Ask questions about your PDFs, on your own computer

In four steps

1. 1

   Open Documents and click “Add a folder”, or pick one under “Add quickly” (Downloads, Desktop, Documents…).

2. 2

   Twoody indexes its files on your computer: PDF, Word, Excel, PowerPoint, OpenDocument, RTF, text and HTML.

3. 3

   Ask your question in the chat: when it clearly speaks of your documents, Twoody searches them before answering.

4. 4

   Click one of the numbered “Sources” to open the file at the page, the slide or the rows it cites.

Scanned PDFs

On a Mac, a scanned PDF is read by macOS character recognition (OCR), on the Mac, up to 300 pages. Scans are read on a Mac only: on Windows and Linux, Twoody reads the PDFs that contain text.

Where your documents are read

On your computer, by the model installed on it, not by an online AI service: Twoody reads the files where they are, without copying them elsewhere. In a conversation kept on your optional account, the passages it read are kept with the answer, encrypted end to end; a private conversation is kept on the computer only. For a client's file, a walled-off project keeps its documents, conversations and notes apart.

Search by words, or by meaning

Twoody finds your documents by their words and by the names of their files and folders, short words included. “Search by meaning”, in Documents, adds a search on what a passage says: a dedicated model, downloaded once, builds that index on your computer, and the text of your documents is not sent.

Answers from your documents only

Under the message, “Tools” › Documents lets you choose all authorized folders, some of them, or none. “Documents only” answers from your documents, with their sources, and turns the internet off.

What to expect

A model running on your computer can be wrong, even when it cites its sources: check what matters. Twoody reads up to 20,000 files per folder; the old .doc, .xls and .ppt formats are not read yet.