PDFToChat implements a complete RAG pipeline for PDF document Q&A. Users upload PDFs which are processed, chunked, and stored as vector embeddings in Chroma Cloud. When a user asks a question, the system retrieves relevant chunks via hybrid dense and sparse vector search using Qwen and SPLADE embeddings, then passes them to Mixtral via Together AI to generate context-grounded answers. The app is built with Next.js, uses Clerk for authentication, and stores files on Bytescale.