Sherlock.ai - Investigative Intelligence System

A citation-grounded investigative system that converts raw evidence into actionable intelligence with complete source attribution. Built for detectives, analysts, and prosecutors who need court-ready evidence bundles.

🎯 Core Features

Evidence Processing

• Multi-source Ingestion: CDR, ALPR, POS, CAD, Sensor, Narrative reports
• SHA-256 Hashing: Every artifact tracked with cryptographic verification
• Citation Tracking: Sentence-level source attribution for all extracted data

Intelligence Generation

• Entity Extraction: Automatic extraction of phones, plates, names, devices, locations
• Event Timeline: Chronological ordering with clock drift correction and contradiction detection
• Relationship Linking:
  • Co-presence detection (same location, same time)
  • Co-travel patterns (consecutive tower connections)
  • Rendezvous identification (repeated meetings)
  • Fraud chains (suspicious transaction sequences)

Visual Analytics

• Crime Scene Mapping: Interactive map with movement routes and polylines
• Entity Graph: Network visualization with typed relationships
• Timeline Scrubber: Temporal analysis with citation popovers
• Lead Scoring: AI-powered ranking of investigative leads

Evidence Packs

• Court-ready ZIP bundles with:
  • Original artifacts
  • SHA-256 manifest
  • Timeline JSON
  • Graph slice
  • One-page summary
  • Verification hashes

🏗️ Architecture

Backend: FastAPI + SQLite + Neo4j + Weaviate
Frontend: React + TypeScript + Tailwind CSS (Glassmorphic UI)
AI Stack: LlamaIndex + OpenAI/Gemini/Friendli AI

┌─────────────┐      ┌──────────────┐      ┌─────────────┐
│   Artifacts │─────▶│   Services   │─────▶│   Output    │
│   (Raw)     │      │  (Process)   │      │  (Intel)    │
└─────────────┘      └──────────────┘      └─────────────┘
     │                      │                      │
     ├─ CDR                ├─ Ingest              ├─ Timeline
     ├─ ALPR               ├─ Extract Entities    ├─ Graph
     ├─ POS                ├─ Eventize            ├─ Leads
     ├─ CAD                ├─ Link (Rules)        └─ Evidence Pack
     ├─ Sensor             └─ Score Leads
     └─ Narratives

🚀 Quick Start

Prerequisites

• Python 3.9+
• Node.js 18+
• Neo4j 5.x
• Weaviate 1.x (or Docker)

1. Clone Repository

git clone https://github.com/yourusername/sherlock-ai.git
cd sherlock-ai

2. Setup Backend

cd backend
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt

# Create .env file
cp .env.example .env
# Edit .env with your API keys and database credentials

3. Start Databases

# Using Docker Compose (recommended)
docker-compose up -d neo4j weaviate

# Or start services manually:
# - Neo4j: http://localhost:7474
# - Weaviate: http://localhost:8080

4. Initialize & Run Backend

# Run demo with synthetic data
python demo/run_demo.py

# Or start API server
uvicorn app.main:app --reload --port 8000

5. Setup Frontend

cd ../frontend
npm install
npm run dev

Visit http://localhost:5173 to see the dashboard!

📊 Demo Walkthrough

The demo script generates a complete investigation with:

• 2 suspects with phone records (10 CDR entries)
• 2 vehicles tracked via ALPR (10 location hits)
• POS theft sequence (DoorOpen → Void transaction)
• 3 officer narratives with embedded contradictions
• CAD dispatch events

Expected Results:

✓ 25+ artifacts ingested
✓ 20+ entities extracted
✓ 30+ events created
✓ 10-15 relationships found
✓ 2 contradictions detected
✓ 3+ high-confidence leads generated
✓ Evidence pack with SHA-256 verification

Run Demo:

cd backend
python demo/run_demo.py

🔒 Security & Compliance

CJIS-Aligned Patterns

• ✅ Least privilege access control
• ✅ Complete audit logging
• ✅ SHA-256 verification on all artifacts
• ✅ PII redaction with role-based reveal
• ✅ Immutable evidence packs
• ✅ No public internet access (configurable)

Citation Requirements

• Every entity must have ≥1 citation
• Every event must have ≥1 citation
• Every edge must have ≥2 citations
• Citations include: artifact_id, source_type, text span

📖 API Documentation

Artifact Ingestion

POST /api/artifacts/ingest-json
{
  "artifact_type": "CDR",
  "raw_payload": "{...}",
  "case_id": "CASE_001"
}

Case Analysis

POST /api/cases/{case_id}/analyze
# Returns: edge counts and top leads

Timeline & Routes

GET /api/cases/{case_id}/timeline
GET /api/cases/{case_id}/routes  # GeoJSON

Evidence Pack

POST /api/cases/{case_id}/export
GET /api/packs/{pack_id}/download
GET /api/packs/{pack_id}/verify

🎨 UI Components

Glassmorphic Design System

• GlassCard: Frosted glass containers with backdrop blur
• GlassButton: Interactive buttons with hover effects
• GlassPanel: Titled panels for structured content
• GlassModal: Overlay modals with blur

Color Palette

• Background: #0a0e27 (dark navy)
• Accent Cyan: #00d9ff
• Accent Purple: #a855f7
• Glass: rgba(255, 255, 255, 0.1) with backdrop-blur

🧪 Testing

# Backend tests
cd backend
pytest

# Frontend tests
cd frontend
npm test

# Integration test (full pipeline)
python backend/demo/run_demo.py

📈 Metrics & Validation

The system tracks:

• Citation Coverage: % of entities/events with citations
• Edge Quality: Average confidence scores
• Contradiction Rate: Temporal impossibilities detected
• Hash Verification: 100% pass rate on evidence packs
• Processing Time: Ingest → Lead generation cycle

🤝 Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Development Guidelines

• Follow PEP 8 for Python
• Use ESLint/Prettier for TypeScript
• Write tests for new features
• Document API changes
• Maintain citation requirements

📝 License

MIT License - see LICENSE for details

🙏 Acknowledgments

• FastAPI for the excellent web framework
• Neo4j for graph database capabilities
• React-Leaflet for mapping
• Cytoscape.js for graph visualization
• LlamaIndex for document processing

📞 Support

• Documentation: docs/
• Issues: GitHub Issues
• Email: support@sherlock-ai.example

---

Built with ❤️ for law enforcement and investigative professionals