Skip to content

README.md

Latest commit

 

History

History
191 lines (157 loc) · 7.12 KB

README.md

File metadata and controls

191 lines (157 loc) · 7.12 KB

SyncOpsAI — AI-Powered Equipment Monitoring

IBM Hackathon POC | Multi-Agent System | watsonx.ai | watsonx Orchestrate | RAG

Overview

SyncOpsAI monitors manufacturing equipment in real time and automatically detects anomalies, diagnoses root causes, and generates maintenance work orders. Three specialized AI agents replace a manual process that previously took hours.

How It Works

graph TB
    A[IoT Sensors] -->|Real-time Data| B[Agent 1: Telemetry Listener]
    B -->|Anomaly Detected| C[Agent 2: Diagnostic Expert]
    C -->|Query| D[RAG System]
    D -->|Context| E[Equipment Manuals]
    C -->|Generate| F[watsonx.ai Granite]
    F -->|Diagnosis| C
    C -->|Report| G[Agent 3: Orchestrator]
    G -->|Create| H[Work Order]
    G -->|Check| I[Inventory]

    style C fill:#4CAF50
    style F fill:#FF9800
    style D fill:#2196F3
Loading
Agent Role
Agent 1 — Telemetry Listener Monitors sensor data, detects anomalies, classifies severity
Agent 2 — Diagnostic Expert Queries equipment manuals via RAG, generates AI diagnosis using IBM Granite
Agent 3 — Orchestrator Checks inventory, creates prioritized work order

Quick Start

# 1. Install dependencies
pip install -r requirements.txt

# 2. Configure environment
cp .env.example .env
# Add your WATSONX_API_KEY, WATSONX_URL, WATSONX_PROJECT_ID

# 3. Start the Mock API
python mock_apis/app.py
# Runs on http://localhost:8787

# 4. Run the dashboard
streamlit run dashboard/app.py
# Opens at http://localhost:8501

Startup Commands

# Option 1: Start everything at once
cd dashboard && ./run_dashboard.sh

# Option 2: Start separately

# Terminal 1 — Mock API
cd mock_apis && python app.py

# Terminal 2 — Dashboard
cd dashboard && streamlit run app.py

Project Structure

SyncOpsAI/
├── poc/                              # Core agent logic (POC)
│   ├── agents.py                     # Agent 1, 2, 3 implementations
│   ├── main.py                       # LangGraph workflow orchestration
│   ├── state.py                      # Shared state schema
│   ├── data.py                       # Demo scenarios & thresholds
│   ├── dashboard.py                  # POC Streamlit interface
│   ├── integration.py                # POC integration layer
│   └── requirements.txt
│
├── dashboard/                        # Production dashboard
│   ├── app.py                        # Streamlit app (4 tabs, live gauges)
│   ├── run_dashboard.sh              # One-command startup script
│   └── requirements.txt
│
├── mock_apis/                        # Mock REST API servers
│   ├── app.py                        # Main Flask API
│   ├── work_order_api.py             # Work order endpoints
│   ├── inventory_api.py              # Inventory endpoints
│   ├── technician_api.py             # Technician endpoints
│   ├── test_api.py                   # API tests
│   ├── start_all.sh                  # Start all API servers
│   └── stop_all.sh                   # Stop all API servers
│
├── orchestrate/                      # watsonx Orchestrate
│   ├── agents/
│   │   ├── telemetry_agent.yaml      # Agent 1 config
│   │   ├── diagnostic_agent.yaml     # Agent 2 config
│   │   └── orchestrator_agent.yaml   # Agent 3 config
│   ├── tools/
│   │   ├── detect_anomaly.py         # Anomaly detection tool
│   │   ├── generate_diagnosis.py     # Diagnosis generation tool
│   │   ├── create_work_order.py      # Work order tool
│   │   └── check_inventory.py        # Inventory check tool
│   └── import-all.sh                 # Deploy all agents & tools
│
├── tools/                            # watsonx Orchestrate tool YAMLs
│   ├── work_order_api.yaml
│   ├── inventory_api.yaml
│   └── technician_api.yaml
│
├── bob_sessions/                     # IBM Bob session logs
│
├── logs/                             # API runtime logs
│
├── data.py                           # Sensor data scenarios
├── manuals.py                        # Equipment manual knowledge base
├── rag.py                            # RAG system
├── diagnosis.py                      # Diagnostic engine
├── agents.py                         # Root-level agent workflow
├── watsonx_integration.py            # watsonx.ai Granite wrapper
├── watson_orchestrate_integration.py # watsonx Orchestrate wrapper
├── pinecone_integration.py           # Pinecone vector DB integration
├── verify_ai_config.py               # AI configuration checker
├── workspace_config.yaml             # Workspace configuration
└── requirements.txt                  # Root dependencies

Demo Scenarios

HVAC Overheating

  • Equipment: HVAC-001
  • Signal: Temperature 22°C → 32°C (threshold: 28°C)
  • AI Diagnosis: Clogged air filter restricting airflow (92% confidence)
  • Parts: AF-2024 — $45, 30 min repair
  • Result: Work order created automatically in <5s

Motor Vibration

  • Equipment: MOTOR-001
  • Signal: Vibration 1.2 → 4.5 Hz (threshold: 3.5 Hz)
  • AI Diagnosis: Worn roller bearings (88% confidence)
  • Parts: RB-500 + AS-KIT — $310, 2–3 hrs repair
  • Result: Work order created automatically in <5s

Tech Stack

Layer Technology
AI Diagnosis IBM watsonx.ai — Granite-13B-Chat-v2
Agent Orchestration IBM watsonx Orchestrate
Development Assistant IBM Bob
Workflow Graph LangGraph
Vector DB Pinecone (production)
UI Streamlit + Plotly
API Flask
Language Python 3.9+

IBM Technology Usage

IBM Bob — Used throughout development for architecture planning, parallel workstream design, and code generation across all modules.

IBM watsonx.ai — Powers Agent 2. Builds a prompt from sensor readings and RAG-retrieved manual context, calls Granite-13B-Chat-v2, and returns a structured JSON diagnosis with root cause, confidence score, and recommended actions. Falls back to template-based diagnosis if unavailable.

IBM watsonx Orchestrate — Three agent YAML configurations and four @tool-decorated Python functions are implemented and ready for deployment. Live connection is the immediate next step — no structural changes required.

Performance

Metric Value
Anomaly detection <1ms
AI diagnosis 2–5s
End-to-end (anomaly → work order) <5s
Confidence score (AI mode) 85–95%
Diagnosis time vs manual 95% faster

Status

Component Status
Multi-agent workflow ✅ Complete
watsonx.ai integration ✅ Complete
RAG system ✅ Complete
Mock API (10 endpoints) ✅ Complete
Streamlit dashboard ✅ Complete
watsonx Orchestrate structure ✅ Complete
Pinecone integration 🏭 Production
Live Orchestrate connection 🏭 Production

Built for IBM Hackathon — Team Manual-Miners