Skip to content

v3.0.10 - 2026-06-27

Latest
Latest

Choose a tag to compare

View all tags
@ASK-03 ASK-03 released this 27 Jun 18:35
v3.0.10
0fd819c
This commit was signed with the committer’s verified signature.
ASK-03 Abhishek Singh Kushwaha
GPG key ID: F3781DB37D8A07D6
Verified
Learn about vigilant mode.

🔁 v3.0.10 - 2026-06-27

tl;dr

  • We have revamped the UI using electron
  • Performance enhancements using indexes in METADATA db and using onxx runtime
  • Functionality to give names to detected faces
  • Added Search with text on images with reranking (You can search "Abhishek trekking" and it will only show images for Abhishek on top of list)
  • Changes in github workflow

SnapSort continues to evolve as your go-to desktop application for offline, AI-powered image sorting and semantic searching. With a complete architectural overhaul and significant feature upgrades, v3.0.0 transitions SnapSort into a professional, standalone desktop application. No more Python prerequisites or manual installations! Simply download the installer for your platform, select a folder of photos, and SnapSort will detect faces, index semantic embeddings, and allow you to:

  • Search for photos by typing natural descriptions (e.g., "tree in sunset", "group of people outdoors") using advanced semantic search.
  • Search for photos by typing a specific tagged person's name combined with context (e.g., "Rahul hiking").
  • Enjoy a true "double-click to run" experience on Windows, macOS, and Linux with our new seamlessly bundled AI engine.

Changes

🧠 Functionality

  • Semantic Image Search (CLIP): A powerful new feature utilizing Xenova CLIP to allow natural language text-to-image searching, enabling you to find photos by simply describing what's in them.
  • Standalone AI Engine: The Python backend (FastAPI, YuNet, SFace, CLIP) is now compiled into a highly optimized, fully bundled executable using PyInstaller. You no longer need to install Python or use pip!
  • Persistent Vector Storage: Upgraded to FAISS for lightning-fast high-dimensional vector similarity search for both face features and semantic embeddings.

🎨 User Interface & Experience

  • Electron + React Frontend: A completely rebuilt, blazing fast desktop interface using React (TypeScript) and Zustand for state management, seamlessly wrapped in an Electron shell.
  • About & Settings Polish: Added the official SnapSort logo to the About page and a convenient "Report a Bug" link in the settings menu that directs straight to GitHub issues.
  • Automated Installers: One-click installers (.exe, .dmg, .AppImage) are now automatically generated via GitHub Actions and electron-builder.

🛠️ Codebase Improvements

  • Multi-Worker Processing: Completely revamped the backend controller to use a multi-processing pool for background indexing, achieving ~2.5 - 3 images per second on standard CPUs without blocking the UI.
  • AI Agent Context: Added an AGENTS.md file to act as canonical context documentation, allowing AI coding assistants to seamlessly contribute to future development.
  • Benchmarking Tools: Included a new benchmark.py script to easily test and measure local hardware performance on the heavy machine learning pipeline.
Assets 10
Loading