RAG-Chat
RAG-Chat is a document chat application that processes files in the background, stores searchable vectors, and returns answers with source citations.
RAG-Chat was built to understand the full path from uploaded document to final answer, not just the chat interface. Users can upload files, let them process in the background, and then ask questions against that content.
RAG-Chat was built to understand the full path from uploaded document to final answer, not just the chat interface. Users can upload files, let them process in the background, and then ask questions against that content.
The system keeps file processing separate from the chat path. Workers extract and split the content, store search-ready vectors in Qdrant, and the chat layer pulls back the most relevant pieces before answering.
How It Was Built
The main technical choices behind the product, from system design to the parts that make it work day to day.
- Separated uploads and processing from the chat path so file work does not block the user-facing experience.
- Built parsing for PDFs, Word files, and spreadsheets, then split that content into chunks sized for search.
- Separated uploads and processing from the chat path so file work does not block the user-facing experience.
- Built parsing for PDFs, Word files, and spreadsheets, then split that content into chunks sized for search.
- Used embeddings and Qdrant vector search so questions are matched to the most relevant document chunks before the model answers.
- Showed processing status and citations in the product so users can see when files are ready and where answers came from.