SmartRAG Assistant

This project allows you to ask questions about any PDF document using a hybrid retrieval-augmented generation (RAG) pipeline. It combines semantic search (FAISS) and keyword search (BM25) with Google Gemini LLM to provide accurate answers.

---

Features

• Upload and read PDF documents
• Chunk and preprocess text for better retrieval
• Semantic embeddings with HuggingFaceBgeEmbeddings
• Keyword search using BM25
• Hybrid retrieval with EnsembleRetriever
• Answer questions using Google Gemini LLM
• Interactive web interface via Gradio

---

Environment Setup

Install required packages:

pip install -q langchain langchain-community langchain-google-genai langchain-text-splitters faiss-cpu pypdf2 sentence_transformers gradio rank_bm25

Set your Google API key:

import os
os.environ["GOOGLE_API_KEY"] = "YOUR_GOOGLE_API_KEY"

---

Usage

• Upload a PDF
• Read the PDF text
• Split text into chunks
• Initialize embeddings
• Set up retrievers
• Initialize Google Gemini LLM
• Create prompt template
• Build RAG pipeline
• Ask questions
• Interactive web interface

---

Requirements

• Python 3.9+
• Google API key with access to Gemini models
• Packages listed in Environment Setup

---

Notes

• Ensure your PDF contains text (not scanned images) for proper extraction.
• The ensemble retriever combines semantic and keyword-based search for better results.
• Designed to run in Google Colab, but can be adapted locally.