shivaprakash23/mcp_server_psetae

MCP Server for Sentinel-1 PSETAE

A multi-agent system implementation using the Model Context Protocol (MCP) to orchestrate Sentinel-1 PSETAE crop classification workflows.

Overview

This repository implements a Model Context Protocol (MCP) server that coordinates multiple AI agents to execute PSETAE (Pixel Set Encoder with Temporal Attention Encoder) workflows for crop classification using Sentinel-1 satellite imagery. The system integrates with Google Earth Engine (GEE) for data extraction and leverages the existing PSETAE codebase for model training and inference. This implementation focuses exclusively on Sentinel-1 data to establish a stable baseline before expanding to other data sources.

Features

• Multi-Agent Architecture: Hierarchical agent structure with Admin and specialized Worker agents
• Workflow Orchestration: End-to-end automation of PSETAE tasks from data extraction to inference
• GEE Integration: Seamless interaction with Google Earth Engine for satellite data retrieval
• Modular Design: Agents encapsulate specific functionality of the PSETAE pipeline
• Central Knowledge Base: Shared context and memory through MCP database

System Architecture

Agent Structure

• Admin Agent: Orchestrates the workflow and delegates tasks
• Worker Agents:
  • Sentinel1DataExtractionAgent: Handles GEE data retrieval and processing for Sentinel-1
  • Sentinel1ModelTrainingAgent: Manages model training and hyperparameter tuning for Sentinel-1
  • Sentinel1InferenceAgent: Applies Sentinel-1 models to new data
  • Sentinel1TileCoverageAgent: Analyzes satellite tile coverage for study areas
  • Sentinel1DocumentationAgent: Prepares the report based on the results of the PSETAE model that got trained.

Installation

1. Clone this repository:

git clone https://github.com/yourusername/mcp_server_psetae.git
cd mcp_server_psetae

2. Install dependencies:

pip install -r requirements.txt

3. Set up Google Earth Engine authentication:

earthengine authenticate

Usage

Starting the MCP Server

python server/server.py --port 8080 --project-dir /path/to/project

Initializing the Admin Agent

# First, retrieve the admin token from the database
# Then initialize the admin agent with:
# "Initialize as an admin agent with this token: [admin-token] Please add the MCD.md file to the project context. Don't summarize it."

Creating Worker Agents

Worker agents are created through the Admin Agent:

1. Ask the Admin Agent to create a worker (e.g., "Create a worker agent with ID 'data-extraction-worker'")
2. The Admin will provide a worker token
3. Initialize the worker with the provided token

Workflow Example

1. Project Initialization:

   • Define study area, time period, and satellite data sources

2. Data Extraction:

   • Extract HLS or Sentinel data using Google Earth Engine
   • Process and normalize the satellite imagery

3. Model Training:

   • Train PSETAE models with the extracted data
   • Tune hyperparameters for optimal performance

4. Inference:

   • Apply trained models to new data
   • Generate crop classification maps

5. Coverage Analysis:

   • Analyze satellite tile coverage for the study area
   • Generate coverage reports

Directory Structure

mcp_server_psetae/
├── server/                  # MCP server implementation
├── agents/                  # Agent implementations
├── utils/                   # Utility functions
├── config/                  # Configuration files
├── docs/                    # Documentation
└── tests/                   # Test scripts

Contributing

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add some amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• PSETAE codebase by Shivaprakash Yaragal, Lund University
• Agent-MCP framework by rinadelph
• Google Earth Engine team for the Python API