Doc Search · private in-browser document search

Search your documents without uploading them

1
Add documents
PDFs, scanned pages, photos of receipts, or text files.
2
They are read here
Text is extracted, and scans are read with OCR, inside this tab.
3
Ask questions
In English, Malay or Chinese, even across languages.

Add your own files

Add documents Drop files here or tap to choose · PDF, images, text Read scans and photos in

Options

New documents go to

Paste a page of text

How it works

Everything runs in this tab: a 33 MB search model (keyword search plus a small multilingual embedder), PDF text extraction and OCR. Your files are never uploaded, and searching makes no network requests.
Documents are kept in one library per language, because the search decides how to combine keyword and meaning matches per language. A question is searched in every library; the one in the question's language is shown first.
The index is saved in this browser, so reopening the page does not read your files again. Clear all removes it.
Research candidate, not a product. What has and has not been measured: research/own_retriever_v4/browser_pack/README.md in Whyme-Labs/doc-exp. Sample documents are fictional and were written by an AI assistant.