IntelliPDF is a high-performance Retrieval-Augmented Generation (RAG) pipeline designed to unlock the visual and textual context of PDF documents.

Unlike traditional text-only search, IntelliPDF uses Late Interaction (ColPali/ColQwen/ColSmol) to process entire PDF pages as images, ensuring charts, diagrams, and layouts are understood alongside text.

Powered by a two-stage hybrid retrieval engine, it combines fast prefetching with high-precision neural reranking, delivering deep contextual understanding at scale.

Key Features

• Visual-First Understanding: Uses advanced Vision-Language Models to index PDF pages as images, capturing layout and non-textual data.
• Two-Stage Retrieval Pipeline: Combines Qdrant-powered fast prefetching with Late Interaction reranking for maximum accuracy and speed.
• GPU-Accelerated & Optimized: Features batch ingestion, multi-vector embeddings, and scalar quantization, achieving up to a 13x reduction in retrieval time.
• Advanced Hybrid Search: Leverages pooled embeddings and multi-vector search to find the most relevant content instantly.
• Context-Aware QA: Delivers precise answers by feeding retrieved visual context directly to the LLM.
• Full-Stack Solution: Includes a local Streamlit UI for effortless batch PDF uploading, indexing, and interactive querying.
• Flexible LLM Support: Compatible with local or remote LLMs via LiteLLM.