4.4.2

This is a maintenance upgrade that adds safe in-app updating for legacy corpus databases, improves startup resource handling, and fixes several corpus, search, and interface issues.

New Features

• Recognized legacy raw-file and Word List corpus databases can now be updated to the current format from within AntConc instead of requiring a complete rebuild.
• Legacy database updates run in the background with progress reporting. AntConc creates and retains a numbered backup before making changes, validates the migrated database, and automatically loads a successfully updated corpus in its requested target or reference role.
• Raw legacy corpus updates now generate lexical token positions and build the indexes required by the current database format. Existing Word List data is preserved when upgrading older Word List corpora.

Improvements

• Frozen-app startup is faster and performs less unnecessary file work. Bundled demo corpora are now synchronized individually using per-corpus versions, so unchanged corpora are not recopied at every startup.
• AI resource initialization now uses manifest versions and no longer recursively copies or scans downloaded model files at startup. User-downloaded models and newer online catalogues are preserved.
• OpenAI, Gemini, and Claude model-list caches are now stored in Application Support. Packaged defaults are seeded only when a cache is missing, while lists explicitly updated by the user are retained.
• Quick Corpora now receive sequential names such as `quick_corpus_1.db`, using the next number above the highest existing Quick Corpus number.
• Newly created Quick Corpora can now be assigned immediately as the target or reference corpus even when the Corpus Manager has not previously been opened.
• Download processing chunks have been increased from 1 KB to 1 MB, reducing overhead during large AI model downloads.
• Whisper command construction and bundled binary permission handling have been improved for cross-platform builds.
• The Linux Flatpak runtime has been updated to version 6.10.

Bug Fixes

• Plot advanced searches containing multiple alternative terms now combine all alternatives before calculating document statistics. Overlapping hits are deduplicated while preserving the existing overlay behavior.
• Changing a KWIC sort preset, including switching from right sorting to center sorting, no longer triggers duplicate searches and repeated “Wait for the previous action to finish” warnings.
• The metadata-filter display now follows native light and dark themes correctly on macOS without showing an unwanted underline.
• Corpus indexing no longer attempts to create indexes for columns that are absent from legacy databases. Legacy upgrades now create all applicable current-format indexes before marking the corpus as indexed.
• Malformed, unrecognized, and newer unsupported corpus database versions are handled more safely and are not offered an inappropriate automatic update.
• Development and smoke-test entry points now consistently use the installed Alice demo corpus from the correct per-user Application Support location instead of creating or opening stray `demo.db` or `test.db` files.