DigUp: search local files by their meaning, page or moment
DigUp is a free, open-source desktop search application for finding what is inside your files. Its official repository describes local search across pictures, documents, recordings and videos, using Google's EmbeddingGemma 2 model through llama.cpp. Mac is supported, with Windows and Linux builds labelled beta.
Find the part of the file you remember
You can describe a scene, a spoken topic or a passage in a document. Results can point to a PDF page or a moment in a recording, rather than just returning the filename. Exact strings such as invoice numbers and error codes also contribute to matching.
Opening at the matching moment depends on the external viewer. The README names supported media players and PDF viewers, and explains that other applications may open at the beginning. DigUp's own preview can still show the matching moment.
Choose the folders and indexing scope
The project says indexed files stay on your computer, with network access used for the initial model download and update checks. Users select the folders to include. The README also describes exclusions for sensitive files, hidden folders, caches and cloud-only files that have not been downloaded.
Code search is a separate, opt-in index. With that enabled for selected repositories, a query prefixed with code: can describe a function or name a symbol and open a result in an editor at the relevant line. It is not enabled automatically for every source folder on the machine.
Allow time for the first pass
The model download is approximately 865MB. Initial indexing can be substantial, particularly for a large media library on a CPU. The project says later passes read new files rather than rebuilding everything.
Platform differences matter. Text recognition inside screenshots is currently available on Mac, while Windows and Linux can search screenshots by visual content. The unsigned Windows installer and beta status also deserve attention before adopting it for a work environment.
Try a small folder containing files you can identify manually. Check a precise string, a descriptive query and a timestamped result. That will tell you more about its fit than starting with an entire archive and waiting for a large first index.