metadata

title: SentinelOne
emoji: 🛰️
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 5.33.0
app_file: app.py
pinned: false
license: mit
short_description: AI Agents for climate risk assessment and monitoring

🛰️ SentinelOne – Climate Risk Assessment Platform

Welcome to our submission for the Hugging Face GenAI Agents & MCP Hackathon!
This project showcases a multi-agent system powered by Gradio and Smolagents, designed to monitor, process, and analyze environmental data to assess climate-related risks and provide recommendations. Link to demo : https://youtu.be/aLK_HUSRl4w

🚨🟡🌍 IMPORTANT NOTICE FOR HACKATHON! 🤖🛰️🚨

🏆 If our project wins a prize at the hackathon, we will use the award to purchase additional API credits so we can continue development and maintain the Agents for everyone. 🏆

🙏 Your support will directly contribute to making this open-source climate risk platform sustainable and accessible! 🙏

💚 Thank you for your trust and your votes! 💚

🌐 Project Overview

This system leverages autonomous agents that collaborate to perform complex data science workflows:

🛰️ Risk Assessment Agent: Analyzes environmental data and assesses potential climate-related risks.
📊 Recommendation Agent: Provides actionable recommendations based on risk assessments.
🎯 Orchestrator: Coordinates the workflow (deterministic for now) between different agents and tools.

🗂️ Data Sources & External APIs

Our platform relies on real, authoritative, and up-to-date environmental data sources. Each data stream is directly queried and processed by dedicated AI tools, enabling both precise geolocation and a truly holistic risk assessment.
Here is a detailed overview of the APIs and data sources leveraged by the agents:

1. OpenStreetMap – Geolocation & Mapping

Purpose: Converts city names or user input into precise latitude/longitude coordinates.
API Used: Nominatim API
Usage: All risk assessments start by locating the point of interest. This is the backbone for all downstream weather, risk, and local resource queries.

2. Open-Meteo – Weather, Climate, Marine & Flood Data

A suite of APIs offering a wide spectrum of meteorological and environmental data:

Current & Forecasted Weather:
- api.open-meteo.com/v1/forecast
- Daily/weekly forecasts: temperature, precipitation, wind, humidity.
Historical Weather:
- archive-api.open-meteo.com/v1/archive
- Up to 50 years of daily meteorological records for any coordinate.
Climate Change Projections:
- climate-api.open-meteo.com/v1/climate
- Multi-model climate projections (1950–2050), covering temperature, rainfall, humidity, wind, radiation, soil moisture, etc.
Flood Risk:
- flood-api.open-meteo.com/v1/flood
- River discharge, flood probability, and trends, updated daily.
Marine & Coastal Data:
- marine-api.open-meteo.com/v1/marine
- Daily oceanic conditions: wave heights, swell, periods, and wind on sea.
Satellite Solar Radiation:
- satellite-api.open-meteo.com/v1/archive
- Hourly solar and terrestrial radiation from satellite observation.
Air Quality:
- air-quality-api.open-meteo.com/v1/air-quality
- Hourly PM10, PM2.5, CO, NO2, SO2, ozone, pollen, and AQI.

Open-Meteo is open, reliable, and integrates data from a variety of meteorological institutes.

3. USGS – Earthquake Data

Purpose: Provides real-time and historical seismic activity near a given location.
API Used: USGS Earthquake Catalog API
Usage: Retrieves all seismic events (magnitude, time, depth, location) in a configurable radius and time window (default: last 30 days, within 100 km).

4. NASA FIRMS – Wildfire Detection

Purpose: Detects active wildfires using data from earth observation satellites.
API Used: NASA FIRMS (Fire Information for Resource Management System)
Data: MODIS & VIIRS satellite observations, real-time fire detections (location, brightness, confidence, time, etc.).
Authentication: Requires a NASA FIRMS API key (set in .env).
Usage: All fire events in the last 2 days within a given radius (default: 50km) of the query point.

5. OpenStreetMap – Local Emergency Resources

Purpose: Finds nearest hospitals, fire stations, and police stations for a given location.
API Used: Overpass API
Usage: Critical for vulnerability assessment and emergency preparedness, especially in high-risk areas.

📦 Data Integration Logic

All data is retrieved in real time and cross-referenced using geospatial coordinates. The system is designed to:

Query and aggregate up-to-date meteorological, geophysical, and environmental signals
Offer long-term context (climate, trends) and immediate alerts (fire, flood, quake, air pollution)
Provide geographic context: local emergency resources and infrastructural exposure
Output both raw data and actionable visualizations (charts, maps, PDF reports)

Example API calls in our workflow:

🌦️ Weather: Open-Meteo
🔥 Wildfire: NASA FIRMS
🌊 Flood: Open-Meteo Flood API
🌐 Mapping & Resources: OpenStreetMap Nominatim & Overpass
🌍 Seismic: USGS
🏭 Air Quality: Open-Meteo Air Quality API
🏝️ Marine and coastal data: Open-Meteo Marine API
☀️ Solar radiations: Open-Meteo Satellite API
📈 Climate Change: Open-Meteo Climate API

🧩 Core Components

👁️ Risk Assessment Agent (`risk_agent.py`)

Role: Analyzes environmental data and assesses risks
Responsibilities:
- Process environmental data
- Evaluate risk levels
- Generate risk reports

🧠 Recommendation Agent (`recommendation_agents.py`)

Role: Provides actionable recommendations
Responsibilities:
- Analyze risk assessments
- Generate tailored recommendations
- Prioritize actions

🔧 Tools (`tools.py`)

Data processing utilities
Mapping utilities (mapping_utils.py)
Configuration management (config.py)

🎨 User Interface (`ui.py`)

Interactive Gradio interface
Real-time data visualization
User-friendly controls

🧑‍💻 Authors

Tristan DARRIGOL – AI Consultant @Wavestone
Clément PEPONNET – AI Agents engineer @Wavestone
Paul BARBASTE – AI Agents architect @Wavestone & AI Researcher @Inclusive Brains

📝 License

This project is licensed under the MIT License