supervertaler 1.9.332__tar.gz → 1.9.334__tar.gz

{supervertaler-1.9.332 → supervertaler-1.9.334}/CHANGELOG.md

@@ -2,9 +2,55 @@
 
 All notable changes to Supervertaler are documented in this file.
 
-**Current Version:** v1.9.332 (February 26, 2026)
+**Current Version:** v1.9.334 (February 27, 2026)
 
 
+## v1.9.334 - February 27, 2026
+
+### New Feature: SDLXLIFF Comment Round-Trip
+
+Supervertaler now fully supports **Trados Studio comments** in SDLXLIFF files — both import and export.
+
+- **Comment import** — Existing Trados Studio comments (`<cmt-def>` / `<mrk mtype="x-sdl-comment">`) are parsed during SDLXLIFF import and displayed in the Notes panel, prefixed with the original author name (e.g. `[mbeijer] comment text`).
+- **Comment export** — Any text entered in the Notes field is written back as a standard Trados Studio comment when exporting to SDLXLIFF or SDLRPX. Comments appear correctly in Trados Studio's Comments pane with proper author attribution, timestamp, and severity metadata.
+- **Full round-trip** — Import → edit/add comments → export → re-import preserves all comments. Double round-trips also work correctly.
+
+### New Feature: User Identity Settings
+
+A new **👤 User Identity** tab in Settings lets you configure the translator name that appears in exported files.
+
+- **Translator name** — Set your preferred name or alias. This is used as the author in SDLXLIFF comments, the `PackageCreatedBy` field in Trados return packages (SDLRPX), and the default creator ID when creating new TMX files.
+- **Privacy-friendly fallback** — If left empty, falls back to your system username. The app name is never stamped into exported files.
+
+### Improvements
+
+- **SDLXLIFF segment ID robustness** — Segment IDs for SDLXLIFF round-trip export are now stored in a dedicated internal field (`sdl_segment_id`) instead of being embedded in user-editable notes. This prevents accidental breakage if users edit or clear the notes field. Old projects with metadata-in-notes are still supported via a fallback parser.
+
+---
+
+## v1.9.333 - February 27, 2026
+
+### New Feature: Okapi Framework Integration (Experimental)
+
+Supervertaler now ships with an **Okapi Framework sidecar** — an industry-standard, open-source file filter engine used by professional translation tools worldwide. This lays the foundation for significantly improved document import fidelity and support for additional file formats in future releases.
+
+- **Okapi sidecar architecture** — A lightweight Java-based REST service (`okapi-sidecar.jar`) runs locally on `127.0.0.1:8090` as a background process alongside Supervertaler. All document processing happens on the user's machine — no files are sent over the internet. The sidecar starts automatically (with a 1.5s deferred launch) and shuts down cleanly when Supervertaler exits.
+- **Bundled minimal JRE** — The sidecar includes a custom 44 MB Java runtime (built with `jlink`) so users do not need to install Java separately. PyInstaller specs are updated to include the sidecar JAR + JRE in Windows and macOS builds.
+- **Optional DOCX import engine** — When importing a DOCX, users are now presented with a choice dialog:
+  - **Standard (recommended)** — the existing built-in python-docx handler with full formatting tag support
+  - **Okapi Framework (experimental)** — industrial-strength extraction via the Okapi sidecar with SRX sentence segmentation
+
+  The dialog only appears if the sidecar is running; otherwise the standard engine is used silently.
+- **Intelligent segment filtering** — The Okapi extraction path filters out non-body content (headers, footers, textbox names) using sub-document tracking, keeping the segment grid clean.
+- **Formatting tag extraction** — The sidecar detects inline formatting from OOXML run properties and converts them to Supervertaler's `<b>`, `<i>`, `<u>`, `<sup>`, `<sub>`, `<s>` tag system.
+- **SRX segmentation** — The Okapi engine uses industry-standard SRX rules for sentence segmentation, correctly handling abbreviations (Dr., Mr., etc.), decimal numbers, and other tricky boundaries.
+- **Seven REST endpoints** — The sidecar exposes: `/health`, `/filters`, `/extract`, `/merge`, `/tmx/read`, `/tmx/validate`, `/segment` — designed for future integration beyond DOCX import.
+- **Graceful degradation** — If the sidecar JAR or Java runtime is not present, Supervertaler functions exactly as before with no errors or warnings.
+
+> **Note:** The Okapi import engine is marked experimental. The standard import engine remains the recommended choice for production work. The Okapi path will be refined in upcoming releases with improved formatting fidelity and broader file format support.
+
+---
+
 ## v1.9.332 - February 26, 2026
 
 ### Critical Bug Fixes

{supervertaler-1.9.332 → supervertaler-1.9.334}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: supervertaler
-Version: 1.9.332
+Version: 1.9.334
 Summary: Professional AI-enhanced translation workbench with multi-LLM support, glossary system, TM, spellcheck, voice commands, and PyQt6 interface. Batteries included (core).
 Home-page: https://supervertaler.com
 Author: Michael Beijer
@@ -80,7 +80,7 @@ Dynamic: requires-python
 
 **Professional AI-enhanced translation workbench** with multi-LLM support (GPT-4, Claude, Gemini, Ollama), translation memory, glossary management, and seamless CAT tool integration (memoQ, Trados, CafeTran, Phrase, Déjà Vu).
 
-**Latest release:** v1.9.332 - Fix DOCX import silently dropping paragraphs; add word count verification safety net; TMX export now uses standard bpt/ept pairs.
+**Latest release:** v1.9.333 - Okapi Framework sidecar integration (experimental); industrial-strength DOCX extraction with SRX segmentation, inline formatting tag detection, and sub-document filtering.
 
 ---
 
@@ -119,7 +119,7 @@ macOS will block the app on first launch because it is not signed with an Apple
 - **Superlookup** - Unified concordance search across TM, glossaries, MT, and web resources
 - **CAT Tool Integration** - memoQ XLIFF, Trados SDLPPX/SDLRPX, CafeTran, Phrase, Déjà Vu X3
 - **Voice Commands** - Hands-free translation with OpenAI Whisper
-- **Document Support** - DOCX, bilingual DOCX/RTF, PDF, Markdown, plain text
+- **Document Support** - DOCX, bilingual DOCX/RTF, PDF, Markdown, plain text + experimental [Okapi Framework](https://okapiframework.org/) sidecar for industrial-strength file extraction
 
 ---
 

{supervertaler-1.9.332 → supervertaler-1.9.334}/README.md

@@ -6,7 +6,7 @@
 
 **Professional AI-enhanced translation workbench** with multi-LLM support (GPT-4, Claude, Gemini, Ollama), translation memory, glossary management, and seamless CAT tool integration (memoQ, Trados, CafeTran, Phrase, Déjà Vu).
 
-**Latest release:** v1.9.332 - Fix DOCX import silently dropping paragraphs; add word count verification safety net; TMX export now uses standard bpt/ept pairs.
+**Latest release:** v1.9.333 - Okapi Framework sidecar integration (experimental); industrial-strength DOCX extraction with SRX segmentation, inline formatting tag detection, and sub-document filtering.
 
 ---
 
@@ -45,7 +45,7 @@ macOS will block the app on first launch because it is not signed with an Apple
 - **Superlookup** - Unified concordance search across TM, glossaries, MT, and web resources
 - **CAT Tool Integration** - memoQ XLIFF, Trados SDLPPX/SDLRPX, CafeTran, Phrase, Déjà Vu X3
 - **Voice Commands** - Hands-free translation with OpenAI Whisper
-- **Document Support** - DOCX, bilingual DOCX/RTF, PDF, Markdown, plain text
+- **Document Support** - DOCX, bilingual DOCX/RTF, PDF, Markdown, plain text + experimental [Okapi Framework](https://okapiframework.org/) sidecar for industrial-strength file extraction
 
 ---