Multimodal RAG Application

An interactive Streamlit dashboard that allows users to upload PDFs (text + images), and get AI-powered answers using Google Gemini, LangChain, and HuggingFace embeddings.

Overview

This app demonstrates a Multimodal RAG (Retrieval-Augmented Generation) system — it can understand and answer questions based on:

• Text content inside PDFs
• Images embedded in the PDF
• Tables
• Combined multimodal reasoning (text + visual)

Simply upload a PDF file, ask any question, and the model will answer based only on your document content.

Tech Stack

• Frontend: Streamlit

• AI Model: Google Gemini 2.5 Flash

• Framework: LangChain

• Vector Store: ChromaDB

• Embeddings: HuggingFace (Sentence Transformers)

• File Handling: PyMuPDF + base64

• Language: Python 3.10+