EO Disaster Analyzer

A Python package for analyzing Earth Observation (EO) data for disaster management, augmented by Large Language Models (LLMs). The goal is to orchestrate the process of identifying potential disaster areas from news, fetching relevant satellite imagery, and generating analytical reports.

Features

• Efficient News Analysis: Implements a two-stage LLM pipeline to first quickly filter relevant disaster news and then perform detailed, structured data extraction.
• Configuration-driven: Manage API keys and settings easily via environment variables (.env).
• Data Providers: Modular connectors for fetching news from Google News and preparing for EO data sources.
• Text and Image Preprocessing: Includes utilities for cleaning text data with spaCy and preparing satellite imagery (tiling, normalization) for vision models.
• Structured LLM Outputs: Uses Pydantic schemas to ensure reliable, structured data from LLM analysis.
• Multiple Interfaces: Interact with the tool via a Command-Line Interface (CLI) or a REST API (FastAPI).

---

Installation

1. Clone the Repository

git clone <your-repository-url>
cd llm-eo-disasters

2. Create and Activate a Virtual Environment

python -m venv .venv
source .venv/bin/activate
# Windows: .venv\Scripts\activate

3. Install the package in editable mode (recommended for development)

pip install -e ".[dev,notebook]"

This reads dependencies from pyproject.toml and installs optional groups.

4. Set up environment variables

Copy the example file and fill in your values (SentinelHub, OpenAI, etc.):

cp .env.example .env
# Then edit `.env` with your configuration

---

Project Structure

The project follows a modern PEP 621 layout using pyproject.toml and a src/ directory.

eo_disaster_analyzer/
├── src/
│   └── eo_disaster_analyzer/
│       ├── __init__.py
│       ├── orchestrator/
│       │   ├── __init__.py
│       │   └── orchestrator.py
│       ├── data/
│       │   ├── __init__.py
│       │   └── sentinelhub_client.py
│       ├── llm/
│       │   ├── __init__.py
│       │   └── reasoning.py
│       ├── vision/
│       │   ├── __init__.py
│       │   └── segmentation.py
│       ├── cli.py
│       └── api.py
├── tests/
│   ├── __init__.py
│   ├── test_orchestrator.py
│   ├── test_data.py
│   ├── test_llm.py
│   └── test_vision.py
├── .env.example
├── pyproject.toml
├── README.md
└── .gitignore

Note: The sample structure above matches the current naming (eo_disaster_analyzer) and a modern package layout (src/), not the older setup.py layout.

---

Usage

Command-Line Interface

Once installed, the CLI entry point is available:

eo-analyzer --help

FastAPI Server

Start the API server:

uvicorn eo_disaster_analyzer.api:app --reload

Then open your browser at:

http://127.0.0.1:8000/docs

---

License

MIT License — feel free to use, modify, and distribute.