From facility telemetry to
operational intelligence
GreenNexa is an AI-powered sustainable facility intelligence platform for institutions and municipalities. Far beyond a passive visualization dashboard, GreenNexa connects operational data, multi-model AI/ML, forecasting, priority ranking, and conversational intelligence into a closed-loop operational decision system.
Engineered for modern institutions & civic administrations
SENSE → UNDERSTAND → DECIDE → ACT → VERIFY
How GreenNexa turns raw sensor telemetry into validated operational resolutions.
Telemetry Ingestion
Ingests readings every 30 seconds via authenticated HTTP IoT endpoints (X-Device-ID + X-API-Key) or isolated module-scoped synthetic simulators for active sensors.
Tri-Layer Analytics
Processes streams through hard threshold limits, statistical 3-sigma Z-score rolling baselines, Scikit-Learn Isolation Forest ML, and Holt-Winters 24h diurnal forecasting.
Priority & Prescriptions
When concurrent anomalies arise, the Priority Engine isolates the single most urgent issue and generates contextual recommendations with estimated kWh, m³, or cost savings.
Operator Resolution
Facility operators review clear priority cards across 4 dashboard styles, interact with the Gemini 2.5 Flash conversational assistant, and dispatch maintenance teams.
Closed-Loop Validation
Subsequent 30-second telemetry cycles stream continuously, mathematically confirming whether physical interventions restored normal baseline operations.
Why GreenNexa exists
Built to solve the operational gap between fragmented telemetry and decisive facility action.
An Operational Decision System (Beyond Static Dashboards)
Most facility software stops at visual charts and passive graphs. GreenNexa treats the dashboard as merely the user-facing presentation layer of an end-to-end operational decision system. It continuously bridges raw sensor telemetry, AI/ML analytics, predictive forecasting, priority evaluation, and prescriptive recommendations into guided human operational action.
Dual-Path Ingestion: Authenticated IoT & 30-Second Simulation
Facilities can stream physical hardware telemetry (such as ESP32 microcontrollers) over authenticated HTTPS using unique device IDs and secret keys. For evaluation, testing, or demonstrations, an isolated 30-second synthetic simulator models realistic diurnal consumption curves and injected faults without polluting production data.
Tri-Layer Anomaly Detection (Rules, Statistics & Machine Learning)
Operational failures rarely conform to simple static rules. GreenNexa combines deterministic threshold boundaries, statistical 3-sigma rolling Z-score deviations, and unsupervised Scikit-Learn Isolation Forest algorithms across 8 features to detect subtle equipment degradation, leaks, and demand spikes before catastrophic failures occur.
The Priority Engine: Eliminating Multi-Anomaly Alert Fatigue
When 3 or more anomalies occur concurrently across facility systems, operators risk alert fatigue. GreenNexa’s Priority Engine automatically evaluates risk severity, persistence, and facility impact to isolate and spotlight the single most critical issue, keeping operators focused on what matters most.
Conversational Intelligence: Gemini 2.5 Flash with 15 Domain Tools
Facility managers and engineers can query operational status in natural language using English, Hinglish, Odia, or Roman Odia. Powered by Google GenAI SDK and Gemini 2.5 Flash, GreenNexa AI leverages 15 specialized domain tools to inspect real-time metrics, anomalies, forecasts, and What-If scenarios with safe offline fallback.
Municipal Civic Monitoring & Government-Only Association
In addition to standard facility modules, municipal administrations gain dedicated civic infrastructure monitoring across Wards—including Street Lighting, Roads & Infrastructure, Parks & Playgrounds, and Drainage & Sewage. GreenNexa enforces strict governance hierarchy: only verified Government organisations can associate with a municipality.
Verified technology architecture
Built with modern, open standards verified directly in the implemented production codebase.
Next.js 14 App Router
React 18, TypeScript, custom Vanilla CSS design tokens, responsive layouts, and React Context-based state management.
FastAPI & Python 3.11+
High-performance asynchronous REST API, Pydantic v2 schema validation, SQLAlchemy 2.0 ORM, and JWT bearer authentication.
SQLite (13 Normalized Tables)
Strict multi-tenant schema isolation, historical immutable telemetry, physical file-size monitoring, and logical org attribution.
Statsmodels & Scikit-Learn
Holt-Winters Exponential Smoothing for 24h diurnal forecasting, statistical 3-sigma Z-scores, and 8-feature Isolation Forest ML.
Gemini 2.5 Flash & GenAI SDK
Natural language interface with 15 live domain tool bindings, multilingual dialogue, What-If simulation, and graceful offline fallback.
Authenticated Ingestion
Dedicated HTTP endpoints (/api/v1/iot/sensor-data) with X-Device-ID and X-API-Key verification, rate-limiting, and sensor catalog validation.