Reduce industrial energy waste by up to 30% through real-time AI monitoring, autonomous agents, edge computing, carbon compliance, and smart demand response.
Problem • Solution • Features • Architecture • Quick Start • Pilot • Support
Nederlands • English
Manufacturing companies waste 15-30% of energy through machines running idle, suboptimal schedules, and zero visibility into consumption patterns.
Most industrial sites still rely on monthly utility bills to identify waste. By the time the data arrives, the money is already spent and the CO2 is already emitted. With increasing EU regulations on carbon reporting and ISO 50001 requirements, companies need more than just monitoring — they need a complete compliance platform.
AI-IDLE connects directly to industrial equipment through Shelly, Tasmota, Tuya, and MQTT devices, collects live power data every 10 seconds, and uses autonomous AI agents, edge computing, demand response, and carbon compliance tracking to detect anomalies, optimize costs, reduce emissions, and ensure regulatory compliance.
| Traditional Monitoring | AI-IDLE |
|---|---|
| Monthly energy bills | Real-time data every 10 seconds |
| Manual walkthroughs | 4 autonomous AI agents monitoring 24/7 |
| Spreadsheet reports | Automated PDF/Excel reports with AI insights |
| Reactive maintenance | Predictive maintenance + self-healing |
| No cost optimization | ENTSO-E Day-Ahead pricing + demand response |
| Cloud-only processing | Edge computing with ONNX Runtime inference |
| No emission tracking | Full Scope 1/2/3 carbon compliance with ISO 50001 |
- Live dashboard with power, cost, and trend data via WebSocket (Socket.IO)
- Per-machine sparklines for instant visual trend analysis
- TV casting on factory floor screens via Chromecast/DLNA
- Mobile-first PWA with offline support and push notifications
4 autonomous agents that monitor, analyze, and act on energy data around the clock. No external API keys required.
| Agent | Capabilities |
|---|---|
| Energy Agent | Real-time power monitoring, idle machine detection, department analysis, device control |
| Cost Agent | EPEX/ENTSO-E price tracking, cost forecasting, budget monitoring, savings opportunities |
| Anomaly Agent | 4-layer anomaly detection, root cause analysis, device health scoring |
| Maintenance Agent | Predictive maintenance, overdue task detection, equipment health overview |
A comprehensive carbon management system for monitoring, reporting, and reducing industrial CO2 emissions.
- Real-time CO2 Dashboard with animated gauges for total emissions, offsets, and net emissions
- Sustainability Score (0-100) with color-coded status and trend tracking
- Scope 1/2/3 Emissions Accounting with both market-based and location-based Scope 2
- Multi-country Emission Factors with 2024 data (NL 0.336, DE 0.364, FR 0.056, EU avg 0.251 kg CO2/kWh)
- Department & Machine Breakdown showing per-source emission contributions
- Environmental Impact Cards translating emissions into tangible equivalents (trees, car km, flights)
- Carbon Offset Management with certificate tracking, verification workflow, and provider management
- Automated Compliance Reports (Daily/Weekly/Monthly/Yearly/Custom) with PDF/Excel export
Full support for ISO 50001:2018 energy management system requirements.
- Clause-by-clause checklist tracking for all ISO 50001 requirements
- Evidence collection and documentation per requirement
- Responsible user assignment with due dates and priority levels
- Periodic energy reviews with baseline comparisons and EnPI results
- Improvement action tracking with findings documentation
- Review/approval workflow (Draft → In Review → Completed)
- Baseline year comparisons with target year reduction percentages
- Scope-specific targets (Scope 1, 2, 3 individually configurable)
- Methodology tracking (SBTi, internal, regulatory)
- Progress visualization with trend analysis against reduction goals
- Embedded Aedes MQTT broker for local IoT device communication
- Protocol adapters: Modbus TCP, OPC-UA, BACnet for industrial equipment
- External MQTT client for connecting to cloud brokers
- Device adapter translating MQTT messages to AI-IDLE data model
- ONNX Runtime inference on edge nodes for low-latency anomaly detection
- Edge-cloud sync with bidirectional data synchronization
- OTA firmware updates with staged rollout strategies (canary, rolling, blue-green)
- ENTSO-E Transparency Platform integration for real NL Day-Ahead prices
- Price signal reactor monitoring EPEX spot prices with configurable thresholds
- Genetic scheduler for multi-device schedule optimization
- Battery arbitrage service for storage charge/discharge strategy
- Constraint solver for operational constraint satisfaction
- Modes: simulation, supervised (manual approve), autonomous
- User-defined profiles with custom phase definitions (power range, duration range per phase)
- Learn from machine auto-generates profiles from cycle history (P5/P95 percentiles)
- Profile editor with drag-and-drop phase ordering and block diagram preview
- Linked to machines for cycle detection with custom thresholds
- 5-factor cycle health scoring: duration drift, efficiency degradation, peak power trend, phase skip rate, cycle count anomaly
- Per-machine health reports with weighted overall score (0-100)
- Integrated into existing predictive maintenance engine with weight renormalization
- Cycle duration constraints: scheduler ensures contiguous blocks match appliance cycle length
- Inter-cycle pause enforcement and max cycles per day limits
- Automatic constraint derivation from cycle history and custom profiles
- Enhanced genetic algorithm fitness with cycle-aware penalties
- 4 new report types: Daily, Cycle, Maintenance, Custom
- Cycle reports: phase breakdown, efficiency trends, top-5 machines by cycle count
- Maintenance reports: task completion rate, active schedules, overdue analysis
- Excel export via ExcelJS with 5 worksheets (Overview, Machines, Cycles, Anomalies, Costs)
- Floating Action Button (FAB) with quick actions: scan device, generate report, agent chat
- Compact mobile KPI cards with trend indicators (visible on mobile only)
- Swipeable tab navigation using motion/react drag gestures
- Pull-to-refresh gesture on dashboard
- PWA shortcuts for Dashboard, Scan, and Reports
- Playbook-based remediation for automatic incident resolution
- Incident detection from device offline events and critical anomalies
- Auto-recovery with configurable confidence thresholds
- FastAPI microservice for advanced ML workloads
- ONNX model serving with hot-reload
- Federated learning for privacy-preserving model training
- Explainability (SHAP/LIME) for model transparency
- 4-layer anomaly detection: Statistical (Z-score), ML (Isolation Forest, VAE), pattern-based, context-aware
- Appliance Signature Recognition: 18+ appliance types with cycle detection
- TensorFlow.js for GPU-accelerated server-side and browser-side inference
- Energy predictions per machine, department, and site
- Interactive 3D factory floor with live energy data (Three.js + React Three Fiber)
- 2D layout editor with drag-and-drop zone management
- Color-coded status: active (green), idle (amber), offline (gray)
- Carbon footprint calculation with Dutch/EU grid emission factors
- Scope 1/2/3 emissions tracking with ISO 50001 compliance
- Gamification: challenges, badges, leaderboards
+---------------------------+
| Nginx (SSL / LB) |
+-------------+-------------+
|
+--------------------+--------------------+
| |
+----------v----------+ +--------------v--------------+
| React 19 Frontend | | Express 5 API |
| Vite 7 . TW 4 |<------------>| Node.js 24 LTS |
| Zustand 5 . TQ 5 | Socket.IO | Prisma 7 (108 models) |
| Three.js . TF.js | | BullMQ 5 . Apollo GQL 5 |
+-----------+----------+ +--------------+--------------+
|
+-----------------------------+-------------------+
| | |
+----------v--------+ +----------v--------+ |
| PostgreSQL 16 | | Redis 7 | |
| + TimescaleDB | | Cache . PubSub | |
+-------------------+ +--------------------+ |
|
+------------------------------------------------------------v-+
| ML Sidecar (Python 3.12 / FastAPI) |
| ONNX Runtime . Federated Learning . SHAP/LIME |
+--------------------------------------------------------------+
| Aedes MQTT Broker + Protocol Adapters (Modbus/OPC-UA/BACnet)|
+--------------------------------------------------------------+
| Edge Computing (ONNX inference . Edge-Cloud Sync . OTA) |
+--------------------------------------------------------------+
| Demand Response (Price Signals . Genetic Scheduler . DR) |
+--------------------------------------------------------------+
| Carbon Compliance (Scope 1/2/3 . ISO 50001 . Offsets) |
+--------------------------------------------------------------+
| Self-Healing Engine (Playbooks . Incident Detection) |
+--------------------------------------------------------------+
| 4 AI Agents (Energy . Cost . Anomaly . Maintenance) |
+--------------------------------------------------------------+
Shelly (Gen1+Gen2+Gen3) . Tasmota . Tuya . MQTT . Modbus
| Layer | Technology |
|---|---|
| Frontend | React 19, TypeScript 5.9 (strict), Vite 7, Tailwind CSS 4, Zustand 5, TanStack Query 5, Recharts 3, Three.js, TensorFlow.js 4, Motion 12, i18next (NL/EN) |
| Backend | Express 5, TypeScript 5.9 (strict), Prisma 7 (108 models), Apollo Server 5 (GraphQL), Socket.IO 4, BullMQ 5, Zod 4, Winston 3 |
| ML Sidecar | Python 3.12, FastAPI, ONNX Runtime, scikit-learn, SHAP, federated learning |
| IoT/Edge | Aedes MQTT broker, Modbus TCP, OPC-UA, BACnet protocol adapters, ONNX edge inference, OTA firmware manager |
| Energy | ENTSO-E Transparency Platform, EPEX Day-Ahead pricing, genetic scheduler, battery arbitrage, constraint solver |
| Carbon | Scope 1/2/3 emissions, multi-country emission factors, ISO 50001 compliance, carbon offset verification |
| AI/ML | Rule-based agent engine, TensorFlow.js, Isolation Forest, VAE, appliance signature recognition |
| Database | PostgreSQL 16 + TimescaleDB, Redis 7 |
| CI/CD | GitHub Actions, Docker Compose, Nginx reverse proxy |
287,000+ lines of code (TypeScript + Python)
108 Prisma database models
118 backend service modules
161 frontend components
146 page components
58 API route files
49 API controllers
23 middleware modules
47 custom React hooks
16 Zustand stores
21 AI agent tools
18 appliance profiles
4 autonomous AI agents
9 carbon compliance models
2 languages (NL/EN)
830+ test files
| Tool | Version |
|---|---|
| Node.js | 24 LTS |
| PostgreSQL | 16+ with TimescaleDB |
| Redis | 7+ |
| Python | 3.12+ (for ML sidecar) |
| Docker | 24+ (recommended) |
git clone https://github.com/WimLee115/ai-idle.git
cd ai-idle
cp backend/.env.example backend/.env
# Edit .env: set DB_PASSWORD, JWT_SECRET, ENCRYPTION_KEY, ENTSOE_API_TOKEN
docker compose up -d# Backend
cd backend
cp .env.example .env
npm install
npx prisma generate --config prisma/prisma.config.ts
npx prisma migrate deploy --config prisma/prisma.config.ts
npm run dev # http://localhost:5000
# Frontend (new terminal)
cd frontend
npm install
npm run dev # http://localhost:3000
# ML Sidecar (new terminal)
cd ml-sidecar
pip install -r requirements.txt
uvicorn src.main:app --port 8001# In backend/.env:
ENTSOE_API_TOKEN=your-token-from-transparency.entsoe.eu
DEMAND_RESPONSE_ENABLED=true
DEMAND_RESPONSE_MODE=simulation # or supervised / autonomous| Version | Highlights |
|---|---|
| v1.8.0 | Carbon compliance & emissions tracking, ISO 50001 support, Scope 1/2/3 accounting, carbon offset management, multi-country emission factors, zero lint warnings |
| v1.7.0 | Custom appliance profiles, cycle health scoring, cycle-aware scheduling, enhanced reporting engine, mobile responsive dashboard |
| v1.5.0 | MQTT broker, edge computing, demand response, self-healing, ML sidecar |
| v1.2.1 | Agentic AI system with 4 autonomous agents, floating agent chat |
| v1.0.1 | Node 22, Express 5, React 19, Motion 12 upgrades |
AI-IDLE is seeking 2 pilot partners in the SME metal/manufacturing sector.
What we offer (free):
- Full platform installation, setup, and onboarding
- Direct access to the developer for customization and support
- All updates and new features during the pilot
- Comprehensive review and optimization report
What the pilot partner provides:
- A production environment with at least 10 machines
- Willingness to provide feedback
- Device costs (Shelly smart plugs, ~EUR 15-25 per machine)
- Email: ai-idle@outlook.com
- Public repo: github.com/WimLee115/ai-idle-platform
- Discussions: GitHub Discussions
Copyright (c) 2024-2026 AI-IDLE. All rights reserved.
This is proprietary software. Contact ai-idle@outlook.com for licensing information.
Built with dedication in the Netherlands
Making industrial energy waste a thing of the past, one machine at a time.