DigUp: local semantic search on Mac powered by Google's EmbeddingGemma 2
DigUp uses Google's EmbeddingGemma 2 to search Mac files, PDFs, screenshots, photos, podcasts and videos by meaning rather than exact keywords, running locally.
DigUp is a search app for macOS built on Google's EmbeddingGemma 2 model. The project is available on GitHub: DigUp.
The app can search not only file names but also their contents: PDFs, screenshots, photos, podcasts and videos.
The capability comes from embedding-based search. Users do not need to know exact words; they can describe what they are looking for in broad terms, and the AI model finds the closest match. For example, a query like where they talk about sleep can jump directly to the relevant moment in a podcast.
The model itself is small, open source and runs locally on the Mac, so files are not sent to external servers.
This kind of search could be implemented in almost any context where local search is used today.
Why it matters
AnalysisLocal embedding search turns documents and media into semantically indexed content, improving discoverability without sending private files to the cloud.
Discuss in community
Share your questions and insights with developers