Skip to content

Releases: newdee/magpie

magpie v0.1.9

Choose a tag to compare

@github-actions github-actions released this 21 Aug 15:48
v0.1.9
a14cfae
v0.1.9

magpie v0.1.7

Choose a tag to compare

@github-actions github-actions released this 21 Aug 13:34
v0.1.7
14f9a06

First public release. magpie is a Spotlight-style desktop launcher that searches everything you saved and forgot — locally, privately.

Highlights

  • Three sources (Tab to switch): explicitly-added local folders, your GitHub stars (full READMEs, chunk-embedded), and browser bookmarks (any Chromium-based browser + Firefox, auto-discovered).
  • Hybrid search: SQLite FTS5 keyword + local multilingual semantic embeddings, cross-language (query in Chinese, hit English docs, and vice versa). Highlighted context snippets.
  • Image search via SigLIP 2: text-to-image in any language, and image-to-image by drag / paste / file picker, with similarity percentages and thumbnails.
  • Full-text indexing: ~80 text/code formats, PDF (pdf-inspector), Word/Excel/PowerPoint; every other file findable by name.
  • Privacy by architecture: no full-disk scanning — only folders you add; everything stays on your machine; models run locally via ONNX.

Install

  • Windows: magpie_0.1.7_x64-setup.exe
  • macOS (Apple Silicon): curl -L https://github.com/newdee/magpie/releases/download/v0.1.7/magpie_aarch64.app.tar.gz | tar xz -C /Applications (skips Gatekeeper quarantine; builds are unsigned)
  • Linux: AppImage / deb / rpm

First launch downloads the embedding models (~700 MB); keyword search works immediately while they warm up. In regions where huggingface.co is unreachable, switch the download source to hf-mirror.com in Settings (tray icon → Settings…).

magpie v0.1.0

Choose a tag to compare

@github-actions github-actions released this 21 Aug 04:10
v0.1.0
6730f20

Spotlight-style local search for everything you saved and forgot: GitHub stars, local folders (full text incl. PDF/Word/Excel), and images (search by description or by similar image). Privacy-first: only folders you add are indexed, everything runs and stays on your machine.

First launch downloads the embedding models (~700 MB); keyword search works immediately while they warm up.

Downloads

  • Windows: magpie_0.1.0_x64-setup.exe (unsigned: SmartScreen will warn, choose "Run anyway")
  • macOS (Apple Silicon): magpie_0.1.0_aarch64.dmg (unsigned: right-click the app, then Open; if it says "damaged", run xattr -cr /Applications/magpie.app)
  • Linux: AppImage / deb / rpm (glibc 2.39+, e.g. Ubuntu 24.04)

macOS and Linux builds are CI-built and lightly tested; feedback welcome.