chore: genericize inbox path examples in docs and tests

chore: drop personal default selected-tags lists from tooling

chore: remove personal paths and album names from tooling

docs: add dual MIT/CC BY licensing and model attribution

docs: add GitHub model weight links and ignore large binaries

refactor: rebrand project from Thresh3r to THR3SHR

feat: add GPU model unload and improve classify cancel

feat: implement SFW classify mode and NSFW tag management

chore: restore taxonomy formatting after SFW veto expansion

feat: expand SFW veto for acts, Voyeur, and specialty cues

fix: slim SFW evidence and veto favourite-folder signals

fix: drop school_uniform from SFW evidence

feat: tighten strict-safe SFW fallback catch-all

Document strict-safe SFW catch-all design.

feat: include group_sex in Doujin favourite folders

feat: expand milf and favourite-folder evidence

feat: add group_sex bucket and veto-based precedence

test: add drawn group_sex precedence matrix

Add implementation plan for group_sex taxonomy refresh.

Document drawn taxonomy group_sex and evidence refresh design.

Add isolated real-life adult tagging domain with local VLM pipeline.

feat: add Doujin works classify/review/migrate mode

feat: route pregnancy_test into fertilization

fix: stop items list from reading full_scores_json blobs

fix: SQLite WAL/busy timeout and allow migrate on failed runs

feat: ML-Danbooru primary runs only route to loli

fix: keep classified items when a run is cancelled

feat: hybrid ML allowlist rescue on WD needs-review

feat: taxonomy catch-alls, scan/migrate fixes, and tag recall/FP benchmarks

fix: stop discarding the ONNX provider probe on every settings save

feat: expand taxonomy funnel with act, fallback, and Voyeur sub-folders

refactor: rename project to Thresh3r and update related documentation

feat: robust GIF/video presence classification pipeline

feat: expand Voyeur soft cues for condom, masturbation, and gestures

feat: add nested Voyeur soft taxonomy with theme precedence

feat: improve taxonomy routing and restore scaled media sampling

feat: hybrid real-life filter, review/migrate fixes, and faster media runs

feat: add experimental style gate and refresh taxonomy priorities

Add Classifier debug real-life vs anime remote eval.

Enhance media preview functionality in backend API. Introduce caching for JPEG stills from GIFs and videos to improve performance under load. Update preview endpoint to support raw media serving and adjust response handling for inline display. Modify frontend to reflect changes in media preview handling and improve UI for video indicators. Add tests to ensure proper functionality for GIF and MP4 previews.

Add real_life taxonomy bucket backed by WD realism probe.

Fix GIF/video preview and classify with length-scaled multi-frame sampling.

Speed up WD inference with real batching and harden reclassify cancel/retry.

Refactor message formatting in backend API for run status updates. Update message strings in start_run and reclassify_run functions to use f-strings for improved readability and consistency.

Enhance backend API and frontend components for reclassification and debugging. Introduce new request and response schemas for reclassification, update inference settings, and improve handling of inference modes. Add a debug interface for SFW evaluation and integrate reclassification functionality into the frontend, allowing users to retry with different models. Update styles for better UI presentation and ensure proper environment variable handling during startup.

Implement destination folder validation in migrate_run function. Add checks to ensure the destination folder exists and is a directory, with appropriate error handling and logging for failed migrations. This enhances the robustness of the migration process by preventing invalid paths from causing issues.

Update taxonomy.json to expand categories and improve evidence handling. Introduce new categories for 'NTR', 'incest', 'nakadashi', and 'monster_girl', along with updated aliases and ignore lists. Adjust priorities for existing categories and enhance tests in test_taxonomy.py to validate new folder routes and evidence logic.

Refactor tagging logic in backend API and services. Replace choose_best_tags with choose_best_destination for improved tag selection. Integrate taxonomy resolution for folder names in save_settings and discover_tag_folders functions, enhancing tag mapping accuracy. Update relevant imports and ensure consistency in tag handling.

Enhance run management in backend API to support cancellation and directory validation. Implement checks for root and categories directories, ensuring proper error handling for missing or invalid paths. Update run status handling to allow immediate cancellation during execution. Add tests to validate new behavior for run initiation and directory checks.

Document multi-model tagging, assignment rules, and updated UI workflow.