DigUp is a free macOS menu-bar app that searches the files on a Mac by what they show or say, using Google DeepMind's EmbeddingGemma 2, a 740M-parameter multimodal embedding model that the README says covers more than 100 languages. The user picks folders, and DigUp indexes pictures and screenshots, PDFs page by page (scans included, up to 5,000 pages a file), documents such as txt, md, rtf, doc, docx, odt and html, audio, and video. Audio is read in overlapping 30-second windows and video as a frame every few seconds plus its soundtrack, so a match can open a clip at the right moment or a PDF at the right page. Every item is stored as a vector in one 768-dimensional space inside a single SQLite file, next to a keyword index, and exact-word matches get a boost so invoice numbers and error codes still match as written. The model runs as an 8-bit build through llama.cpp on the Mac's GPU in a helper process that quits when indexing is done. Code search is off by default and is turned on per folder; it uses its own index, and indexing all of llama.cpp is quoted at about 20 minutes and 130 MB. The README says nothing indexed leaves the Mac, with no account or telemetry, and that the app goes online only to download the model once (865 MB) and to check for updates daily. It requires Apple Silicon and macOS 14 or later, opens with Shift-Command-Space, and installs from a DMG or Homebrew; Windows and Linux versions are listed as on the way. The evals cover only English, Bengali and Arabic, and Google notes the model is not equally strong in every language. Written in Swift and released under the MIT license.