🌳 ReforestAI – Geospatial Assistant

Google Cloud Hackathon Project

🛰️ A Streamlit app powered by an LLM Agent (Google ADK) to analyze, query, and visualize reforestation data in Madagascar, in order to support the NGO Interaide in their reforestation and sustainable development mission.

---

🌍 Overview

ReforestAI combines AI-powered geospatial reasoning and interactive mapping to support reforestation analysis.
It integrates:

• 🗺️ GeoJSON layers (Fokontany boundaries, Grevillea plantations, Reforestation zones)
• 🤖 Google ADK Agent for natural language queries
• 📊 Folium maps for beautiful visualizations

---

🎬 Demo

Watch a quick demo of the ReforestAI agent in action. See the demo at demo.mp4

---

✨ Features

✅ Interactive chat interface to run spatial and attribute queries
✅ Dynamic Folium maps with Stamen Terrain base layer
✅ Tools for:

• Spatial joins & geometry enrichment
• Attribute filtering
• Map export to PNG/HTML

---

🗂️ Repository Structure

📦 ReforestAI
 ┣ 📂 app/
 ┃ ┗ 📜 streamlit_app.py        → Streamlit frontend (UI + API calls)
 ┣ 📂 reforestAI-agent/
 ┃ ┣ 📜 agent.py                → LLM Agent definition (Google ADK)
 ┃ ┗ 📜 tools.py                → Geospatial utilities (GeoPandas, Folium)
 ┣ 📂 Map_Data/geojson/         → Sample GeoJSON datasets
 ┃ ┣ 📜 Fokontany.geojson
 ┃ ┣ 📜 Grevillea.geojson
 ┃ ┗ 📜 Reboisement.geojson
 ┣ 📂 output/                   → Generated maps & exports
 ┣ 📜 requirements.txt
 ┗ 📜 README.md

---

⚙️ Prerequisites

Local (without Docker)

• 🐍 Python 3.10+ (3.12 recommended)
• 🔑 Google API Key for the LLM agent
• 📦 Dependencies listed in requirements.txt (GeoPandas, Fiona, Folium, etc.)

Docker

• 🐳 Docker 20.10+
• 🧩 Docker Compose v2+

💡 On Windows, use Docker Desktop. On Linux/macOS, install Docker Engine + Compose Plugin.

---

🧰 Local Setup

macOS / Linux

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Windows (PowerShell)

python -m venv .venv
. .venv\Scripts\Activate.ps1
pip install -r requirements.txt

---

🔐 API Configuration

1️⃣ Copy the example environment file:

cp reforestAI-agent/.env.example reforestAI-agent/.env

2️⃣ Add your API key:

GOOGLE_API_KEY=<your_key>

You can create an API key at Google AI Studio

---

🚀 Running (Local)

1) Start the ADK server

adk api_server

The server must listen on http://localhost:8000 and expose the app reforestAI-agent.

2) Launch the Streamlit app

streamlit run app/streamlit_app.py

In the sidebar, create a session and start exploring spatial data through chat prompts.

---

🐳 Running with Docker

Provided structure

The repository includes a Dockerfile (for the app) and a docker-compose.yml orchestrating:

• agent : Google ADK server
• web : Streamlit web interface
• A volume for output/ to retrieve exported maps

Environment variables: ensure reforestAI-agent/.env contains your GOOGLE_API_KEY.

1) Build & Run

docker compose up --build

2) Access the application

• Streamlit (Web UI): http://localhost:8501

🧠 Example Queries

• “Show me a map of the Antavibe with grevillea patches on the Antavibe Fokonany.”
• “Generate a map that includes only the 'Reboisement' and 'Grevillea' layers restricted to the Ambalona Fokontany region.”
• “List me the Fokontanys in the Sandrohy Commune.”
• "Show me a map of Grevillea plantations where the surface area is greater than 5000 m²."

---

🧩 Tech Stack

Layer	Technology
🌐 Frontend	Streamlit
🧠 Agent Framework	Google ADK
🗺️ Geospatial	GeoPandas, Folium, Shapely, Fiona
💾 Data	GeoJSON
🔌 API	HTTP (Agent @ :8000)
🐳 Runtime	Docker & Docker Compose
🧰 Language	Python 3.12

---

🌐 Online Deployment

The app is live and accessible online! 🚀

You can explore the interactive interface and test the LLM agent directly here:
👉 https://reforestai-agent-front-396062328299.europe-west9.run.app/

💡 The online version uses the same ADK Agent backend and geospatial datasets but may have limited export capabilities.

---

⭐ If you like this project, give it a star on GitHub! ⭐