neonwatty/meme-search v2.2.0

Release v2.2.0

on GitHub

Highlights

• Added clipboard image paste support on the upload screen, including validation, generated filenames, and end-to-end coverage.
• Normalized image inputs before captioning so transparent, palette, and grayscale images are converted consistently for the vision model.
• Fixed Docker Compose database URLs when a custom host DB_PORT is set.
• Hardened Compose persistence paths for uploads, SQLite data, models, and generator-side storage.
• Refreshed the multi-arch Docker images for AMD64 and ARM64.

Contributor Thanks

Thanks to the issue reporters and reviewers who helped pin down the upload, captioning, and Docker Compose edge cases covered by this release.

Docker Images

Multi-arch images are available for Linux AMD64 and ARM64:

• ghcr.io/neonwatty/meme_search:v2.2.0
• ghcr.io/neonwatty/image_to_text_generator:v2.2.0

The latest tags have also been refreshed.

What's Changed

• chore: run GitHub Actions on Node 24 by @neonwatty in #160
• Normalize image inputs before captioning by @neonwatty in #161
• Fix Compose internal database URL by @neonwatty in #162
• Harden Compose persistence paths by @neonwatty in #164
• Add clipboard image upload support by @neonwatty in #163
• Prepare v2.2.0 release by @neonwatty in #165

Full Changelog: v2.1.0...v2.2.0