Open-source Mac app DigUp brings on-device multimodal semantic search via EmbeddingGemma 2
DigUp, an open-source macOS menu bar app, has shot up GitHub's trending list for its fully local, cross-modal content search, and was reshared and recommended by Google's official Gemma account. Built on Google DeepMind's multimodal embedding model EmbeddingGemma 2, it lets users find photos, PDFs, audio recordings, videos, and code with natural language descriptions of the content — even if they've completely forgotten the file name.
Confirmed
- DigUp was open-sourced by developer Abdur Rahim; it's a macOS menu bar search tool that can search photos, documents, code, and audio/video files within selected folders.
- The underlying model is Google DeepMind's EmbeddingGemma 2 (740M parameters, Apache 2.0 license), which encodes text, images, audio, and video into unified 768-dimensional vectors for local cross-modal semantic search.
- Everything runs on-device with no cloud involved; unlike macOS Spotlight's filename-based search, DigUp matches by content semantics.
- The project hit GitHub trending; on star counts, different posts cite 3k (m4) and 2.6k (m5), both as of posting time.
- The project itself is open-sourced under the MIT license.
Why it matters
- It demonstrates a practical deployment of a small multimodal embedding model (EmbeddingGemma 2) on consumer devices: semantic search no longer depends on the cloud, balancing privacy and convenience.
- Unified cross-modal vector representations mean you can 'describe what you remember in one sentence' and retrieve years-old recordings, photos, or documents — a representative case of the local AI tooling ecosystem.
2026-10-14 ~ 2026-10-14 · 5 related posts
Primary sources
- DigUp: open-source Mac app brings 100% local semantic search via EmbeddingGemma — TheMoonMidas · 2026-10-14
- DigUp: Open-Source On-Device Mac Semantic Search Built on EmbeddingGemma 2 — aigclink · 2026-10-14
- [source] Inside DigUp: How a 740M-Param Multimodal Embedding Model Enables Local Semantic File Search on Mac — xiaohu · 2026-10-14