This repository is a part of submission for Reboot the EarthGlobal Tech Challenge 2025 Hackathon organized by United Nations, Salesforce and CMU.

AgroSynth-RAG

AgroSynth-RAG combines real-time weather alerts with a retrieval-augmented LLaMA pipeline to give farmers both immediate climate warnings and localized pest-infestation risk forecasts.

Team Members:

• Tanishq (Myself)
• Shresth
• Aditya
• Hannah
• Tram

---

These are the major components in this repository:

1. weatherAlert.py: This file is used to provide real time weather alerts if in case there are any anomalies or severe weather that can potenially impact the crops.
2. query_llama_rag.py: This file utilizes a corpus of data related to Agricultural pests in New York state which includes certain open source articles, and information about the species of the pests with regards to the climatic conditions and the counties they are usually found in. We utilized Llama 3.2 and FAISS database to implement this RAG.
3. RAG Indexing: This folder contains the data(context) utilized by the RAG app and also transforming the data into vectors to implement vector search.

🔍 Features

• Weather Alerts (weatherAlert.py & thresholds.js)

  • Heavy rain (>40 mm), flood risk, high winds (>60 km/h), frost (<5 °C) and heat (>38 °C) notifications
  • City-to-lat/lon geocoding via Open-Meteo API
  • Simple CLI tool (python weatherAlert.py)

• Pest-Risk RAG (query_llama_rag.py)

  • FAISS-indexed pest datasheets (region → top 3 pests)
  • 7-day forecast (GDD, humidity, rain) from Open-Meteo
  • OpenRouter LLaMA call to answer “Will there be infestation? If yes, which pest?”
  • Deterministic “No infestation expected.” or “Yes – (including the probable species)” along with a brief generative explaination as to Why?

---

Agrosynth Frontend and Framework

• https://github.com/shresthkapoor7/agrosynth

🚀 Getting Started

1. Clone & Install

git clone https://github.com/Tans37/AgroSynth-RAG.git
cd AgroSynth-RAG
python3 -m venv venv
source venv/bin/activate    # Windows: venv\Scripts\activate
pip install --upgrade pip
pip install -r requirements.txt

requirements.txt should include:
requests, fastapi, uvicorn, langchain, faiss-cpu,
sentence-transformers, openai, transformers, python-dotenv

2. Configure Environment

Create a .env (or set environment variables) for OpenRouter LLaMA:

export OPENAI_API_KEY="your_openrouter_api_key"
export OPENAI_API_BASE="https://openrouter.ai/api/v1"
export OPENAI_MODEL="openrouter-llama-2-7b"

3. Run Weather Alerts

python weatherAlert.py
# Follow the prompt to enter a city name

4. Run Pest-Risk Query

python query_llama_rag.py
# Or integrate into FastAPI/Ngrok for React frontend:
uvicorn backend:app --reload --host 0.0.0.0 --port 8000

---

📁 Repository Structure

AgroSynth-RAG/
├── weatherAlert.py        # City → weather alerts CLI
├── thresholds.js          # JS constants for frontend alerts
├── query_llama_rag.py     # Script: region/crop → pest-infestation prediction
├── backend.py             # (optional) FastAPI wrapper for both flows
├── requirements.txt       # Python dependencies
└── README.md

---

🛠️ Integration

• React Frontend
  • Call /weather-alerts (POST { city }) → returns alerts: string[]
  • Call /pest-predict (POST { latitude, longitude, region, crop, planting_date }) → returns prediction: string
  • Display results in your UI with buttons or map-click handlers.