Skip to content

Releases: M4yankkkk/sift

Release list

Sift 0.1.0

Sift 0.1.0 Latest
Latest

Choose a tag to compare

View all tags
@M4yankkkk M4yankkkk released this 16 Oct 10:54
v0.1.0
616a0bc

Find files on Linux by what's in them. Sift searches photos, screenshots, PDFs, notes, and code by meaning, exact text, and the text inside images. It runs fully on your machine, with Google DeepMind's EmbeddingGemma 2 in a single Rust process.

Install (Ubuntu / Debian)

sudo apt install ./Sift_0.1.0_amd64.deb

apt installs the dependencies too (WebKitGTK, Tesseract OCR, poppler). Then open Sift from your app menu, or run sift. The default hotkey is Ctrl+Alt+Space.

First launch

  • The window opens right away. Filename, keyword, and text-in-image search work immediately.
  • Sift downloads the search model once (about 435 MB), with live progress in the status bar. If the connection drops it resumes, and if you're offline it says so and offers Retry.
  • When the download finishes, meaning search turns on and your files are embedded in the background. After that, Sift doesn't use the network.
  • Add the folders to index from the footer (Folders, or Ctrl+Shift+F).

What's in 0.1.0

  • Search by meaning: sunset finds a sunset photo whatever its filename is.
  • Text and photos share one model, so a text query is compared directly against images.
  • OCR makes error text inside screenshots searchable.
  • Exact keyword search (BM25), with PDFs opening at the matching page.
  • Query filters: sunset photo, resume pdf, type:pdf invoice, ext:png logo.
  • Visual mode shows a photo grid. The footer shows live memory and CPU use.

Requirements and limits

  • 64-bit Ubuntu / Debian-based Linux, developed on Ubuntu with GNOME. X11 and Wayland are supported.
  • Runs on the CPU. Indexing a large photo library takes a while (about 0.6 s per image).
  • No audio or video search yet.

Bug reports and bad-search-result reports are welcome. The issue templates describe what to include.

Assets 3
Loading