🌾 Agricult-AIcult RAG Bot

A Retrieval-Augmented Generation (RAG) bot designed to assist farmers and agriculture experts by answering crop-related queries using a document-based question-answering system. This project utilizes advanced LLM techniques, document retrieval, multi-query decomposition, reranking, and automatic evaluation using RAGAS.

📌 Features

• 💬 Accepts user questions about crops (e.g., onion cultivation)
• 📄 Extracts context from domain-specific PDFs
• 🔍 Performs:
  • Sub-question decomposition
  • Step-back question generation
  • Multi-query expansion
  • Reciprocal Rank Fusion (RRF)-based retrieval
  • LLM-based reranking
• 🧠 Generates answers with GPT-3.5 via OpenRouter
• ✅ Automatically evaluates responses using RAGAS (Faithfulness, Context Precision, Context Recall)
• 📊 Exports evaluation results to Excel

---

🧰 Tech Stack

Component	Tool/Library
Embedding	sentence-transformers/all-MiniLM-L12-v2
Vector Store	Chroma + FAISS
LLM	GPT-3.5-turbo via OpenRouter
Evaluation	RAGAS (Faithfulness, Context Precision, Context Recall)
PDF Parsing	PyMuPDF (fitz)
Query Handling	LangChain + custom chains
Excel Export	pandas + openpyxl

📂 Directory Structure

Agricult-AIcult-RAG ├── data/ # Folder containing crop PDFs ├── vectorstore/ # Chroma DB vector store (auto-created) ├── ragas_evaluation_report.xlsx # Evaluation results (auto-generated) ├── main.py # Full RAG pipeline script ├── requirements.txt # Required Python dependencies └── README.md # You're here!

---

🔧 Setup Instructions

1️⃣ Install Dependencies

Make sure you have Python 3.11.9+ installed.

Install all required packages:

pip install -r requirements.txt
* installed.

If you face issues with datasets, install it separately:
pip install datasets
2️⃣ Add Your PDF Files
Place your crop-specific PDF files inside the data/ directory.
Each file should be named according to the crop it covers, all lowercase with no spaces. For example:
data/
├── onion.pdf
├── wheat.pdf
├── tomato.pdf
3️⃣ Add Your OpenRouter & OpenAI Keys
In main.py, replace the following lines with your actual API keys:
# OpenRouter key for LLM calls
openai_api_key = "sk-..."  

# OpenAI key for RAGAS evaluation (used internally by ragas)
import os
os.environ["OPENAI_API_KEY"] = "sk-..."
🙏 Acknowledgments
LangChain

OpenRouter

RAGAS

HuggingFace Sentence Transformers