Document demo
Retrieval is the first half of RAG: split a document into chunks, turn each chunk into a vector, and find the ones closest in meaning to a question. This page runs all of it in your browser. Files never leave your machine.
1 Add a document
Plain text, Markdown, CSV or PDF, up to 10 MB. Only the first 60,000 characters are used.
Embedding model: waiting
2 See how it gets chunked
A chunk is the unit that gets embedded and retrieved. Change the strategy and the chunks redraw instantly.
Document map (two lanes so overlaps show; click a chunk)
Similarity between each sentence and the next (dips below the dashed line become cuts)
Document text, coloured by chunk
3 Ask for a chunk
Ask a question in your own words. It gets embedded the same way as the chunks and compared by cosine similarity.
4 Retrieved chunks
Top three matches. Click a result to find it in the document above.
Results appear here after you ask a question.