CrimeCopilot AI - Decentered Intelligence Dashboard for Karnataka State Police

CrimeCopilot AI is a premium, high-performance, AI-driven crime analysis console built for the Karnataka State Police (KSP) Datathon 2026. It integrates relational database filters (SQLite), graph network traversals (Neo4j), TF-IDF semantic case matching, offender risk timelines, and bilingual voice recognition (English/Kannada) into a unified "Stadium Noir" command dashboard.

---

1. Prototype Brief (Project Overview)

A. Problem Statement Addressed

Modern police intelligence relies on fragmented silos (Excel, PDF files, isolated databases). When investigating crime syndicates, officers face:

1. Hidden Connections: Difficulty identifying links between suspects, shared vehicles, and bank accounts.
2. Information Overload: Scanning thousands of written FIR descriptions to find patterns (modus operandi).
3. Language Barriers: Inability to easily query or record voice notes in native languages like Kannada.
4. Explainability Deficit: Black-box AI systems that make claims without verifiable sources.

B. Key Features & Functionality

• Conversational AI Console (SSE Streaming): Natural language routing with real-time progress logs.
• Interactive Cytoscape.js Network Graph: Visualizes 2-hop criminal syndicates, communications, and financial channels.
• Geospatial Hotspot Leaflet Map: Renders regional crime hotspots on custom dark-theme GIS tiles.
• TF-IDF Semantic Similar Case Search: Matches crime descriptions using local Cosine Similarity vectors.
• Offender Risk Profiling: Renders suspect cards with gang banners, timelines, and SVG risk gauges.
• Bilingual Voice Input (Speech-to-Text): Browser SpeechRecognition supporting Kannada and English transcription and automated query translation.
• Official PDF Report Export: Custom @media print rules styling printable case sheets with 1 click.
• Grounded Explainability (Hallucination Guard): Verifies all AI response tokens against evidence trails.

C. Technology Stack

• Frontend: Next.js 16 (App Router + TS), Tailwind CSS (Stadium Noir Theme), Cytoscape.js, React-Leaflet.
• Backend: FastAPI (Python 3.13), LangGraph (Agent State Orchestration), SQLAlchemy ORM.
• Database: SQLite (Relational), Neo4j AuraDB (Graph database + Cypher engine).
• LLM Engine: Gemini 1.5 Flash (via google-generativeai).

D. Proposed Impact

Reduces time-to-insight for field investigators from hours to seconds. Promotes native language usage in administrative interfaces, and ensures 100% auditability through verifiable evidence trails.

---

2. Setup & Local Execution

A. Backend Setup

1. Navigate to the project root:

   cd crimecopilot-ai

2. Create and activate a Python virtual environment:

   python -m venv .venv
   # On Windows:
   .venv\Scripts\activate
   # On macOS/Linux:
   source .venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Copy the environment template and add your API Keys:

   copy .env.example .env

   Edit .env to supply:

   • GEMINI_API_KEY (Get from Google AI Studio)
   • NEO4J_URI, NEO4J_USER, NEO4J_PASSWORD (Optional: Neo4j Aura DB credentials)

5. Seed the relational and graph databases:

   python generate_data.py
   python validate_data.py
   python load_sql.py
   # Optional Neo4j graph seed:
   python load_neo4j.py

6. Start the FastAPI development server:

   python -m uvicorn main:app --host 127.0.0.1 --port 8000

B. Frontend Setup

1. Navigate to the frontend directory:

   cd frontend

2. Install npm dependencies:

   npm install

3. Start the Next.js development server with Turbopack:

   npm run dev

4. Access the command console at: http://localhost:3000

---

3. Zoho Catalyst Deployment Guide

To deploy the solution to the Zoho Catalyst platform, follow these steps:

A. Deploy Backend on AppSail

Catalyst AppSail hosts full-stack python apps.

1. Install Catalyst CLI globally:

   npm install -g zcatalyst-cli

2. Login to your Zoho account:

   catalyst login

3. Initialize the AppSail service in the root directory:

   catalyst init
   # Select 'AppSail' and associate it with your Catalyst project.

4. Configure app-config.json inside the root folder to point to main:app using uvicorn:

   {
     "build": {
       "command": "pip install -r requirements.txt"
     },
     "run": {
       "command": "uvicorn main:app --host 0.0.0.0 --port $PORT"
     }
   }

5. Deploy to Catalyst AppSail:

   catalyst deploy

B. Deploy Frontend on Catalyst Slate

1. Navigate to frontend/ and configure Next.js for static export by editing next.config.ts:

   const nextConfig = {
     output: 'export'
   };
   export default nextConfig;

2. Build the static distribution:

   npm run build
   # This generates an 'out/' directory.

3. Initialize Catalyst Slate hosting inside the frontend/ folder:

   catalyst init
   # Select 'Hosting' or 'Slate', point the public source directory to 'out'.

4. Deploy the frontend:

   catalyst deploy --only hosting