FAQ

A search app for the files on your computer. It runs Google DeepMind's EmbeddingGemma 2, a model that puts text, pictures, sound and video in one shared space, on your own machine. You describe what you remember, and it finds the photo, the PDF page, the minute in a recording or the moment in a video. It's free and open source, on a Mac, on Windows and on Linux.

They find files by their names, text and metadata. DigUp also finds them by what they show or say, and takes you to the page or moment that matched. They work fine side by side.

No. DigUp is a search: you describe what you remember and it finds the file and the place in it. No language model writes answers, and nothing is sent anywhere.

Yes. DigUp is open source under the MIT license, with no account, subscription or paid tier. If it helps you, you can support it, but you don't have to.

From this site, the GitHub repository and the Homebrew tap arahim3/tap. Every download comes from GitHub Releases. Other sites offering DigUp aren't the project's; don't download from them.

No. DigUp is built around EmbeddingGemma 2 because it puts text, pictures, sound and video in one space, which is what lets one search rank a video frame against a PDF page. Everything in the index is tied to that model's vectors.

Not yet. An option to use your own embedding server is planned. Until then, the model runs on the same computer as DigUp.

Nothing you index leaves your computer. DigUp goes online for two things: to download the model once, from Hugging Face, and to check for updates once a day, on GitHub. The update check sends nothing about your files, and you can turn it off in Settings → General. No account, no telemetry. Privacy and security has the details.

Not by DigUp. The index is one file in your user account's folders, which other users on the computer can't open, and on a Mac with FileVault it's on the encrypted disk like everything else. Anyone who can read it as you can also read the files it came from, but it is one copy of what's in them, so keep that in mind on a shared computer.

Keys, certificates and similar files aren't among the kinds DigUp reads, and hidden folders like .ssh are skipped. In code folders, files named like keys or password stores, and any file holding a private key, are skipped too. A plain text document is read like any other document, so keep notes with secrets in a folder you don't choose, or skip that folder.

Yes. Don't choose it, or skip it in Settings → Folders with Skip a Folder… or Skip these file types. Skipping or removing a folder takes its files out of the index right away.

That's the model: the text part (310 MB) and the image and audio encoders (555 MB). It downloads once. The app itself is about 20 MB on a Mac, and 40 to 140 MB on Windows and Linux, depending on the package, since those carry their own PDF, video and image readers.

Only the first pass is heavy, and it happens once. After that DigUp reads only what's new or changed. The model runs in a separate helper that quits when it's done, and the search side only loads while you search. On a Mac, Low Power Mode and a hot Mac make the first pass wait, and Index only at set hours can keep it to the night.

On a Mac, a few minutes for a few hundred files, and about an hour for a big Downloads folder. On Windows and Linux it depends on whether the model runs on the graphics card: on a processor alone a picture takes a few seconds (about five on a 2017 laptop's Core i5), so a large photo library can take hours. Setup and Settings show the estimate for each folder, measured on your computer.

It depends on what your files are more than on how much space they take. Measured on the project's test files: a photo adds about 1.6 KB, a screenshot about 3 KB with its text, a PDF page about 6 KB, an hour of audio about 0.25 MB, an hour of video 1 to 2 MB, and documents 2 to 3 times the size of their text. So 100,000 photos come to about 160 MB, and text is what adds up: about 2 MB for a 300-page book. The model is 865 MB on top.

No. After the first pass DigUp doesn't rescan: the system tells it which folders changed, and it reads only what's new or changed. Reading doesn't wear an SSD; writing does, and after the first pass that's a few KB for each new file.

On a Mac the app idles at about 12 to 30 MB. While you search, a helper holds the text part of the model, about 250 MB, and quits after ten minutes without a search. During the first pass the indexing helper uses up to about 2 GB on a Mac, and quits when it's done.

Pictures and screenshots, PDFs (scans too), documents (txt, md, typ, rtf, docx, odt, html, and on a Mac doc), spreadsheets and slides (xlsx, ods, pptx, odp), audio and video of any length, and code in the folders you pick for code search. Pages, Keynote, Numbers and EPUB files aren't read. What DigUp reads has the full list.

The model knows more than 100 languages, and DigUp searches in all of them, across languages too: a search in English can find a note in Bengali or Arabic. DigUp's tests cover English, Bengali and Arabic because those are the languages of the files it was tested on, and Google notes that the model isn't equally strong in every language. Two limits apply to exact words, not to meaning: the text inside screenshots is read only on a Mac, by Apple's text recognition (about 25 languages), and languages written without spaces, like Chinese, Japanese and Thai, aren't split into words.

Yes. Every kept frame (about one per shot) and every 30 seconds of sound has its own place in the index, with its time. A video's score is its best moment, which is also where it opens, so one scene that fits isn't averaged away by the rest.

Not yet. A search covers all the folders you chose; the search window filters results by kind.

Not read yet: the pictures sit inside the library package, which DigUp doesn't open. Photos exported to a folder are read like any other pictures.

No. Mail apps keep messages in their own stores, which DigUp doesn't open.

No. DigUp needs an Apple Silicon Mac (M1 or later) with macOS 14 or later.

The search is the same: the same model and the same results. Two things are missing for now: the text inside screenshots isn't read yet, so a screenshot is found by what it shows rather than by an exact code typed into the search; and old Word .doc files, RAW photos and SVG pictures are read only on a Mac. Both are betas, so expect some rough edges, and please report them. The download page compares the three.

On a Mac, always, with Metal. On Windows and Linux, a quick check at setup compares the graphics card with the processor and uses the card only when it's faster and gives the same results. The last line of speed.log says which one DigUp picked: "device": "gpu" or "cpu". It's in %LOCALAPPDATA%\com.abdurrahim.DigUp\logs on Windows and ~/.local/share/com.abdurrahim.DigUp/logs on Linux.

Not yet. The Windows and Linux packages are for 64-bit Intel and AMD processors.

The project builds the .deb and the AppImage, and the AppImage runs on most distributions. Packages made by others aren't checked by the project.

Uninstall has the steps for each system, including where the model and the index are.