You know that moment: you took a screenshot of a competitor's pricing page three months ago, saved it somewhere sensible, and now it is completely gone. Not deleted. Just un-findable. Spotlight returns nothing. File names are useless. You give up and redo the work.
DigUp fixes that. It is a free, open-source desktop app that lets you search local files by describing what is IN them, not what you named them. Type "invoice from the plumbing company" and it finds the right PDF. Type "photo of the store before the renovation" and it surfaces the right image.
What's happening
This week DigUp got significantly smarter. The project just integrated Google DeepMind's EmbeddingGemma 2, a 740-million-parameter open multimodal embedding model designed specifically for on-device search. That model understands text, images, audio, and video in one unified search space, which means DigUp can now find meaning inside screenshots and podcast recordings, not just text files and PDFs.
Everything runs on your machine. No account, no subscription, no telemetry. The first-run model download is 865 MB, and after that your files stay yours. The DigUp project itself is MIT-licensed, maintained by indie developer Abdur Rahim, and open for anyone to inspect or build on.
Current stable builds are for Apple Silicon Macs running macOS 14 or later. Windows and Linux beta builds are on the releases page now, with rough edges.
Try this this week
If you are on an Apple Silicon Mac:
- Go to github.com/ARahim3/DigUp/releases and download the latest release
- Install it and point it at a folder you actually use: Documents, a client work folder, or your Downloads pile
- Let it index on first run (do it overnight, the embedding pass takes time on large folders)
- Try a search that would stump Spotlight: "quote PDF from the roofing company" or "that screenshot of the booking form"
- Add your Screenshots folder too, especially if you screenshot a lot for reference
On Windows or Linux, grab the beta from the same releases page. It works, just expect the occasional hiccup.
The bigger picture
A year ago, semantic search over your local files meant either a paid app or a cloud service that ingested your documents. Now an indie developer wrapped a free open model into a menu bar app, and the whole thing runs on hardware most of us already own. No per-seat pricing, no vendor lock-in, no privacy trade-off. That is the pattern worth watching, and it keeps moving in one direction: capable AI as a local utility. The smaller your team, the bigger the relative win, because you get the same technology without the subscription that larger orgs absorb without blinking.