        A4["Predictions<br/>Forecast + Anomaly"]
        A5["PDF Reports<br/>SmartBrowz"]
        A6["EN / KN<br/>i18next"]
    end

    subgraph API["Backend — Python 3.13 + Flask · 28 Endpoints"]
        direction TB
        B1["REST API Layer"]
        B2["In-Memory Cache<br/>2hr TTL"]
        B3["ZCQL Query Builder"]
    end

    subgraph Platform["Zoho Catalyst Platform"]
        direction TB
        C1["Functions<br/>Advanced I/O"]
        C2["Data Store<br/>27 Tables"]
        C3["Cache"]
        C4["Connections<br/>OAuth 2.0"]
        C5["QuickML<br/>GLM-4.7-Flash"]
        C6["SmartBrowz<br/>PDF Rendering"]
        C7["Slate<br/>Static Hosting"]
    end

    A1 & A2 & A3 & A4 & A5 & A6 -->|"REST API"| B1
    B1 --> B2 --> B3 --> C2
    B1 -->|"LLM Insights"| C4 --> C5
    B1 -->|"PDF Export"| C6
    C7 -.->|"hosts"| Client

📊 Data Scale

Metric	Value
Total records	~433,000
Tables	27
Districts	30
Police stations	240
Cases	50,000
Accused persons	87,102
Victims	62,495
Complainants	67,433
Employees	2,000
Translation keys	425+ per language

📊 Data Sources

Source	Data
Synthetic Data Generator	27 relational tables
Census 2011	Literacy, urbanization, sex ratio, population
QuickML GLM-4.7-Flash	LLM-powered natural language insights

⚙️ Technology Stack

☁️ Zoho Catalyst Platform

Service	Purpose
Functions (Advanced I/O)	Entire backend — Python 3.13, 28 endpoints
Data Store (ZCQL)	All queries against 27 tables
Cache	2hr TTL on all endpoints
Connections	OAuth for QuickML LLM API
SmartBrowz	Server-side PDF report generation
Slate	Frontend static hosting
API Gateway	Clean `/api/*` routing, production-grade API management

🤖 Machine Learning & AI

Tech	Purpose
GLM-4.7-Flash (30B MoE)	LLM-powered natural language insights
Statistical Engine	Fallback when LLM times out

🌐 Backend

Tech	Version
Python	3.13
Flask	3.0
Zoho Catalyst SDK	Latest

🌐 Frontend

Tech	Version
React	18.3.1
Vite	5.3
Tailwind CSS	3.4
Recharts	3.9
Leaflet + React-Leaflet	1.9 / 4.2
react-force-graph-3d + Three.js	1.29 / Latest
React Router DOM	7.18
i18next + react-i18next	26.3 / 17.0

🔌 API Endpoints (29 Routes)

Category	Endpoint	Description
🏥 Health	`/api/health`	Health check
📊 Dashboard	`/api/dashboard/kpis`	Key performance indicators
	`/api/dashboard/by-district`	Cases grouped by district
	`/api/dashboard/by-category`	Cases grouped by crime category
	`/api/dashboard/by-status`	Cases grouped by status
🗺️ Map	`/api/map/crimes`	Crime locations with lat/lng
	`/api/map/hotspots`	District-level hotspots
📈 Trends	`/api/trends/monthly`	Monthly crime counts
	`/api/trends/hourly`	Hourly distribution
	`/api/trends/dayofweek`	Day-of-week distribution
	`/api/trends/gravity`	Gravity/offence breakdown
🕸️ Network	`/api/network/graph`	3D graph data (nodes + edges)
	`/api/network/offenders`	Ranked offender list
	`/api/network/person`	Person detail profile
	`/api/network/associations`	Person associations
	`/api/network/statistics`	Network statistics
🔮 Predictions	`/api/predictions/forecast`	30-day forecast
	`/api/predictions/risk`	District risk scores
	`/api/predictions/anomalies`	Anomaly detection
👮 Operational	`/api/operational/io-performance`	Investigation officer stats
	`/api/ops/ranks`	Police rank list for IO Performance filter
	`/api/operational/victim-profile`	Victim demographics
	`/api/operational/victim-gender-timeline`	Gender over time
	`/api/operational/accused-profile`	Accused age distribution
	`/api/operational/seasonal`	Seasonal crime patterns
	`/api/operational/socioeconomic`	Census 2011 correlation
🤖 AI	`/api/ai/insights`	LLM-generated crime analysis
	`/api/ai/risk-explanation`	AI risk narrative
📄 Reports	`/api/reports/pdf`	PDF report generation

🚀 Scalability & Innovation

Innovation	Description
🆓 Zero Infrastructure Cost	Runs entirely on Zoho Catalyst free tier
🚫 Zero GPU Dependency	Catalyst handles compute; BigQuery-style parallelism
🛡️ Statistical Fallback	LLM timeout → data-driven insights always available
🌐 Bilingual AI	All insights in English + Kannada
🔒 SQL Injection Protection	All date parameters validated
🎯 Filter-Aware Handlers	Every endpoint respects district/crime/date filters
⚡ 2hr Cache TTL	Fast repeated loads, reduced latency
📱 Responsive Design	Desktop, tablet, and mobile

🔮 Future Enhancements

📱 Mobile Application — Native iOS/Android app
🔔 Real-Time Alerts — Push notifications for anomalies
📊 Historical Trend Analytics — Multi-year deep dive
🧠 Vertex AI Integration — Advanced ML models
📡 IoT Sensor Support — CCTV, ANPR integration
⚡ Real-Time Streaming — Live data ingestion
🔐 User Authentication — Role-based access control
🌐 Multi-Language AI — Tamil, Telugu, Hindi support

🚀 Quick Start

Frontend

cd aarakshaka-ui
npm install
npm run dev

Runs on http://localhost:5173. Vite proxy routes /api/* to the Catalyst backend.

Backend

npm install -g zcatalyst-cli
catalyst login
catalyst project:use <your-project-id>
catalyst deploy --only functions:aarakshaka_ksp_2026_function

📂 Folder Structure

Datathon/
│
├── aarakshaka-ui/                  # React frontend
│   ├── src/
│   │   ├── components/             # UI components (ai, common, dashboard, map, network, operational, predictions, reports, trends)
│   │   ├── pages/                  # Page components (Dashboard, Map, Network, Offenders, Predictions, Login)
│   │   ├── hooks/                  # Custom React hooks
│   │   ├── api/                    # API client
│   │   ├── i18n/                   # i18next config
│   │   └── utils/                  # Formatters, colors, district names
│   └── public/
│       ├── locales/                # EN + KN translation files (425+ keys each)
│       └── karnataka-districts.geojson
│
├── functions/                      # Zoho Catalyst backend
│   └── aarakshaka_ksp_2026_function/
│       ├── handlers.py             # All 29 API handlers (~2100 lines)
│       ├── main.py                 # Route dispatch + CORS
│       └── requirements.txt        # Python dependencies
│
├── screenshots/                    # 19 PNG screenshots (login, dashboard, crime map, network, offenders, predictions, tablet views)
├── catalyst.json                   # Catalyst project config
├── catalyst-user-rules.json        # API Gateway routing rules
└── .gitignore

📄 License

Competition submission — Karnataka State Police Datathon 2026.

Built for Karnataka State Police | Powered by Zoho Catalyst | AI by GLM-4.7-Flash

Name		Name	Last commit message	Last commit date
aarakshaka-ui		aarakshaka-ui
functions/aarakshaka_ksp_2026_function		functions/aarakshaka_ksp_2026_function
screenshots		screenshots
.gitignore		.gitignore
README.md		README.md
catalyst-user-rules.json		catalyst-user-rules.json
catalyst.json		catalyst.json

Latest commit

History

Folders and files

Repository files navigation

🛡️ AARAKSHAKA — Crime Intelligence Platform

🎯 Problem Statement

🛡️ Why AARAKSHAKA

Real-world Impact

📸 Application Screenshots

Login

Dashboard — Overview

Dashboard — Analysis

Dashboard — Operations

Crime Map

Network Graph

Offenders

Predictions

Predictions

📱 Tablet View

✨ Key Features

🎛️ Dashboard & Analytics

🗺️ Interactive Crime Map

🕸️ Criminal Network Graph (3D)

🔮 Predictions & Risk Scoring

🤖 AI-Powered Insights

🌐 Bilingual Support

🏗️ Architecture

📊 Data Scale

📊 Data Sources

⚙️ Technology Stack

☁️ Zoho Catalyst Platform

🤖 Machine Learning & AI

🌐 Backend

🌐 Frontend

🔌 API Endpoints (29 Routes)

🚀 Scalability & Innovation

🔮 Future Enhancements

🚀 Quick Start

Frontend

Backend

📂 Folder Structure

📄 License

About

Topics

Resources

Stars

Watchers

Forks

Contributors

Languages