Rag-For-Engineering-Documents-with-Images-and-Diagrams

This project is a Retrieval-Augmented Generation (RAG) system designed to parse, extract, and search structured technical data from engineering drawings in PDF format. It uses Multimodal Large Language Models (LLMs), specifically Gemini, to process drawings page-by-page. The system extracts drawing details such as title blocks, revision history, dimensions, tolerances, and GD&T symbols into a structured JSON schema. These extracted data are then indexed into a vector database, enabling a conversational interface for asking technical questions with semantic context retrieval. The system supports both local and cloud-based operation modes, with local mode using FAISS for vector search and cloud mode using PostgreSQL with pgvector. It includes features like OCR using Gemini Vision, native PDF text extraction, hybrid information merging, semantic chunking, and a Flask web interface for real-time progress updates.