How abhayj04/mplads-sentinel-ai works
MPLADS Sentinel AI is an AI-assisted audit and decision-support platform: project records are screened for risk signals, consolidated into explainable scores, prioritized for human investigation, and acted on through evidence review and audit logging. The README also documents dashboards, agency profiling, GIS risk views, data-quality checks, and a copilot. Sampled source confirms startup seeding, scoring, authentication, dashboard queries, and investigation actions; several documented feature routers are present in the tree but their internal wiring was not sampled.
Main entry points: main.py FastAPI application · generator.py Dataset screening · database.py Application database · investigations.py Investigation queue
Components 25
Access and API
- Web interface
frontend - FastAPI application
backend/app/main.py - Authentication
backend/app/api/auth.py - Project records
backend/app/api/projects.py - Administration
backend/app/api/admin.py
Risk Intelligence
- Agency profiles
backend/app/api/agencies.py - Data quality
backend/app/api/data_quality.py - Dataset screening
backend/app/seed/generator.py - Demo scenarios
backend/app/seed/scenarios.py - Cost anomaly detection
backend/app/ml/cost_anomaly.py - Payment mismatch detection
backend/app/ml/payment_anomaly.py - Timeline delay detection
backend/app/ml/delay_detector.py - Duplicate detection
backend/app/ml/duplicate_detector.py - Agency risk profiling
backend/app/ml/agency_profiler.py - Composite risk and explanations
backend/app/ml/risk_engine.py
Review and Oversight
- Dashboard analytics
backend/app/api/dashboard.py - Risk alerts
backend/app/api/alerts.py - Investigation queue
backend/app/api/investigations.py - Evidence locker
backend/app/api/evidence.py - GIS risk views
backend/app/api/gis.py - Decision copilot
backend/app/api/copilot.py - Audit trail
backend/app/core/audit.py
Platform State
- Application database
backend/app/database.py - Domain records
backend/app/models.py
Other components
- Scheme officers
Connections 27
- Scheme officers → Web interface: use
- Web interface → FastAPI application: API requests
- FastAPI application → Authentication: registers
- FastAPI application → Project records: registers
- FastAPI application → Dashboard analytics: registers
- FastAPI application → Risk alerts: registers
- FastAPI application → Investigation queue: registers
- FastAPI application → Evidence locker: registers
- FastAPI application → GIS risk views: registers
- FastAPI application → Agency profiles: registers
- FastAPI application → Data quality: registers
- FastAPI application → Decision copilot: registers
- FastAPI application → Administration: registers
- FastAPI application → Dataset screening: seeds on startup
- Dataset screening → Demo scenarios: builds examples
- Dataset screening → Cost anomaly detection: detects costs
- Dataset screening → Payment mismatch detection: detects mismatches
- Dataset screening → Timeline delay detection: detects delays
- Dataset screening → Duplicate detection: finds duplicates
- Dataset screening → Composite risk and explanations: calculates scores
- Dataset screening → Agency risk profiling: profiles agencies
- Investigation queue → Audit trail: logs actions
- Authentication → Application database: reads users
- Investigation queue → Application database: reads and writes
- Dashboard analytics → Application database: queries metrics
- Dataset screening → Application database: writes seeded records
- Application database → Domain records: maps records