Changelog

All notable changes to this project will be documented here.

[Unreleased]

Style-rule edits no longer invalidate the TM cache — style_rules is dropped from the cache key, so editing ai_style_rules.txt keeps the warmed cache intact instead of re-translating every segment. Style rules still drive translation and the QA pass on a cache miss. Trade-off: already-cached segments keep being served under the old rules until you clear the cache manually (DELETE FROM translation_memory). (OMP-032)

Batch pre-translation — on file open, a popup offers to pre-translate all untranslated segments in the background. Each segment is processed by the same orchestrator (glossary, style rules, QA, TM cache) as the live MT pane, so the MT pane responds instantly from cache when you navigate to a pre-translated segment. Non-destructive: nothing is inserted into OmegaT's translation fields. (OMP-025, OMP-026, OMP-027)
/batch-translate service endpoint — accepts a list of fully-contexted TranslateRequest objects, processes each via translate_segment(), and returns per-segment results with per-segment error isolation so one failure does not abort the batch. (OMP-025)
QA_ENABLED / QA_MODEL settings — opt-in QA self-critique pass that checks each new translation against the approved glossary and style rules, auto-corrects violations, and logs each fix to the service log and OmegaT's own log. Off by default. (OMP-023, OMP-024)
TM_CACHE_ENABLED setting — server-side translation memory cache (on by default). Set false to force a fresh LLM call per segment. (OMP-022)