Drop or choose files
Files are read into memory through the browser File API and are never sent to a server.
SecureRAG Cross-format file indexing.
PDF, DOCX, TXT, Markdown, CSV, HTML and JSON. Up to 25 MB per file and 40 files per library. The first visit downloads a ≈ 25 MB embedding model; after that it works offline.
- Files never leave the device
- No account, no tracking pixels
- Not used for training
- Works offline once cached
How to use it
Three steps to a cited answer
Wait for the local index
Parsing, chunking and embedding run in a Web Worker with live progress.
Ask, then verify
Every answer carries numbered citations that jump to the source paragraph and page.
FAQ
Are my files really not uploaded?
Correct. There is no backend endpoint and no upload code path in the page. Open DevTools → Network, add a document and ask a question: apart from the one-time model download you will see no requests at all.
Does my library survive closing the tab?
Chunks and vectors live in your browser IndexedDB, so the next visit is instant. Clearing site data deletes everything.
Which formats work?
PDF with a text layer, DOCX, TXT, Markdown, CSV, HTML and JSON. Encrypted PDFs, scans without a text layer and legacy .doc are refused with a specific reason.