🌍 Astra Agent

An intelligent assistant that converts natural language queries into executable GIS workflows using AI and open-source geospatial tools.

🚀 Features

• Natural Language Processing: Convert plain English queries to GIS operations
• Multi-Model AI Integration: Mistral-7B, LLaMA-3-8B via LangChain
• Comprehensive GIS Tools: QGIS, GRASS GIS, GDAL/OGR, GeoPandas
• Real Data Sources: Bhoonidhi, OpenStreetMap, STAC APIs
• Interactive Interface: Streamlit + OpenLayers/Leaflet mapping
• Scalable Backend: FastAPI + PostgreSQL/PostGIS

🏗️ Architecture

User Query → Streamlit UI → FastAPI → AI Models → GIS Tools → Spatial Data → Results

🛠️ Tech Stack

• AI: Mistral-7B-Instruct, LLaMA-3-8B, LangChain, Transformers
• GIS: QGIS, GRASS GIS, GDAL/OGR, GeoPandas, PostGIS
• Backend: Python FastAPI, PostgreSQL
• Frontend: Streamlit, OpenLayers, Leaflet
• Data: Bhoonidhi, OpenStreetMap, STAC APIs

📦 Installation

Prerequisites

• Python 3.9+
• PostgreSQL with PostGIS
• QGIS with PyQGIS
• Docker (optional)

Quick Setup

# Clone repository
git clone https://github.com/proavipatil/geospatial-ai-assistant.git
cd geospatial-ai-assistant

# Install dependencies
pip install -r requirements.txt

# Setup environment
cp .env.example .env
# Edit .env with your configurations

# Setup database
python scripts/setup_database.py

# Run the application
python scripts/run_app.py

Docker Setup

# Build and run with Docker
docker-compose up --build

🚀 Usage

Example Queries

# Natural language queries the system can handle:
"Find all schools within 1km of hospitals in Mumbai"
"Calculate population density by district using census data"
"Identify flood-prone areas using elevation and rainfall data"
"Generate land use classification from satellite imagery"

API Usage

import requests

# Submit query
response = requests.post("http://localhost:8000/api/query", 
                        json={"query": "Find schools near hospitals in Mumbai"})

# Get results
result = response.json()
print(result['workflow'])  # AI-generated workflow
print(result['results'])   # GIS analysis results

Web Interface

1. Start the application: python scripts/run_app.py
2. Open browser: http://localhost:8501
3. Enter natural language query
4. Review AI-generated workflow
5. Execute analysis and view results

📁 Project Structure

geospatial-ai-assistant/
├── src/
│   ├── ai/                 # AI processing modules
│   │   ├── query_processor.py
│   │   ├── workflow_generator.py
│   │   └── model_manager.py
│   ├── gis/                # GIS processing engines
│   │   ├── qgis_engine.py
│   │   ├── gdal_processor.py
│   │   └── spatial_operations.py
│   ├── data/               # Data access and management
│   │   ├── osm_client.py
│   │   ├── stac_client.py
│   │   └── bhoonidhi_client.py
│   ├── api/                # FastAPI backend
│   │   ├── main.py
│   │   ├── routes/
│   │   └── models/
│   └── frontend/           # Streamlit application
│       ├── app.py
│       ├── components/
│       └── utils/
├── config/                 # Configuration files
├── data/samples/           # Sample datasets
├── tests/                  # Test suites
├── docker/                 # Docker configuration
├── docs/                   # Documentation
└── scripts/                # Utility scripts

🧪 Testing

# Run all tests
pytest tests/

# Run specific test suite
pytest tests/unit/
pytest tests/integration/

# Run with coverage
pytest --cov=src tests/

📖 Documentation

• API Documentation
• User Guide
• Developer Guide
• Architecture Overview

🤝 Contributing

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

📄 License

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

🙏 Acknowledgments

• OpenStreetMap contributors
• QGIS development team
• Hugging Face for AI models
• Indian Space Research Organisation (ISRO) for Bhoonidhi

---