RAGify
Chat with any document, video, or website using RAG
A Retrieval-Augmented Generation tool that lets you upload PDFs, Word docs, PowerPoints, Excel sheets, or paste a YouTube/website link, then ask natural-language questions and get context-aware answers.
RAGify was built to remove the friction of manually digging through long documents, videos, or web pages to find an answer. Instead of reading everything, a user can drop in a file or a link and simply ask what they want to know.
- Information is scattered across many formats — PDFs, slide decks, spreadsheets, videos, and web pages — each requiring a different tool to search through.
- Manually skimming long-form content to answer a specific question is slow and error-prone.
- Most people don't want to set up a custom AI pipeline just to query their own files.
A Streamlit front end handles uploads and chat, while a backend pipeline chunks, embeds, and retrieves content before generating an answer.
User uploads a document/link via the Streamlit UI
Content is parsed and split into text chunks
Chunks are embedded and stored in a vector index
User question is embedded and matched against the index
Relevant chunks + question are sent to the LLM
LLM returns a context-grounded answer to the user
Key Architectural Decisions
- ✓Built a unified ingestion layer that extracts text from PDFs, DOCX, PPTX, XLSX, YouTube transcripts, and scraped web pages.
- ✓Used LangChain to chunk and embed the extracted content into a searchable vector store.
- ✓Implemented a retrieval pipeline that finds the most relevant chunks and passes them to an LLM for a grounded, context-aware answer.
- ✓Wrapped the whole flow in a simple Streamlit interface so non-technical users can just upload and ask.
- Handling multiple file formats cleanly requires a consistent internal text representation before chunking.
- Chunk size and overlap have a big impact on retrieval quality.
- Streamlit is a fast way to validate an AI product idea before investing in a full front end.