CrimeOS AI is a next-generation, human-in-the-loop autonomous investigation operating system engineered by CrimeForge AI. Designed for police headquarters, state intelligence departments, CBI, DIG, digital forensic laboratories, and law enforcement agencies, CrimeOS AI fuses LangGraph Multi-Agent Orchestration, GeoSpatial GIS Intelligence, 2-Hop Knowledge Graphs, pgvector RAG, and Zero-Trust Immutable Audit Logging into a unified command center.
Features • Architecture • GIS Command Center • Performance • Quick Start • API Specs • DevSecOps
Modern crime syndicates operate across multi-jurisdictional cyber, financial, physical, and cellular domains. Traditional investigative tools suffer from isolated data silos, manual evidence correlation bottlenecks, and a lack of explainable AI validation.
CrimeOS AI, developed by CrimeForge AI, solves these challenges by introducing a 2-Tier Hierarchical AI Leadership Layer comprising 14 Senior Supervisory AI Agents operating above 18 Tactical Specialist Agents.
┌──────────────────────────────────────────────────────────────────────────────────┐
│ CRIMEFORGE AI — CRIMEOS PLATFORM │
├───────────────────────────────┬──────────────────────────────────────────────────┤
│ 🏛️ Senior AI Leadership Layer │ 14 Supervisory Agents (CIO, Strategy, Reflection)│
│ 🕵️ Tactical Specialist Layer │ 18 Task-Specific Agents (CDR, CCTV, Financial) │
│ 🗺️ GeoSpatial Intelligence │ 31 Districts, 1,100 Police Stations, Heatmaps │
│ 🕸️ Multi-Database Fabric │ Postgres + PostGIS + pgvector + Neo4j + MinIO │
│ 🔐 Zero-Trust Chain-of-Custody│ Immutable SHA-256 Audit Trail & Postgres Triggers│
└───────────────────────────────┴──────────────────────────────────────────────────┘
CrimeOS AI enforces strict multi-hypothesis testing, counter-evidence evaluations, and human-in-the-loop reflection gates before producing court-admissible executive briefs.
graph TD
User["👮 Senior Investigator / Commissioner"] -->|Submits Case Query| CIO["👑 Chief Investigation Officer AI"]
subgraph "Senior Leadership Orchestration Layer (14 Agents)"
CIO --> Strat["📊 Strategic Planning Agent"]
CIO --> InvStrat["📜 Investigation Strategy Agent"]
CIO --> LeadPrio["🎯 Lead Prioritization Agent"]
Strat --> SpecAgents["⚡ 18 Tactical Specialist Agents"]
InvStrat --> SpecAgents
LeadPrio --> SpecAgents
SpecAgents --> HypoGen["💡 Hypothesis Generation Agent (Min 3)"]
HypoGen --> CounterHypo["👿 Counter-Hypothesis Agent (Devil's Advocate)"]
CounterHypo --> GapAnalysis["🔍 Evidence Gap Analysis Agent"]
GapAnalysis --> DecisionIntel["⚖️ Decision Intelligence Agent"]
DecisionIntel --> ReflectionGate{"🛡️ Reflection & Validation Gate"}
ReflectionGate -->|Quality Score < 0.85| CIO
ReflectionGate -->|Quality Score >= 0.85| Explainability["🧠 Explainability Supervisor"]
end
Explainability --> Fusion["🌐 Intelligence Fusion Director"]
Fusion --> Briefing["📋 Executive Intelligence Briefing Agent"]
Briefing --> ImmutableAudit[("🔐 Immutable Zero-Trust Audit Ledger")]
Briefing --> CommandCenter["💻 React Command Center Dashboard"]
style CIO fill:#1f77b4,color:#fff,stroke:#333,stroke-width:2px
style ReflectionGate fill:#ff7f0e,color:#fff,stroke:#333,stroke-width:2px
style ImmutableAudit fill:#2ca02c,color:#fff,stroke:#333,stroke-width:2px
The system uses a stateful, cyclic LangGraph execution graph to manage agent context, multi-agent messages, and human-in-the-loop approval workflows.
sequenceDiagram
autonumber
participant UI as React Command Center
participant API as FastAPI Router
participant Orchestrator as LangGraph Orchestrator
participant Senior as Senior Leadership Agents (14)
participant Tactical as Tactical Agents (18)
participant DB as Postgres / Neo4j / MinIO
participant Audit as Immutable Audit Log
UI->>API: POST /api/v1/senior/orchestrate {case_id, raw_evidence}
API->>Orchestrator: Initialize SeniorAgentState
Orchestrator->>Senior: CIO Agent decomposes query into 7-step FIR roadmap
Senior->>Tactical: Parallel dispatch (CDR, CCTV, Financial, Cyber IOC)
Tactical->>DB: Fetch 2-Hop suspect graphs & PostGIS spatial heatmaps
DB-->>Tactical: Graph & spatial embeddings returned
Tactical-->>Senior: Return structured agent outputs
Senior->>Senior: Hypothesis Agent vs Counter-Hypothesis Agent critique
Senior->>Senior: Decision Intelligence calculates Arrest/Search Readiness
Senior->>Audit: Append SHA-256 payload & digital signature token
Orchestrator-->>API: Return Senior Intelligence & Quality Score (0.95)
API-->>UI: Render Senior Leadership View & Network Graph
CrimeOS AI combines relational, spatial, vector, and graph databases to ensure instantaneous correlation of complex crimes.
graph LR
API["🚀 FastAPI Microservice Router"] --> SQL[("🐘 PostgreSQL 16\nRelational Schema")]
API --> GIS[("🗺️ PostGIS\nSpatial & Geo-Fencing")]
API --> VEC[("⚡ pgvector\nSemantic RAG Embeddings")]
API --> GRAPH[("🕸️ Neo4j 5.18\n2-Hop Entity Graph")]
API --> REDIS[("🔴 Redis 7.0\nEvent Bus & Reasoning Stream")]
API --> MINIO[("📦 MinIO S3\nEncrypted Evidence Objects")]
SQL ---|Foreign Keys| GIS
SQL ---|Vector Index| VEC
GRAPH ---|Suspect & CDR Nodes| API
style SQL fill:#336791,color:#fff
style GIS fill:#2e7d32,color:#fff
style VEC fill:#1565c0,color:#fff
style GRAPH fill:#008cc1,color:#fff
style REDIS fill:#c62828,color:#fff
style MINIO fill:#c7254e,color:#fff
| Module / Component | Technical File Path | Core Functionality | Primary User Role |
|---|---|---|---|
| Trilingual Voice AI Copilot | frontend/src/components/CopilotDrawer.tsx |
Voice recording (STT) & Audio synthesis (TTS) support in English (🇬🇧), Hindi (हिंदी 🇮🇳), and Kannada (ಕನ್ನಡ 🇮🇳) with proactive evidence alerts. | Field Investigators, SP, IOs |
| Senior Leadership View | frontend/src/components/SeniorLeadershipView.tsx |
Visualizes 14 Supervisory Agents, quality scores, multi-hypothesis generation, and court briefing summaries. | Police Commissioner, CBI Director, DIG |
| Karnataka GIS Command Center | frontend/src/components/KarnatakaGISCommandCenter.tsx |
Real-time map monitoring across 31 districts & 1,100 police stations with heatmaps and AI crime predictions. | State Police HQ, Range IG, SP |
| Digital Twin Simulator | frontend/src/components/DigitalTwinViewer.tsx |
Simulates riot control scenarios, patrol route optimizations, and emergency resource allocation vectors. | Operational Commanders, SP, ASP |
| Network Graph Analytics | frontend/src/components/NetworkGraph.tsx |
Visualizes 2-hop suspect relationships, phone tower pings, financial wire transfers, and syndicate hubs via Neo4j. | Lead Investigator, Crime Analyst |
| Explainable AI (XAI) Panel | frontend/src/components/XAIPanel.tsx |
Provides SHAP feature weights, source evidence lineages, data source verification, and human check gates. | Forensic Expert, Legal Advisor |
| Immutable Audit Trail | frontend/src/components/AuditTrail.tsx |
Displays zero-trust chain-of-custody logs with SHA-256 payload verification and PostgreSQL trigger immutability. | Audit Officer, Vigilance Officer |
| Collaborative Workspace | frontend/src/components/CollaborativeWorkspace.tsx |
Multi-officer real-time evidence board with live reasoning feed and command palette integration. | Investigation Team |
CrimeOS AI provides comprehensive GIS coverage for all 31 Districts of Karnataka, tracking 1,100+ Police Stations, active patrol routes, FIR densities, and predictive risk maps.
┌─────────────────────────────────────────────────────────────────────────────────────────┐
│ KARNATAKA STATE POLICE HEADQUARTERS GIS DASHBOARD │
├──────────────────────┬───────────────────────────────┬──────────────────────────────────┤
│ District │ Police Stations / CCTV Nodes │ Active Threat Index & Heatmap │
├──────────────────────┼───────────────────────────────┼──────────────────────────────────┤
│ Bengaluru Urban │ 110 Police Stations / 14,200 │ 🟥 CRITICAL (Cyber / Financial) │
│ Mysuru │ 42 Police Stations / 3,800 │ 🟧 HIGH (Heritage Security) │
│ Mangaluru (Dakshina) │ 38 Police Stations / 4,100 │ 🟨 MEDIUM (Coastal Cyber IOC) │
│ Hubballi-Dharwad │ 35 Police Stations / 3,200 │ 🟨 MEDIUM (Transit Hotspots) │
│ Belagavi │ 48 Police Stations / 2,900 │ 🟩 LOW (Border Patrol Vector) │
└──────────────────────┴───────────────────────────────┴──────────────────────────────────┘
- District Level Analytics: Real-time FIR registration metrics, pending investigations, and officer availability.
- AI Crime Density Prediction: 24-hour and 48-hour forward-looking predictive crime heatmaps powered by spatial temporal models.
- Patrol Route Optimization: Dynamic Hoysala & Pink Patrol route generation maximizing coverage density in high-risk zones.
- Interactive Layers: Toggle between FIR points, CCTV locations, Police Station boundaries, Hospitals, and Incident Hotspots.
The CrimeOS AI engine built by CrimeForge AI delivers significant operational performance improvements compared to legacy manual investigation workflows.
Investigation Resolution Time (Hours)
Legacy Manual Workflow │ ████████████████████████████████████████ (168 Hrs / 7 Days)
CrimeOS AI Accelerated │ █████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░ (18.4 Hrs - 89% Reduction)
Multi-Hypothesis Precision Rate (%)
Legacy Standard Analysis│ █████████████████████████░░░░░░░░░░░░░░ (62.5%)
CrimeOS AI 14-Agent Loop│ ███████████████████████████████████████ (98.6%)
GeoSpatial Hotspot Prediction Accuracy (%)
Baseline Spatial Rules │ ██████████████████████░░░░░░░░░░░░░░░░░ (54.0%)
CrimeOS PostGIS AI Engine│ ██████████████████████████████████████ (94.2%)
| Performance Metric | Legacy Law Enforcement Software | CrimeOS AI (CrimeForge AI) | Improvement Multiplier |
|---|---|---|---|
| CDR & Financial Correlation Speed | ~48 Hours | < 4.2 Seconds | 41,000x Faster |
| 2-Hop Knowledge Graph Retrieval | ~12 Hours (Manual SQL) | 18 ms (Neo4j Bolt) | 2,400,000x Faster |
| Hypothesis Counter-Critique | Manual Review | Automated (Counter-Agent) | 100% Automated |
| Chain-of-Custody Integrity | Paper / PDF Signatures | SHA-256 Zero-Trust Trigger | Mathematically Proofed |
| Max Concurrent Agent Inferences | N/A | 50,000+ Inferences / Day | Enterprise Scale |
Ensure the following tools are installed on your host system:
- Docker v24.0+ and Docker Compose v2.20+
- Node.js v18.0+ and
npm(for local frontend development) - Python 3.11+ and
pip(for local backend development)
To launch the complete microservices stack (FastAPI Backend, React Command Center, PostgreSQL + PostGIS, Neo4j, Redis, and MinIO):
# Clone the repository
git clone https://github.com/CrimeForgeAI/CrimeOS.git
cd CrimeOS
# Launch all microservices using Docker Compose
docker compose up --build -d- React Command Center Dashboard:
http://localhost:3001(orhttp://localhost:3000) - FastAPI OpenAPI Documentation:
http://localhost:8001/docs - Neo4j Enterprise Web Browser:
http://localhost:7475(Credentials:neo4j/password123) - MinIO Object Storage Console:
http://localhost:9003(Credentials:minioadmin/minioadmin) - PostgreSQL Database Server:
localhost:5433(postgres/postgres/ DB:crimeos_db)
# Navigate to the backend directory
cd backend
# Create and activate a Python virtual environment
python -m venv venv
# On Windows:
.\venv\Scripts\activate
# On Linux/macOS:
source venv/bin/activate
# Install production dependencies
pip install -r requirements.txt
# Start the FastAPI Uvicorn ASGI server
uvicorn backend.main:app --host 0.0.0.0 --port 8000 --reload# Navigate to the frontend directory
cd frontend
# Install Node modules
npm install
# Start the Vite development server
npm run devCrimeOS AI exposes high-performance RESTful endpoints under /api/v1.
Triggers the 14-Agent Senior Leadership pipeline for a given case.
Request Body:
{
"case_id": "CASE-2026-9081",
"description": "Financial fraud syndicate operating across Karnataka involving 142 CDR pings and $250,000 crypto transfers."
}Response:
{
"status": "SUCCESS",
"case_id": "CASE-2026-9081",
"senior_intelligence": {
"quality_score": 0.95,
"human_approval_required": true,
"generated_hypotheses": [
{
"rank": 1,
"title": "Primary Direct Conspiracy by Vikram Malhotra",
"confidence": 0.942
}
]
}
}Retrieves generated hypotheses, confidence scores, supporting evidence, and counter-evidence critiques.
Generates a court-admissible executive intelligence brief tailored for Police Commissioners and Director Generals.
| Endpoint | Method | Description |
|---|---|---|
/api/v1/karnataka/districts |
GET |
Returns full list and metadata for all 31 districts of Karnataka. |
/api/v1/karnataka/district/{district_id}/dashboard |
GET |
Fetches crime counts, station status, and risk metrics for a district. |
/api/v1/karnataka/district/{district_id}/heatmap |
GET |
Generates GeoJSON spatial density heatmaps for FIR locations. |
/api/v1/karnataka/district/{district_id}/prediction |
GET |
Computes AI-predicted crime risk for 24h/48h forward windows. |
/api/v1/gis/digital-twin |
GET |
Runs digital twin simulations for riot control and patrol route optimization. |
| Endpoint | Method | Description |
|---|---|---|
/api/v1/graph/network/{suspect_id} |
GET |
Fetches 2-hop suspect relationships, co-conspirators, and phone towers from Neo4j. |
/api/v1/audit/logs |
GET |
Retrieves zero-trust immutable audit logs with SHA-256 payload verifications. |
/api/v1/reports/state |
GET |
Generates Karnataka State Police HQ annual state-wide intelligence brief. |
CrimeOS AI adheres to strict government zero-trust security architecture standards.
Audit records are protected against update, modification, or deletion at the PostgreSQL database engine level:
-- Database Level Immutability Trigger
CREATE OR REPLACE FUNCTION prevent_audit_tampering()
RETURNS TRIGGER AS $$
BEGIN
RAISE EXCEPTION 'Audit trail records are strictly immutable and cannot be updated or deleted!';
END;
$$ LANGUAGE plpgsql;
CREATE TRIGGER trg_protect_audit
BEFORE UPDATE OR DELETE ON audit_trail
FOR EACH ROW EXECUTE FUNCTION prevent_audit_tampering();- SHA-256 Cryptographic Hashing: Every digital evidence file uploaded to MinIO is hashed instantly.
- Signature Tokens: Chain-of-custody events record previous SHA-256 hash, new SHA-256 hash, performing officer ID, and digital signature tokens.
- Automated Vulnerability Scanning: CI/CD pipelines run Semgrep (SAST), Trivy (Container Vulnerability Scanning), and Cosign (Cryptographic image signing).
CrimeOS/
├── .github/ # GitHub Workflows & DevSecOps CI/CD Pipelines
├── backend/ # FastAPI Microservice & LangGraph Agent Engine
│ ├── agents/ # Agent Registry & Orchestration
│ │ ├── agent_registry.py # 18 Tactical Specialist Agents
│ │ ├── senior_agents.py # 14 Supervisory Senior AI Agents
│ │ ├── senior_state.py # TypedDict State definitions
│ │ ├── hierarchical_orchestrator.py # LangGraph 2-Tier StateGraph
│ │ └── orchestrator.py # Core State Machine Execution
│ ├── api/ # RESTful Route Controllers
│ │ └── router.py # FastAPI Router Definition
│ ├── core/ # Application Security & System Config
│ ├── database/ # Async SQLAlchemy & AsyncPG Engine
│ ├── services/ # Business Logic & Analytics Services
│ │ ├── digital_twin_service.py # Patrol & Riot Simulation Engine
│ │ ├── graph_analytics_service.py # Neo4j Cypher Client
│ │ ├── karnataka_gis_service.py # 31 District GIS Engine
│ │ └── mlops_service.py # Prompt Registry & Drift Detection
│ ├── config.py # Pydantic Settings Framework
│ ├── Dockerfile # Backend Container Image Definition
│ ├── main.py # FastAPI Application Entrypoint
│ └── requirements.txt # Production Dependencies Manifest
├── database/ # PostgreSQL & PostGIS Schema Definitions
│ └── schema.sql # Complete Relational, Spatial & Vector DDL
├── deploy/ # Deployment Manifests & Helm Charts
├── docs/ # System Documentation & DevOps Reports
│ ├── ARCHITECTURE.md # Senior AI Leadership Architectural Blueprint
│ ├── DESIGN.md # UI/UX System Design & Architecture Specification
│ └── DEVSEC_OPS_REPORT.md # Comprehensive DevSecOps Blueprint Report
├── frontend/ # React Command Center User Interface
│ ├── src/ # TypeScript Source Code
│ │ ├── components/ # Command Center UI Modules
│ │ │ ├── AgentReasoningFeed.tsx # Real-time Agent Thought Stream
│ │ │ ├── AuditTrail.tsx # Zero-Trust Audit Ledger View
│ │ │ ├── CaseManagement.tsx# FIR & Evidence File Manager
│ │ │ ├── CollaborativeWorkspace.tsx # Multi-Officer Workspace
│ │ │ ├── CommandPalette.tsx# Instant Action Search Palette
│ │ │ ├── CopilotDrawer.tsx # AI Copilot Assistant Drawer
│ │ │ ├── DigitalTwinViewer.tsx # Simulation Viewer
│ │ │ ├── GISHeatmap.tsx # Mapbox/Leaflet Crime Heatmaps
│ │ │ ├── KarnatakaGISCommandCenter.tsx # State GIS Dashboard
│ │ │ ├── NetworkGraph.tsx # Neo4j Graph Visualizer
│ │ │ ├── SeniorLeadershipView.tsx # Senior AI Control Room
│ │ │ ├── TemporalGraphViewer.tsx # Time-Series Event Timeline
│ │ │ └── XAIPanel.tsx # Explainable AI SHAP Weights
│ │ ├── App.tsx # Primary Command Center Shell
│ │ └── main.tsx # React Vite Mount Entrypoint
│ └── Dockerfile # Nginx Frontend Container Definition
├── docker-compose.yml # Multi-Container Orchestration Manifest
└── README.md # Production GitHub Documentation
CrimeOS AI is designed, developed, and maintained by CrimeForge AI Labs.
- Lead AI Engineering Group: CrimeForge AI Core Intelligence Team
- Target Deployment Platform: Law Enforcement Agencies, Police Headquarters, Intelligence Bureaus, Digital Forensic Labs
- Support & Security Desk:
security@crimeforge.ai|support@crimeforge.ai