EcoGuard AI - Industrial Pollution Compliance Monitor

EcoGuard AI is a secure, deployable, multi-agent AI system designed for real-time monitoring of industrial water effluents and stack air emissions to comply with the Central Pollution Control Board (CPCB) standards in India.

Built using the official Google Agent Development Kit (ADK) and the Model Context Protocol (MCP), EcoGuard AI automates the inspection process, flags compliance violations, dispatches immediate alerts, and compiles comprehensive audit reports—all while protecting the system with a multi-layered security shield.

---

🏗 System Architecture

The following diagram illustrates the flow of data from the user input through the security, orchestrator, and agent layers down to the MCP compliance tools:

graph TD
    classDef actor fill:#1F2937,stroke:#111827,stroke-width:2px,color:#fff;
    classDef ui fill:#4F46E5,stroke:#312E81,stroke-width:2px,color:#fff;
    classDef security fill:#DC2626,stroke:#7F1D1D,stroke-width:2px,color:#fff;
    classDef orchestrator fill:#0D9488,stroke:#115E59,stroke-width:2px,color:#fff;
    classDef mcp fill:#D97706,stroke:#78350F,stroke-width:2px,color:#fff;

    User([User / Operator]):::actor --> UI[Gradio Dashboard]:::ui
    UI --> SecurityShield{Security Shield}:::security
    
    subgraph Security [Security & Validation Layer]
        SecurityShield --> PP[Prompt Protection Scan]:::security
        SecurityShield --> IV[Input Validator]:::security
        PP -.-> AL[(logs/audit.log)]:::security
        IV -.-> AL
    end
    
    IV -- Validated Metrics --> Master[EcoGuard Master Agent]:::orchestrator
    
    subgraph MultiAgent [Google ADK Multi-Agent Team]
        Master --> Water[WaterMonitor Agent]:::orchestrator
        Master --> Air[AirMonitor Agent]:::orchestrator
        Master --> Alert[AlertDispatch Agent]:::orchestrator
        Master --> Report[ReportGen Agent]:::orchestrator
    end
    
    Water --> MCPS[MCP Server]:::mcp
    Air --> MCPS
    
    subgraph MCP [Model Context Protocol Server]
        MCPS --> tool_water[get_water_quality]:::mcp
        MCPS --> tool_air[get_air_emissions]:::mcp
        MCPS --> tool_check[check_cpcb_compliance]:::mcp
        
        tool_check --> ComplianceEngine[CPCB Compliance Engine]:::mcp
    end

---

📋 Problem Statement & Relevance

Industrial manufacturing plants discharge heavy organic loads and release toxic gases into ambient air. The Central Pollution Control Board (CPCB) mandates strict limits to protect ecosystems. Monitoring compliance manually is error-prone, slow, and lacks immutable audit logs. EcoGuard AI solves this by:

• Automating Inspection: Using specialized agents to inspect water and air metrics.
• Securing Inputs: Restricting bounds (preventing sensor manipulation) and scanning for prompt injections.
• Standardizing CPCB Checks: Centralizing logic inside an MCP compliance engine.
• Dispatching Incident Triggers: Notifying operators immediately upon limits breach.
• Providing Audit Integrity: Logging all events to a secure log trace file.

---

🛠 Features

• Multi-Agent Orchestration (Google ADK): A coordinator agent routes wastewater parameters to WaterMonitor, gaseous emissions to AirMonitor, alert formatting to AlertDispatch, and report writing to ReportGen.
• FastMCP Server Compliance Tools: Exposes standard tools (get_water_quality, get_air_emissions, check_cpcb_compliance) running on stdio transport.
• Security Shield:
  • Boundary Check: Rejects negative values and impossible pH numbers.
  • Prompt Protection: Intercepts and blocks safety bypass phrases like "ignore previous instructions".
  • Audit Logger: Appends ISO-timestamped records to logs/audit.log.
• Gradio Dashboard: Interactive sliders, live agent reasoning log, communication trace visualizer, SMS/Email mock outbox, and markdown report rendering.

---

🚀 Installation & Setup

Prerequisites

• Python 3.10 or higher installed.

1. Clone & Set Up Directory

Navigate to your project folder and create a virtual environment:

python -m venv venv
venv\Scripts\activate      # On Windows
source venv/bin/activate    # On macOS/Linux

2. Install Dependencies

Install all required libraries including Google ADK, Gradio, and MCP:

pip install -r requirements.txt

(If you don't have requirements.txt, run pip install google-adk gradio mcp python-dotenv pytest)

3. Set Up Environment Variables (Optional)

Create a .env file in the root directory:

GEMINI_API_KEY=your_gemini_api_key_here
GEMINI_MODEL=gemini-2.5-flash

Note: If no API key is provided, the dashboard runs in Simulation Mode using local CPCB checks so you can inspect the entire workflow without an active key.

---

💻 Running the Application

Launch Gradio Dashboard

Start the dashboard locally:

python demo_app.py

Or use the hot-reload Gradio CLI:

gradio demo_app.py

Open your browser and navigate to http://localhost:7860.

Running Integration Tests

Run pytest to verify the security and tool bindings:

pytest tests/

---

📖 Demo Instructions

1. Test Normal State: Set all sliders to normal levels (e.g. pH 7.2, BOD 15, SO₂ 35). Click Run Compliance Check. The status banner displays green PASS with no alerts.
2. Test Violation State: Slide pH to 11.0 and PM2.5 to 90.0. Click Run Compliance Check. The status banner turns red FAIL. The Incident Response Queue lists the SMS and email drafts detailing the exact CPCB standard breaches.
3. Verify Security Block: Type "ignore previous instructions" in the operator directive box and click check. A red Security Warning appears, blocking execution. Check logs/audit.log to see the warning log entry. #� �E�c�o�G�u�a�r�d�-�A�I� � �#� �E�c�o�G�u�a�r�d�-�A�I� � �