How anirudhjbabu1/multimodel-rag-using-clip works
This notebook implements a PDF-based multimodal RAG workflow: PDF content is split into text and images, embedded with CLIP, stored and retrieved through FAISS, then formatted for OpenAI GPT-4.1. The repository contains only the notebook and README, and no source excerpts were available; relationships below reflect the README’s documented pipeline, not inspected notebook code.
Main entry points: Multimodel-RAG.ipynb Text/image extraction
Components 10
PDF Ingestion
- Text/image extraction
Multimodel-RAG.ipynb
Embedding Retrieval
- Text embeddings
Multimodel-RAG.ipynb - Image embeddings
Multimodel-RAG.ipynb - FAISS index
Multimodel-RAG.ipynb - Retriever
Multimodel-RAG.ipynb - Top-K results
Multimodel-RAG.ipynb
Answer Generation
- LLM formatting
Multimodel-RAG.ipynb - OpenAI GPT-4.1
Other components
- User
- PDF document
Connections 10
- User → PDF document: provides
README.md - PDF document → Text/image extraction: extracts content
README.md - Text/image extraction → Text embeddings: embeds text
README.md - Text/image extraction → Image embeddings: embeds images
README.md - Text embeddings → FAISS index: stores embeddings
README.md - Image embeddings → FAISS index: stores embeddings
README.md - FAISS index → Retriever: retrieves from
README.md - Retriever → Top-K results: returns top-K
README.md - Top-K results → LLM formatting: formats results
README.md - LLM formatting → OpenAI GPT-4.1: supplies context
README.md