2494 Agents Full Reference V19313 2026 02 23 B581da1a
ASecurity
> **This is the single source of truth for AI coding assistants working on this project.** > **Last Updated:** February 23, 2026 | **Version:** v1.9.313 ---
Installs into .claude/skills of the current project.
Are you the author of 2494 Agents Full Reference V19313 2026 02 23 B581da1a?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/tools-only-2494-agents-full-reference-v19313-2026-02-23-b581d)
# Supervertaler - AI Agent Documentation
> **This is the single source of truth for AI coding assistants working on this project.**
> **Last Updated:** February 23, 2026 | **Version:** v1.9.313
---
## ⚡ QUICK START FOR AI AGENTS
**IMPORTANT: If you're continuing from a previous session or ran out of context:**
1. **Skip to the end of this file** - The most recent development context is in the **"🔄 Recent Development History"** section (search for the latest date)
2. **Current version: v1.9.313** - SDLPPX multi-file views, locked segment toggle, memoQ submenu
3. **Read only what you need** - The Project Overview and Architecture sections are reference material; the dated history entries contain the actual working context
**Quick Navigation:**
- **Latest context:** See CHANGELOG.md for v1.9.311-313 (Trados SDLXLIFF/SDLPPX improvements)
- **Module list:** Search for `## 🔌 Complete Module List`
- **Architecture:** Search for `## 🏗️ Architecture Patterns`
- **Common pitfalls:** Search for `## ⚠️ Common Pitfalls`
---
## 🎯 Project Overview
**Supervertaler** is a professional desktop translation application built with Python and PyQt6. It serves as a companion tool for translators, integrating AI-powered translation with traditional CAT (Computer-Assisted Translation) tool workflows.
| Property | Value |
|----------|-------|
| **Name** | Supervertaler |
| **Version** | v1.9.313 (February 2026) |
| **Framework** | PyQt6 (Qt for Python) |
| **Language** | Python 3.10+ |
| **Platform** | Windows (primary), Linux compatible |
| **Repository** | https://github.com/michaelbeijer/Supervertaler |
| **Website** | https://supervertaler.com |
| **Main File** | `Supervertaler.py` (~56,000+ lines) |
| **Modules** | 85+ specialized modules in `modules/` directory |
### Key Capabilities
- **Multi-LLM AI Translation**: OpenAI GPT-4, Anthropic Claude, Google Gemini, Local Ollama, Custom OpenAI-Compatible API (Volcengine/Doubao, Tongyi/Qwen, DeepSeek, etc.)
- **CAT Tool Integration**: Trados SDLPPX/SDLRPX, memoQ XLIFF/DOCX/RTF, Phrase/Memsource DOCX, CafeTran DOCX, Déjà Vu X3 RTF
- **Translation Memory**: Fuzzy matching TM with TMX import/export + Supermemory (ChromaDB vector search)
- **Terminology Management**: SQLite-based termbases with priority highlighting and automatic extraction
- **Document Handling**: DOCX, bilingual DOCX, PDF (via OCR), simple TXT, **Markdown (MD)**, **Multi-file folder import**
- **Quality Assurance**: Spellcheck, tag validation, consistency checking
- **Superlookup**: Unified concordance hub with TM, Termbase, Supermemory, MT, and Web Resources
- **Batteries Included**: Default install includes all major features and dependencies, except heavy optional components (Supermemory + offline Local Whisper)
---
## 📦 Installation (Simplified)
Supervertaler uses a batteries-included *core* install: `pip install supervertaler` pulls in everything needed for the major features, while keeping heavy optional components out of the default install.
**Note:** Offline dictation via Local Whisper is optional because it pulls in very large dependencies (PyTorch). The default/recommended voice path uses the OpenAI Whisper API.
**Legacy note:** older “extras” (e.g. `supervertaler[all]`) are still accepted for backward compatibility, but they no longer change what gets installed.
### Installation Options
| Command | Notes |
|---------|-------|
| `pip install supervertaler` | Recommended core install (excludes heavy optional Local Whisper) |
| `pip install supervertaler[local-whisper]` | Adds offline Local Whisper (heavy; pulls PyTorch) |
| `pip install supervertaler[all]` | Legacy alias (no-op; kept for compatibility) |
### Feature Modules
| Module | Included by default | Size | Description |
|--------|----------------------|------|-------------|
| **Supervoice** | ✅ | ~150 MB | Voice dictation & commands (OpenAI Whisper API recommended) |
| **Web Browser** | ✅ | ~100 MB | Built-in browser for Superlookup |
| **PDF Rescue** | ✅ | ~30 MB | PDF text extraction/OCR |
| **MT Providers** | ✅ | ~30 MB | DeepL, Amazon Translate |
| **Hunspell** | ✅ | ~20 MB | Advanced spellcheck |
| **AutoFingers** | ✅ (Windows) | ~10 MB | Windows automation |
### Settings → Features Tab
Users can enable/disable features in **Settings → 📦 Features**.
### Feature Manager Module
The `modules/feature_manager.py` provides:
- `FeatureManager` class for checking feature availability
- `FEATURE_MODULES` dict defining all optional features
- `lazy_import_*()` functions for conditional imports
- `check_feature(id)` quick availability check
---
## 🪟 Windows EXE Releases (CORE + FULL)
Supervertaler Windows releases are published as **two separate ZIP assets** on the same GitHub Release:
1) **CORE** (recommended for most users)
- Smaller download
- Does **not** bundle the heavy ML stack (Supermemory + offline Local Whisper)
2) **FULL** ("batteries included")
- Larger download
- Bundles **Supermemory** + **offline Local Whisper** (PyTorch / sentence-transformers / ChromaDB)
### Key Rule (PyInstaller one-folder builds)
The EXE must be run from the extracted distribution folder. Do **not** separate `Supervertaler.exe` from `_internal/`.
If users see an error like missing `python312.dll`, they are almost always:
- running the wrong EXE (from an intermediate build folder), or
- moving the EXE away from `_internal/`.
### How to Build (recommended)
Use the automated build script:
```powershell
# Build BOTH core + full and create both ZIP assets
powershell -NoProfile -ExecutionPolicy Bypass -File .\build_windows_release.ps1
# Build only core
powershell -NoProfile -ExecutionPolicy Bypass -File .\build_windows_release.ps1 -CoreOnly
# Build only full
powershell -NoProfile -ExecutionPolicy Bypass -File .\build_windows_release.ps1 -FullOnly
# Clean build venvs + rebuild (keeps dist/ so multi-flavor builds can coexist)
powershell -NoProfile -ExecutionPolicy Bypass -File .\build_windows_release.ps1 -Clean
```
What it does:
- Uses two isolated build environments: `.venv-build-core` and `.venv-build-full`
- Runs PyInstaller with the corresponding spec:
- `Supervertaler.core.spec` → `dist\Supervertaler-core\...`
- `Supervertaler.full.spec` → `dist\Supervertaler-full\...`
- Zips each output using `create_release_zip.py` and writes a `README_FIRST.txt` into each dist folder.
### Output Files
After a successful run, you should have:
- `dist\Supervertaler-v<version>-Windows-CORE.zip`
- `dist\Supervertaler-v<version>-Windows-FULL.zip`
### GitHub Release Posting
Attach **both** ZIP files to the same GitHub Release tag.
### VS Code Tasks
There are build tasks wired up in `.vscode/tasks.json`:
- "Build Windows EXE (core)"
- "Build Windows EXE (full)"
- "Build Windows EXE (core + full)"
---
## 📁 Project Structure
```
Supervertaler/
├── Supervertaler.py # Main application (~56,000+ lines)
├── modules/ # 85+ specialized modules
│ ├── feature_manager.py # Modular feature management (NEW)
│ ├── llm_clients.py # OpenAI, Anthropic, Google Gemini, Ollama
│ ├── translation_memory.py # TM matching and storage
│ ├── termbase_manager.py # Terminology management
│ ├── docx_handler.py # DOCX import/export
│ ├── sdlppx_handler.py # Trados Studio packages
│ ├── phrase_docx_handler.py# Phrase/Memsource bilingual
│ ├── cafetran_docx_handler.py # CafeTran bilingual
│ ├── supermemory.py # Vector-indexed semantic TM (ChromaDB)
│ ├── spellcheck_manager.py # Spellcheck with pyspellchecker/Hunspell
│ ├── prompt_library.py # AI prompt management
│ └── ... # See module list below
├── user_data/ # User content (gitignored)
│ ├── prompts/ # .svprompt files
│ ├── termbases/ # .db termbase files
│ ├── translation_memories/ # .db TM files
│ ├── dictionaries/ # Custom spellcheck words
│ └── supermemory/ # ChromaDB vector database
├── assets/ # Icons, images
├── docs/ # Documentation site
├── beijerterm/ # 🔗 GIT SUBMODULE - Beijerterm glossary website
├── tests/ # Test files
└── legacy_versions/ # Historical Tkinter version
```
---
## 🔗 Beijerterm Website Submodule (IMPORTANT FOR AI AGENTS)
The `beijerterm/` folder is a **Git submodule** - a separate repository embedded inside Supervertaler.
### What is a Submodule?
- It's a **pointer** to a specific commit in another repository
- Regular `git clone` does NOT download submodule contents (saves ~15 MB)
- The submodule has its own `.git` and tracks `michaelbeijer/beijerterm` separately
### Two Separate Projects
| Project | Location | Purpose |
|---------|----------|---------|
| **Supervertaler** | Root folder | PyQt6 translation app |
| **Beijerterm** | `beijerterm/` subfolder | Static glossary website (beijerterm.com) |
**Note**: The "Superlookup" name refers to the in-app lookup panel inside Supervertaler - don't confuse it with the Beijerterm website!
### Working with the Submodule
**To make changes to Beijerterm website:**
```bash
cd beijerterm/
# Edit files...
git add .
git commit -m "your message"
git push origin main # Pushes to michaelbeijer/beijerterm
```
**Then update the parent reference:**
```bash
cd .. # Back to Supervertaler root
git add beijerterm # Stage the new submodule commit pointer
git commit -m "chore: Update beijerterm submodule"
git push origin main # Pushes to michaelbeijer/Supervertaler
```
### ⚠️ Common Pitfalls
1. **Commits in submodule aren't automatically tracked** - After committing inside `beijerterm/`, you must also commit in the parent repo to update the pointer.
2. **Changelogs are separate** - Beijerterm website changes go in `beijerterm/CHANGELOG.md`, Supervertaler changes go in `CHANGELOG.md` at root.
3. **Features in the wrong repo** - The "Superlookup" panel inside Supervertaler.py is part of **Supervertaler**, not the Beijerterm website. Don't confuse them!
4. **Building the website** - Run `python scripts/build_site.py` from inside `beijerterm/`, not from root.
---
## 🔧 Key Technical Details
### Main Application (`Supervertaler.py`)
The main file is a large monolithic PyQt6 application. Key sections:
| Line Range | Purpose |
|------------|---------|
| 1-700 | Imports, constants, Project dataclass |
| 700-2000 | Custom widgets (grid editors, checkboxes) |
| 2000-4500 | MainWindow initialization, UI setup |
| 4500-8000 | Menu actions, file operations |
| 8000-12000 | Settings dialogs |
| 12000-18000 | Grid operations, navigation |
| 18000-25000 | Import/Export handlers |
| 25000-32000 | AI translation, batch operations |
### Key Classes
```python
@dataclass
class Project:
segments: List[Segment]
source_lang: str
target_lang: str
original_docx_path: Optional[str] = None
memoq_source_path: Optional[str] = None
sdlppx_source_path: Optional[str] = None
phrase_source_path: Optional[str] = None
original_txt_path: Optional[str] = None
# ... 20+ fields total
@dataclass
class Segment:
source: str
target: str = ""
status: str = "Not Started"
notes: str = ""
segment_type: str = "text"
# ... additional fields
```
### File Extensions
| Extension | Purpose |
|-----------|---------|
| `.svproj` | Supervertaler project files (JSON) |
| `.svprompt` | Prompt files (JSON) |
| `.svntl` | Non-translatables lists (JSON) |
---
## 🔌 Complete Module List
### AI & LLM (`modules/`)
- `llm_clients.py` - OpenAI, Anthropic Claude, Google Gemini, Ollama, Custom OpenAI-Compatible API integration
- `model_version_checker.py` - Auto-detect new LLM models from providers
- `model_update_dialog.py` - UI for selecting new models
- `prompt_library.py` - Prompt management and favorites
- `prompt_assistant.py` - AI-powered prompt generation
- `unified_prompt_library.py` - Unified prompt system
- `unified_prompt_manager_qt.py` - Prompt manager UI
- `voice_dictation.py` - Whisper-based voice input
- `voice_commands.py` - Talon-style voice command system (NEW)
- `ai_actions.py` - AI action system for prompt library
- `ai_attachment_manager.py` - File attachment persistence
- `ai_file_viewer_dialog.py` - File viewing dialog
### Translation Memory & Terminology
- `translation_memory.py` - Fuzzy matching TM system
- `supermemory.py` - ChromaDB vector semantic search (2100+ lines)
- `termbase_manager.py` - SQLite-based terminology
- `term_extractor.py` - Automatic term extraction
- `termbase_entry_editor.py` - Term editing UI
- `termbase_import_export.py` - TMX/TBX import/export
- `tm_manager_qt.py` - TM management UI
- `tm_metadata_manager.py` - TM metadata handling
- `tm_editor_dialog.py` - TM editing dialog
- `tmx_editor.py` / `tmx_editor_qt.py` - TMX file editing
- `tmx_generator.py` - TMX file generation
### File Handlers
- `docx_handler.py` - Standard DOCX import/export
- `sdlppx_handler.py` - Trados Studio SDLPPX/SDLRPX packages (767+ lines)
- `phrase_docx_handler.py` - Phrase/Memsource bilingual DOCX
- `cafetran_docx_handler.py` - CafeTran bilingual DOCX
- `trados_docx_handler.py` - Trados bilingual review DOCX
- `dejavurtf_handler.py` - Déjà Vu X3 bilingual RTF (NEW in v1.9.91)
- `mqxliff_handler.py` - memoQ XLIFF files
- `simple_segmenter.py` - Text segmentation
### Spellcheck & QA
- `spellcheck_manager.py` - Dual-backend spellcheck (pyspellchecker + Hunspell)
- `non_translatables_manager.py` - Non-translatable term management
- `tag_cleaner.py` - CAT tool tag removal
- `tag_manager.py` - Tag handling
### UI Components
- `ribbon_widget.py` - Ribbon-style toolbar
- `translation_results_panel.py` - Match display panel
- `termview_widget.py` - Inline term display
- `superlookup.py` - Unified lookup window
- `superbrowser.py` - Multi-chat AI browser
- `quick_access_sidebar.py` - Quick access panel
- `keyboard_shortcuts_widget.py` - Shortcut management
- `project_home_panel.py` - Project home UI
### Utilities
- `database_manager.py` - SQLite database operations
- `database_migrations.py` - Database schema migrations
- `config_manager.py` - Settings management
- `file_dialog_helper.py` - File dialog utilities
- `find_replace.py` - Find and replace functionality
- `shortcut_manager.py` - Keyboard shortcut handling
- `theme_manager.py` - UI theme management
- `statuses.py` - Segment status definitions
### Specialized Tools
- `pdf_rescue_Qt.py` - AI OCR for PDF extraction
- `image_extractor.py` - Extract images from DOCX
- `figure_context_manager.py` - Image context for AI
- `document_analyzer.py` - Document analysis
- `encoding_repair.py` / `encoding_repair_Qt.py` - Fix encoding issues
- `autofingers_engine.py` - memoQ AutoFingers automation
- `tracked_changes.py` - Track changes analysis
- `supercleaner.py` / `supercleaner_ui.py` - Text cleaning
### Benchmarking
- `llm_leaderboard.py` - LLM quality benchmarking
- `superbench_ui.py` - Benchmark UI
- `local_llm_setup.py` - Ollama setup wizard
---
## 🏗️ Architecture Patterns
### UI Pattern
- PyQt6 with custom styled widgets
- Consistent styling:
- Checkboxes: `CheckmarkCheckBox` (Standard), `PinkCheckmarkCheckBox` (Project), `BlueCheckmarkCheckBox` (Global)
- Radio Buttons: `CheckmarkRadioButton` (Standard Green)
- Tag-based text formatting (`<b>`, `<i>`, `<u>`, `<li-o>`, `<li-b>`)
- Grid-based segment editor with source (read-only) and target (editable) columns
### Data Flow
1. Import file → Parse to segments → Display in grid
2. User translates/edits → Status updates → Grid refreshes
3. Export → Reconstruct original format with translations
### Settings Storage
- `settings/settings.json` - Unified configuration with `api_keys`, `general`, `ui`, and `features` sections
- `settings/` - Satellite files: `themes.json`, `shortcuts.json`, `recent_projects.json`, `find_replace_history.json`, `superlookup_history.json`, `voice_commands.json`, `model_version_cache.json`
- Legacy settings files are one-time migrated on startup and renamed to `.migrated`
- `.svproj` files - Per-project settings
---
## 📝 Development Guidelines
### When Editing Supervertaler.py
1. The file is large (~38K+ lines) - use line ranges when reading
2. Search for method names with `grep_search` before editing
3. Follow existing patterns for new features
4. `__version__` auto-reads from `pyproject.toml` (dev) or `importlib.metadata` (pip) — no manual update needed
5. To bump version: edit `pyproject.toml`, `CHANGELOG.md`, and `docs/index.html` (see Version Bump Checklist below)
### When Adding New Modules
1. Create in `modules/` directory
2. Add import in main file where needed
3. Follow existing module patterns (docstrings, type hints)
### Documentation Updates
**Always update these files after changes:**
1. `AGENTS.md` - Add dated entry to development history
2. `CHANGELOG.md` - Add version entry
3. `README.md` - Update version badge if needed
### Commit Messages
Use semantic prefixes:
- `feat:` - New feature
- `fix:` - Bug fix
- `docs:` - Documentation
- `refactor:` - Code restructuring
- `style:` - Formatting
---
## ⚠️ Common Pitfalls
1. **ElementTree namespaces** - Always use namespace dict when working with SDLXLIFF:
```python
NAMESPACES = {'sdl': 'http://sdl.com/FileTypes/SdlXliff/1.0'}
element.find('.//sdl:seg', NAMESPACES)
```
2. **Grid widget access** - Use `table.cellWidget()` for QTextEdit, `table.item()` for QTableWidgetItem
3. **File paths** - Store absolute paths, use `os.path.exists()` before accessing
4. **Status updates** - Remember to update both internal data and grid display
5. **Signal blocking** - When setting text programmatically, use `blockSignals(True/False)` to prevent cascading events
6. **Qt event queue** - Be aware that `setPlainText()` queues events even when signals are blocked
7. **Hidden widget styling** - Qt may not apply stylesheets to hidden widgets. If theme colors aren't appearing, apply styles when the widget becomes visible, not just at creation time
8. **Race conditions & timing** - When debugging UI issues, consider whether the problem might be timing-related:
- Widgets created before theme manager is initialized
- Styles applied while widgets are hidden
- Signals firing before handlers are connected
- Use `QTimer.singleShot()` for deferred initialization when needed
---
## 📋 Planned Features & Refactoring
### 🔑 Inline API Key Editing in Settings UI (Completed in v1.9.240)
**Status:** Completed and released in v1.9.240
Users can now manage API keys directly in Settings > AI Settings with password-masked `QLineEdit` fields and show/hide toggles. Keys are read from and written to unified settings storage.
---
### 🔧 Configuration File Consolidation (Completed in v1.9.240)
**Status:** Completed and released in v1.9.240
Configuration was consolidated into `settings/settings.json` with four top-level sections:
- `api_keys`
- `general`
- `ui`
- `features`
The following files were also relocated under `settings/`: `themes.json`, `shortcuts.json`, `recent_projects.json`, `find_replace_history.json`, `superlookup_history.json`, `voice_commands.json`, and `model_version_cache.json`.
A one-time startup migration converts legacy files and renames originals to `.migrated`.
---
## � Future Investigation: TMX Tag Export Format
**Issue discovered December 22, 2025**: Our TMX Editor stores formatting tags (like `<b>`, `<i>`) as escaped text (`<b>`) in the `<seg>` element. This is valid XML but may not be the optimal approach.
**Three approaches exist:**
| Approach | Example in TMX file | Pros | Cons |
|----------|---------------------|------|------|
| **Escaped text** | `<seg><b>text</b></seg>` | Simple, valid XML | Tags treated as plain text, lost semantic meaning |
| **TMX inline elements** | `<seg><bpt i="1" type="bold"><b></bpt>text<ept i="1"></b></ept></seg>` | TMX 1.4 compliant, preserves tag semantics | More complex to implement |
| **Raw XML** (invalid) | `<seg><b>text</b></seg>` | Human readable | Invalid XML unless DTD defines `<b>` |
**What other tools do:**
- **memoQ**: Uses TMX inline elements (`<bpt>`, `<ept>`) for proper tag preservation
- **Trados Studio**: Similar, uses inline elements with type attributes
- **OmegaT**: Generally uses escaped text for simplicity
- **Many web tools**: Just escape everything
**Recommendation for future**: Consider implementing proper TMX inline elements (`<bpt>`, `<ept>`, `<ph>`) when exporting. This would:
1. Maintain compatibility with professional CAT tools
2. Preserve tag type information (bold, italic, etc.)
3. Allow round-tripping without tag loss
**Files to modify**: `modules/tmx_editor.py` - `TmxParser.save_file()` method
---
## �💡 Problem-Solving Tips for AI Agents
When stuck on a difficult bug, consider these approaches:
1. **Think about timing**: Is this a race condition? Are things happening in the wrong order?
- Widget creation vs. theme application timing
- Signal connections vs. signal emissions
- Hidden vs. visible widget state changes
2. **Think outside the box**: The obvious solution may not work
- If stylesheets aren't applying, try QPalette as an alternative
- If a method isn't being called, check if the widget is even visible
- If changes aren't reflected, check if there's caching involved
3. **Add debug output**: When behavior is mysterious, add logging to trace execution flow
- Print method entry/exit with timestamps
- Log parameter values and state
- Write to a debug file if console output is too fast
4. **Question assumptions**: What do you THINK is happening vs. what is ACTUALLY happening?
- The code might be running but not having the expected effect
- A different code path might be executing
- Something else might be overriding your changes
---
## 🧪 Testing
### Running Tests
```bash
pytest tests/
```
### Manual Testing Checklist
- [ ] Import DOCX, translate segment, export
- [ ] Save/load .svproj project
- [ ] TM matching works
- [ ] Termbase highlighting works
- [ ] AI translation (if API keys configured)
- [ ] Spellcheck toggles correctly
- [ ] SDLPPX import/export round-trip
---
## 🔑 API Keys
**Unified Settings Storage (v1.9.240+):**
API keys are stored in `settings/settings.json` under the `api_keys` section.
| Platform | Default Data Folder | API Key Storage |
|----------|---------------------|-----------------|
| **All Users (default)** | `~/Supervertaler/` | `settings/settings.json` → `api_keys` |
| **Windows (default)** | `C:\Users\Username\Supervertaler\` | `settings\settings.json` → `api_keys` |
| **Development** | `user_data_private\` (git-ignored) | `settings\settings.json` → `api_keys` |
**Backward compatibility:**
- Legacy `api_keys.txt` remains supported as a fallback input path.
- New writes persist to `settings/settings.json`.
**Supported keys:**
- `openai`, `claude`, `google`, `gemini`, `custom_openai`, `deepl`, `google_translate`, `ollama_endpoint`
**Notes:**
- `google` and `gemini` are aliases.
- `custom_openai` is for OpenAI-compatible endpoints; endpoint/model are configured in Settings > AI Settings.
- `ollama_endpoint` can override the default (`http://localhost:11434`).
---
## 🔄 Recent Development History
### February 8, 2026 - Unified Settings + Inline API Key Editing (v1.9.240)
**✨ Feature Summary**
Completed the unified settings migration and inline API key editing workflow.
**Status:** Implementation complete, version 1.9.240 released to PyPI
**Highlights:**
- Consolidated core settings into `settings/settings.json` with `api_keys`, `general`, `ui`, and `features` sections
- Added password-masked API key fields with show/hide toggles in Settings > AI Settings
- Implemented one-time startup migration that renames legacy files to `.migrated`
- Moved satellite settings/history files into the `settings/` subfolder
- Removed dead code for deprecated settings paths
---
### February 7, 2026 - Custom OpenAI-Compatible API Provider (v1.9.236)
**✨ Feature Summary**
Added a generic "Custom (OpenAI-Compatible API)" provider that enables any OpenAI SDK-compatible endpoint — Volcengine (ByteDance Doubao), Alibaba Tongyi (Qwen), DeepSeek, Mistral, Groq, and more. This addresses GitHub issue #155 requesting Chinese AI services that handle Chinese translation better.
**Status:** Implementation complete, version 1.9.236 released to PyPI
**Issue Addressed:**
- GitHub #155 - Feature request for Volcengine (Doubao) and Tongyi (Qwen) support
**Approach:**
Rather than adding each Chinese AI provider individually, implemented a single `custom_openai` provider that reuses `_call_openai()` with a user-specified `base_url`. Users configure: endpoint URL, API key (in `api_keys.txt` as `custom_openai = <key>`), and model name (free-text input since model names vary by provider).
**Files Modified:**
1. **`modules/llm_clients.py`**:
- Added `base_url: Optional[str] = None` parameter to `__init__`
- `translate()` routes `"custom_openai"` to `_call_openai()` (same as `"openai"`)
- `_call_openai()` passes `base_url` to `OpenAI()` constructor when set
- `DEFAULT_MODELS` includes `"custom_openai": "custom-model"`
- `load_api_keys()` includes `"custom_openai": ""` in defaults
- API key validation skipped for `custom_openai` (some endpoints don't need one)
2. **`Supervertaler.py`** — Settings UI:
- Added `custom_radio = CustomRadioButton("🔌 Custom (OpenAI-Compatible API)")` in `_create_ai_settings_tab()`
- Added `custom_endpoint_input` (QLineEdit for URL) and `custom_model_input` (QLineEdit for model name)
- Added `custom_enable_cb = CheckmarkCheckBox("Enable Custom (OpenAI-Compatible)")`
- Save handler (`_save_ai_settings_from_ui()`) saves `custom_openai_model`, `custom_openai_endpoint`
- `load_llm_settings()` includes `custom_openai_model` and `custom_openai_endpoint` defaults
- `load_provider_enabled_states()` includes `llm_custom_openai: True`
3. **`Supervertaler.py`** — Helper method:
- Added `create_llm_client(provider, model, api_keys, settings)` method that handles `base_url` logic in one place
- Used at ~12 `LLMClient()` instantiation points to avoid duplicating custom_openai logic
- `PreTranslationWorker` accepts `base_url` parameter for batch translation
4. **`modules/quicktrans.py`**:
- Added `("Custom", "CUS", "custom_openai", "mtql_custom_openai")` to `llm_defs`
- `_call_llm_translation()` handles base_url for custom_openai, reads endpoint from parent_app settings
5. **`user_data_private/api_keys.example.txt`**: Added `custom_openai` section with example endpoints
**LLM Provider Architecture (5 providers total):**
- `openai` — OpenAI GPT models
- `claude` — Anthropic Claude models
- `gemini` — Google Gemini models
- `ollama` — Local Ollama models (no API key needed)
- `custom_openai` — Any OpenAI-compatible endpoint (reuses `_call_openai()` with custom `base_url`)
**Settings stored in `settings/settings.json` (`ui` section):**
- `llm_settings.provider` — active provider name
- `llm_settings.custom_openai_model` — model name or endpoint ID
- `llm_settings.custom_openai_endpoint` — base URL (e.g., `https://ark.cn-beijing.volces.com/api/v3/`)
- `provider_enabled_states.llm_custom_openai` — enabled toggle
---
### February 7, 2026 - Version Display Fix for pip Users (v1.9.235)
**🐛 Bug Fix**
Fixed `_read_version()` which caused all pip-installed users to see version 1.9.227 regardless of the actual installed version. The function only tried `pyproject.toml` (which isn't included in pip wheels) and had a hardcoded fallback. Now uses a two-step approach:
1. Try `pyproject.toml` via `tomllib` (works in dev/source checkout)
2. Try `importlib.metadata.version("supervertaler")` (works after pip install)
3. Fallback to `"0.0.0"` instead of a misleading real version number
---
### February 7, 2026 - Multi-file Export & Batch Fixes (v1.9.232-234)
**✨ Features:**
- **Multi-file export "Original Format" option** (v1.9.234): Exports each file back to its source format (`.txt`, `.md`, or `.docx`), useful for mixed-format projects
- **Saved Views** (v1.9.232): Named views that filter the grid to selected files, persisted in project file
- **File boundary separators** (v1.9.232): Blue separator lines between files in multi-file projects
- **Markdown in multi-file import** (v1.9.232): `.md` files recognized alongside `.docx` and `.txt`
- **Tabbed Project Info dialog** (v1.9.232): Overview and File Progress tabs merged into one dialog
**🐛 Bug Fixes:**
- **Batch pre-translation SQLite thread error** (v1.9.233): Worker thread was calling main thread's SQLite connection for AI-inject glossary terms. Terms now pre-fetched on main thread and passed to worker.
---
### February 7, 2026 - memoQ RTF Fixes & UI Improvements (v1.9.228-231)
**🐛 Bug Fixes:**
- **memoQ RTF Unicode/formatting loss** (v1.9.231): Unicode escapes and RTF character control words were stripped by the generic cleanup regex. All Unicode escapes, hex escapes, and named character control words now decoded before the generic strip.
- **memoQ RTF combined formatting** (v1.9.231): Segments with bold+underline lost all formatting. Replaced pair-matching regex with direct marker-to-tag conversion.
- **memoQ RTF missing import options dialog** (v1.9.231): RTF import now shows the same formatting options dialog as DOCX import.
- **TM Read/Write settings persistence** (v1.9.228): Stale global TM activations could override project-specific settings on restart. Project-specific settings now always take priority.
**🎨 UI Improvements:**
- **Settings panel reorganized** (v1.9.227): AI Translation Preferences section reorganized with sub-headings
- **Version auto-read from pyproject.toml** (v1.9.227): `__version__` no longer needs manual updates
---
### February 6, 2026 - WYSIWYG Fix & Import Language Memory (v1.9.224-226)
**🐛 Bug Fixes:**
- **WYSIWYG/Tags toggle corrupted target text** (v1.9.225, #142): View mode toggle permanently destroyed whitespace/indentation. Fixed with `white-space: pre-wrap` CSS and `_suppress_target_change_handlers` guard.
- **Import dialogs ignored saved language pair** (v1.9.226, #143): Text/Markdown and multi-file import dialogs always defaulted to English → Dutch. Now all three import dialogs share language memory via `general_settings.json`.
---
### February 6, 2026 - memoQ Bilingual RTF Support (v1.9.223)
**✨ Feature Summary**
Added full import/export support for memoQ bilingual RTF files, addressing GitHub issue #145. This enables users with older memoQ versions (or those who prefer RTF format) to use the same bilingual table workflow as DOCX.
**Status:** Implementation complete, version 1.9.223 released
**Issue Addressed:**
- GitHub #145 - Feature request for memoQ RTF bilingual file support
**Implementation Details:**
The memoQ RTF bilingual format uses the identical 5-column table structure as memoQ DOCX:
- Column 1: Segment ID (number + GUID)
- Column 2: Source text (with formatting)
- Column 3: Target text
- Column 4: Comments
- Column 5: Status ("Not started", "Edited", "Confirmed", etc.)
**New Module - `modules/memoqrtf_handler.py`:**
```python
class MemoQRTFHandler:
"""Handler for memoQ bilingual RTF files."""
def load(self, file_path: str) -> bool:
"""Load and parse memoQ bilingual RTF."""
def save(self, output_path: str) -> bool:
"""Save RTF with updated translations."""
def get_source_texts(self) -> List[str]:
"""Get source segments for translation."""
```
**Key Features:**
- Parses RTF table structure using `\cell` and `\row` markers
- Handles RTF formatting codes (`\b` bold, `\i` italic, `\ul` underline)
- Decodes Unicode escapes (`\uNNNN?`) and hex character codes (`\'XX`)
- Preserves RTF structure for clean round-trip export
- Auto-detects source/target languages from header row
**Menu Integration:**
- Import: File → Import → memoQ Bilingual Table (RTF)...
- Export: File → Export → memoQ Bilingual Table - Translated (RTF)...
**Files Modified:**
1. **`modules/memoqrtf_handler.py`** (NEW):
- `MemoQRTFHandler` class for parsing/saving memoQ RTF
- `MemoQSegment` dataclass for segment representation
- RTF escape/decode utilities
- Language detection from header row
2. **`Supervertaler.py`**:
- Added menu actions for memoQ RTF import/export (lines ~8165, ~8226)
- Added `import_memoq_rtf()` method (lines ~27679-27805)
- Added `export_memoq_rtf()` method (lines ~28407-28515)
- Stores `memoq_rtf_source_path` in project for persistence
**Technical Notes:**
- RTF parsing uses regex to find cell content between `\cell` markers
- Unicode handling: `\uc0\uNNNN` and `\uNNNN?` patterns
- Negative Unicode values converted per RTF spec: `code + 65536`
- Target cell replacement done in reverse order to preserve positions
---
### February 5, 2026 - Custom Tooltips & Clean Slate Project Imports (v1.9.222)
**✨ Feature Summary**
Fixed black tooltip rendering issue on certain systems by implementing custom tooltip widgets. Also clarified and restored the "clean slate" behavior for new project imports.
**Status:** Implementation complete, version 1.9.219 released
**Issues Addressed:**
- Black tooltip rectangles when hovering over status icons (PyQt6/Qt tooltip rendering issue)
- GitHub #140 investigation ongoing (TM not readable after re-import on macOS)
**Custom Tooltip Implementation:**
Qt's built-in `QToolTip` rendered as black rectangles on some systems due to platform-specific styling issues. Solution: bypass Qt's tooltip system entirely with custom `QLabel` popup widgets.
**Key Code - Custom Tooltip Widget** (`Supervertaler.py`):
```python
def _get_custom_tooltip(self):
"""Get or create the custom tooltip label widget."""
if not hasattr(self, '_custom_tooltip') or self._custom_tooltip is None:
self._custom_tooltip = QLabel()
self._custom_tooltip.setWindowFlags(
Qt.WindowType.ToolTip | Qt.WindowType.FramelessWindowHint
)
self._custom_tooltip.setStyleSheet("""
QLabel {
background-color: #f5f5f5;
color: #333333;
border: 1px solid #d0d0d0;
padding: 4px 8px;
font-size: 12px;
}
""")
self._custom_tooltip.hide()
return self._custom_tooltip
```
**Event Filter for Status Icons:**
- Intercepts `QEvent.Type.ToolTip` events on status icon labels
- Shows custom tooltip popup instead of Qt's default
- Hides popup on `QEvent.Type.Leave`
**Clean Slate Project Imports (Design Clarification):**
The `_deactivate_all_resources_for_new_project()` function ensures new projects start with a clean slate:
- **Design Intent:** When importing a new document, NO TMs or glossaries should be pre-selected
- **User Workflow:** Import document → Select only needed TMs/glossaries → Work with relevant resources
- This prevents resource "pollution" from previous projects carrying over
**Files Modified:**
1. **`Supervertaler.py`**:
- Added `_get_custom_tooltip()` method for creating styled tooltip widgets
- Added `_show_status_tooltip()` and `_hide_status_tooltip()` methods
- Event filter on status icon labels to intercept tooltip events
- Confirmed `_deactivate_all_resources_for_new_project()` deactivates resources (clean slate)
2. **`modules/theme_manager.py`**:
- Added QToolTip styling in stylesheet (#f5f5f5 background, #333333 text)
- Added QPalette tooltip colors as fallback for Qt versions that ignore stylesheet
**Tooltip Color Choice:**
- Background: `#f5f5f5` (light gray) - professional, readable in all themes
- Text: `#333333` (dark gray) - high contrast
- Border: `#d0d0d0` - subtle definition
**GitHub #140 Investigation Notes:**
Bug report indicates TM not readable after re-importing document on macOS:
- Checkbox appeared checked but TM wasn't being searched
- Workaround: Uncheck, restart, recheck
- **Hypothesis:** When re-importing creates a new project_id, the TM activation records from the old project_id don't apply
- **Awaiting clarification** from bug reporter on exact workflow
---
### February 1, 2026 - Dark Mode Refinements & UI Improvements
**✨ Feature Summary**
Improved dark mode text visibility and fixed TM navigation arrows that were invisible or rendering incorrectly across light and dark themes.
**Status:** Implementation complete, version 1.9.184 released
**Files Modified:**
1. **`Supervertaler.py`** - Multiple UI improvements:
- TermView source text color: Changed to #FFFFFF in dark mode for better contrast (lines 175, 456)
- HTML tag colors: Light pink (#FFB6C1) in dark mode for `<b>`, `</b>` tags (lines 1403-1413, 2733-2743)
- Navigation arrows: Implemented ClickableArrow class with Unicode symbols (◀ ▶) and theme-aware colors (lines 29674-29733)
- Table header font: Reduced from `font_size + 1` to `font_size` for better proportions (line 30834)
- Theme refresh: Added arrow color updates in `refresh_theme_colors()` (lines 43993-43999)
2. **`modules/unified_prompt_manager_qt.py`** - Fixed Issue #112:
- Prompt edits now immediately reflected in Prompt Library and Preview Combined
- Cache refresh for both active primary and attached prompts (lines 2377-2384)
3. **`build_windows_release.ps1`** - Added Start Menu shortcut scripts to release packages (lines 111-116)
4. **`create_release_zip.py`** - Added shortcut creation instructions to README (lines 28-31)
**Key Code Locations:**
**ClickableArrow Class** (Supervertaler.py:29684-29705):
```python
class ClickableArrow(QLabel):
clicked = pyqtSignal()
def __init__(self, arrow_symbol, parent=None):
self.arrow_symbol = arrow_symbol
super().__init__("", parent)
self.setCursor(Qt.CursorShape.PointingHandCursor)
def set_color(self, color):
"""Update arrow color for current theme"""
self.setStyleSheet(f"""
QLabel {{
color: {color};
background: transparent;
border: none;
font-size: 11px;
font-weight: bold;
}}
""")
self.setText(self.arrow_symbol)
```
**Theme Color Logic:**
- Dark mode: White arrows (#FFFFFF), light pink tags (#FFB6C1)
- Light mode: Dark gray arrows (#333333), standard tag colors
**Development Notes:**
- Initially tried PNG arrow images but they rendered fuzzy
- Attempted Unicode angle brackets (❮ ❯) but font support was inconsistent
- Final solution: Unicode triangle symbols (◀ ▶) render crisply on all systems
- Arrow visibility issues were caused by arrows being created at startup before theme was applied
- Solution: ClickableArrow class with `set_color()` method called during theme refresh
**PowerShell Scripts Created:**
- `create_start_menu_shortcut.ps1` - For end users (Supervertaler.exe)
- `create_dev_start_menu_shortcut.ps1` - For developers (run.cmd)
**Related Issues:**
- Fixed #112: Prompt editing bug where saved prompts weren't updating in UI
---
### January 30-31, 2026 - Total Recall Architecture & Build System Unification
**✨ Feature Summary**
Implemented CafeTran-inspired "Total Recall" architecture for instant grid navigation. Instead of querying giant TMs on every segment click, relevant segments are extracted into lightweight in-memory structures on project load.
**Status:** Implementation complete, ready for testing
**Files Created:**
1. **`modules/project_tm.py`** - In-memory TM for instant lookups
2. **`modules/extract_tm.py`** - Persistent TM extraction to .svtm files
**Files Modified:**
1. **`Supervertaler.py`** (~938 lines changed)
2. **`modules/database_manager.py`** - Reduced candidate limit
3. **`build_windows_release.ps1`** - Unified build system
---
#### Implementation 1: In-Memory Termbase Index (Quick Win)
On project load, builds a Python dict mapping `lowercase_word -> [term_info, ...]` for O(1) termbase lookups.
**Key code locations in Supervertaler.py:**
- `self.termbase_index` initialization: ~line 6243
- `_build_termbase_index()`: ~line 22502
- `_search_termbase_index()`: ~line 22645
- Integration (called on project load): ~line 22265
---
#### Implementation 2: ProjectTM - In-Memory TM
**File:** `modules/project_tm.py`
On project load, extracts relevant TM segments (fuzzy matches ≥75%) into an in-memory SQLite database with FTS5 for fast fuzzy search.
**Features:**
- `ProjectTM` class with in-memory SQLite + FTS5
- `extract_from_database()` - extracts segments matching project content
- `search()` - instant lookup (exact match first, then FTS5 fuzzy)
- Thread-safe with locking
**Integration in Supervertaler.py:**
- `self.project_tm` initialization: ~line 6250
- Progress signal: `_project_tm_progress_signal` at ~line 6128
- Background extraction: `_start_project_tm_extraction_background()` at ~line 7553
- UI indicator in status bar: `self.project_tm_indicator` at ~line 7439
---
#### Implementation 3: ExtractTM - Persistent TM Extraction
**File:** `modules/extract_tm.py`
Extracts relevant segments from selected TMs into a `.svtm` file (SQLite) that persists across sessions.
**Features:**
- `ExtractTM` class with persistent SQLite storage
- `extract_and_save()` - extracts and saves to file
- `load()` - loads existing extraction
- `search()` - fast lookup with FTS5
- `export_to_tmx()` - export to standard TMX format
- File saved as `{ProjectName}_Extract.svtm` next to project
**Integration in Supervertaler.py:**
- Menu action: **Bulk → Extract from TMs...**
- Dialog: `show_extract_tm_dialog()` at ~line 33840
---
#### Implementation 4: Database Manager Optimization
**File:** `modules/database_manager.py`
Reduced TM candidate limit from 500 to 100 (~line 1016):
```python
# Before
candidate_limit = max(500, max_results * 50)
# After
candidate_limit = max(100, max_results * 10)
```
---
#### Implementation 5: Unified Build System
**File:** `build_windows_release.ps1`
Removed CORE/FULL split, now uses single unified `Supervertaler.spec`:
```powershell
# New usage
.\build_windows_release.ps1 # Build release
.\build_windows_release.ps1 -Clean # Clean build
# Output
dist\Supervertaler-v{version}-Windows.zip
```
---
#### Expected Performance Improvements
| Metric | Before | After |
|--------|--------|-------|
| Termbase lookup | 20-100ms | <1ms |
| TM lookup (grid) | 50-200ms | <5ms |
| Project load | Fast | Slightly slower (background extraction) |
---
#### Testing Checklist
- [ ] Open project with TMs - see ProjectTM extraction progress indicator
- [ ] Grid navigation feels faster after extraction
- [ ] Termbase matches appear instantly
- [ ] Bulk → Extract from TMs... dialog works
- [ ] .svtm file created next to project
- [ ] Build script: `.\build_windows_release.ps1`
---
#### Bug Fix (January 31, 2026)
Fixed attribute name mismatch: segment classes use `source_text`, but ProjectTM/ExtractTM were looking for `source`. Changed to try both:
```python
source = getattr(seg, 'source', None) or getattr(seg, 'source_text', None)
```
#### Git Status (Uncommitted)
```
M Supervertaler.py
M build_windows_release.ps1
M modules/database_manager.py
?? modules/extract_tm.py
?? modules/project_tm.py
```
---
#### Related Documentation
- **Design doc:** `CLAUDE_SESSION_HANDOFF.md` - original architecture design
---
### January 30, 2026 - Global UI Font Scale Feature (v1.9.180)
**✨ Feature Summary**
A user-configurable setting (50%-200%) that scales the entire application UI. Particularly useful for Linux/macOS users where Qt applications may render with smaller fonts, or for high-DPI displays.
**Files Modified:**
1. **`modules/theme_manager.py`**
2. **`Supervertaler.py`**
3. **`CHANGELOG.md`**
4. **`FAQ.md`**
**Implementation Details:**
**Step 1: ThemeManager (`modules/theme_manager.py`)**
In `__init__` method, added after `self.custom_themes`:
```python
# Global UI font scale (50-200%, default 100%)
self.font_scale: int = 100
```
In `apply_theme` method, added at the beginning (after getting `theme = self.current_theme`):
```python
# Calculate scaled font sizes based on font_scale (default 100%)
base_font_size = int(10 * self.font_scale / 100) # Base: 10pt at 100%
small_font_size = max(7, int(9 * self.font_scale / 100)) # Small text (status bar)
# Font scaling rules (only applied if scale != 100%)
font_rules = ""
if self.font_scale != 100:
font_rules = f"""
/* Global font scaling ({self.font_scale}%) */
QWidget {{ font-size: {base_font_size}pt; }}
QMenuBar {{ font-size: {base_font_size}pt; }}
# ... (all Qt widget types)
"""
```
Then prepended `font_rules` to the stylesheet: `stylesheet = font_rules + f"""...`
**Step 2: Supervertaler.py**
- Replaced "Settings Panel Font Size" UI (~line 17721) in `_create_view_settings_tab()` with "Global UI Font Scale"
- Changed SpinBox range from 80-200 to 50-200
- Updated setting key from `settings_ui_font_scale` to `global_ui_font_scale`
- Replaced `_apply_settings_ui_font_scale()` with `_apply_global_ui_font_scale()`:
```python
def _apply_global_ui_font_scale(self, scale_percent: int):
"""Apply font scale to the entire application UI"""
general_settings = self.load_general_settings()
general_settings['global_ui_font_scale'] = scale_percent
# Remove old key if present (migration)
if 'settings_ui_font_scale' in general_settings:
del general_settings['settings_ui_font_scale']
self.save_general_settings(general_settings)
# Update ThemeManager and reapply theme
if hasattr(self, 'theme_manager') and self.theme_manager is not None:
self.theme_manager.font_scale = scale_percent
self.theme_manager.apply_theme(QApplication.instance())
# Update status bar and main tabs fonts
self._update_status_bar_fonts(scale_percent)
self._update_main_tabs_fonts(scale_percent)
self.log(f"✓ Global UI font scale set to {scale_percent}%")
```
- Added helper methods: `_update_status_bar_fonts()`, `_update_main_tabs_fonts()`, `_get_global_ui_font_scale()`
- Applied font scale at startup after `self.theme_manager = ThemeManager(...)`:
```python
saved_font_scale = self._get_global_ui_font_scale()
self.theme_manager.font_scale = saved_font_scale
```
**Settings Storage:**
- Key: `global_ui_font_scale` (replaces `settings_ui_font_scale`)
- File: `general_settings.json`
- Default: 100
- Range: 50-200
---
### January 28, 2026 - Fresh Projects Start Clean (v1.9.172)
**🐛 Bug Fix: TM/Glossary Deactivation on Project Load**
Fixed bug where TMs and glossaries remained activated from previous sessions when loading or creating new projects. Users expected a clean slate but saw resources from previous work.
**Root Cause:**
In `load_project()`, TM deactivation only ran if the project had saved `activated_tm_ids` in its `tm_settings`. For older projects or newly created projects without `tm_settings`, deactivation was skipped entirely.
**Fix:**
Now follows the same pattern as glossaries:
1. **Always** deactivate all TMs for the project first (start clean)
2. **Then** restore saved TM activations if they exist in the project file
3. **Always** refresh the TM UI after (moved outside the conditional)
**Files Modified:**
- `Supervertaler.py` - Restructured TM restoration in `load_project()` to unconditionally deactivate all TMs first
---
### January 28, 2026 - TM Target & Alt+0 Badge Regression Fix (v1.9.171)
**🐛 Bug Fix: TM Target Display & Alt+0 Shortcut Fully Restored**
Fixed multiple issues with the TM Target display and Alt+0 shortcut in the Match Panel:
1. **Indentation Bug**: Fixed Python indentation error in `_update_match_panel_tm_display()` where TM Target update code was incorrectly nested inside the TM Source `else` block, causing TM Target to never display when valid matches existed.
2. **Alt+0 Handler**: Fixed `_handle_compare_panel_alt0_shortcut()` which checked for mode `'compare'` but `_get_active_match_shortcut_mode()` never returned that value. Updated mode detection to return `'match'` when Match Panel is active.
3. **Badge Styling**: Moved badge from HTML (which doesn't support `border-radius` in QTextEdit) to a proper QLabel widget in the title bar. Changed color from `#2196F3` to `#1976D2` to match Termview badges exactly.
**Root Causes:**
- TM Target text code was inside wrong indentation block (never executed for valid matches)
- Alt+0 shortcut handler checked for non-existent mode value
- HTML badges in QTextEdit render as squares, not circles
**Files Modified:**
- `Supervertaler.py`:
- Fixed indentation in `_update_match_panel_tm_display()` (lines 28746-28754)
- Added `shortcut_badge_text="0"` to Match Panel TM Target box creation
- Changed badge color to `#1976D2` in `_create_compare_panel_box()`
- Stored `self.match_panel_widget` reference for mode detection
- Updated `_get_active_match_shortcut_mode()` to return `'match'` for Match Panel
- Updated `_handle_compare_panel_alt0_shortcut()` to handle Match Panel mode
---
### January 27, 2026 - Scratchpad Tab in Right Panel (v1.9.170)
**📝 Scratchpad Tab Enhancement**
Added Scratchpad as a permanent tab in the right panel for easier access alongside the existing popup dialog.
**Feature Details:**
- **Location**: Right panel tabs → last tab after "Session Log"
- **Auto-Update**: Content automatically syncs with project's scratchpad notes
- **Dual Access**: Available both as popup dialog (`Ctrl+Shift+P`) and as permanent tab
- **Project-Aware**: Tab clears when creating new project, populates when loading project
**Implementation:**
- Added `_scratchpad_widget_for_right_panel` QPlainTextEdit widget
- Added new tab "📝 Scratchpad" to right panel tabs
- Added `_on_scratchpad_changed()` handler for text changes
- Added `_update_scratchpad_for_project()` method to populate tab
- Called in `load_project()` and `new_project()` for project sync
**Files Modified:**
- `Supervertaler.py` - Right panel tabs, scratchpad widget, change handlers
- `CHANGELOG.md` - v1.9.170 entry
- `AGENTS.md` - Updated version and history
---
### January 27, 2026 - Scratchpad for Private Notes (v1.9.169)
**📝 Scratchpad Feature**
Added a private notes scratchpad for translators to store personal notes during a job.
**Feature Details:**
- **Access**: `Tools → 📝 Scratchpad...` or `Ctrl+Shift+P` keyboard shortcut
- **Privacy**: Notes stored only in `.svproj` file, **never** exported to CAT tools
- **Persistence**: Notes saved with project and restored on reopen
**Use Cases:**
- Terminology decisions and rationale
- Client preferences and style notes
- Research findings and reference links
- Questions to ask the project manager
- Personal reminders and to-do items
**Implementation:**
- Added `scratchpad_notes: str = ""` field to `Project` dataclass
- Added serialization/deserialization in `to_dict()` and `from_dict()`
- New `ScratchpadDialog` class with QPlainTextEdit and monospace font
- New `show_scratchpad()` method in MainWindow
**Files Modified:**
- `Supervertaler.py` - Project dataclass, ScratchpadDialog, show_scratchpad(), Tools menu item
- `CHANGELOG.md` - v1.9.169 entry
- `AGENTS.md` - Updated version and history
---
### January 27, 2026 - Markdown Import with Syntax Highlighting (v1.9.168)
**📝 Markdown File Import Support** ([#127](https://github.com/michaelbeijer/Supervertaler/issues/127))
Added support for importing Markdown files (`.md`) with full syntax highlighting to make Markdown codes stand out visually during translation.
**New Features:**
- **File Filter**: Import → Text / Markdown File (TXT, MD)... now accepts `.md` files
- **Markdown Detection**: Automatically detects `.md` extension and enables syntax highlighting
- **Smart Dialog**: Shows Markdown-specific import instructions
**Syntax Highlighting Colors:**
| Element | Pattern | Color | Style |
|---------|---------|-------|-------|
| Headings | `#`, `##`, etc. | Blue (#0066CC) | Bold |
| Bold/Italic markers | `**`, `*`, `__`, `_` | Violet (#C71585) | Bold |
| Inline code | `` ` ``, `` ``` `` | Orange (#D2691E) | Bold |
| Links/Images | `[]()`, `![]()` | Purple (#6A5ACD) | Normal |
| Blockquotes | `>` | Green (#228B22) | Bold |
| Lists | `-`, `*`, `+`, `1.` | Orange (#FF6600) | Bold |
**Implementation Details:**
- New class flag: `TagHighlighter._is_markdown_project` (similar to existing `_is_cafetran_project`)
- New method: `TagHighlighter._highlight_markdown_syntax()` with 11 regex patterns
- Format definitions added to `update_tag_format()`: `md_bold_format`, `md_heading_format`, `md_code_format`, `md_link_format`, `md_quote_format`, `md_list_format`
- Flag is reset in DOCX import and set appropriately in text/MD import
**Design Decision:**
Markdown syntax is preserved as plain text (not converted to internal tags like `<b>`, `<i>`) because:
1. Many translators work with Markdown frequently and understand the syntax
2. Enables round-trip export back to `.md` without loss
3. Highlighting makes the codes visually distinct without hiding them
**Files Modified:**
- `Supervertaler.py` - TagHighlighter class, `import_simple_txt()`, DOCX import flag reset
- `CHANGELOG.md` - v1.9.168 entry
---
### January 26, 2026 - Proactive Grid Highlighting (v1.9.161)
**👁️ See-Ahead Glossary Highlighting**
New "proactive highlighting" feature that highlights glossary matches in upcoming segments while you're still working on the current one. You can now see glossary terms highlighted in segments 255, 256, 257 while editing segment 254.
**The Problem:**
- Even with cached prefetch data, glossary highlighting only appeared AFTER navigating to a segment
- User would press Ctrl+Enter → cursor moves instantly → but still wait for highlighting to apply
- Desired behavior: highlighting should already be visible before navigation
**The Solution:**
Added a signal-based system to apply highlighting proactively from the prefetch worker:
1. **New Signal**: `_proactive_highlight_signal = pyqtSignal(int, str)` - emits segment_id and JSON-encoded termbase matches
2. **New Slot**: `_apply_proactive_highlighting(segment_id, termbase_matches_json)` - finds the row and applies highlighting on the main thread
3. **Prefetch Worker Update**: After caching termbase matches, emits the proactive highlight signal
**How It Works:**
- Prefetch worker finds termbase matches for upcoming segments
- For each segment with matches, it emits `_proactive_highlight_signal(segment_id, matches_json)`
- Signal is automatically queued to main thread's event loop (Qt thread safety)
- Main thread receives signal and applies green highlighting to that segment's source cell
- Result: User sees highlighted terms appearing in upcoming rows while still working on current segment
**Key Code Changes:**
```python
# In _prefetch_worker_run():
if tb_count > 0:
with self.termbase_cache_lock:
termbase_raw = self.termbase_cache.get(segment_id, {})
if termbase_raw:
termbase_json = json.dumps(termbase_raw)
self._proactive_highlight_signal.emit(segment_id, termbase_json)
```
**Files Modified:**
- `Supervertaler.py` - Added `_proactive_highlight_signal`, `_apply_proactive_highlighting()`, updated `_prefetch_worker_run()`
---
### January 25, 2026 - Prefetch Now Does Direct Termbase Lookups (v1.9.160)
**⚡ Instant Glossary Matches After Ctrl+Enter**
Fixed issue where glossary matches still took 3-5 seconds to appear after navigation, despite prefetch system being in place.
**The Problem:**
- User reported: navigation is instant (v1.9.159 fix worked), but glossary matches still take 3-5 seconds
- Expected: prefetched results should appear instantly from cache
- Root cause: **Race condition between two caches**
**Root Cause Analysis:**
The prefetch system relied on `termbase_cache` (populated by a separate batch worker), but if prefetch ran BEFORE the batch worker processed a segment:
1. `_fetch_all_matches_for_segment()` checked `termbase_cache` → empty
2. Prefetch found 0 termbase matches → didn't cache the segment (only caches if matches found)
3. User navigated to segment → cache miss → slow lookup
**The Solution:**
Give the prefetch worker its own SQLite connection to do **direct termbase lookups** instead of relying on the batch worker's cache:
1. `_prefetch_worker_run()` now creates a thread-local SQLite connection
2. `_fetch_all_matches_for_segment(segment, thread_db_cursor)` now accepts an optional cursor
3. If termbase_cache doesn't have the segment AND cursor is provided → direct lookup via `_search_termbases_thread_safe()`
4. Results also populate `termbase_cache` for future use (other components benefit too)
**Key Code Changes:**
```python
# If not in cache and we have a thread-local cursor, do direct lookup
if termbase_matches_raw is None and thread_db_cursor is not None:
termbase_matches_raw = self._search_termbases_thread_safe(
segment.source, thread_db_cursor, source_lang, target_lang
)
# Also populate the termbase cache for future use
if termbase_matches_raw:
with self.termbase_cache_lock:
self.termbase_cache[segment.id] = termbase_matches_raw
```
**Files Modified:**
- `Supervertaler.py` - Updated `_prefetch_worker_run()` to create thread-local DB connection
- `Supervertaler.py` - Updated `_fetch_all_matches_for_segment()` to accept cursor and do direct lookup
---
### January 25, 2026 - Idle Prefetch for Instant Ctrl+Enter (v1.9.158)
**⚡ Predictive TM/Glossary Loading While You Work**
New idle prefetch system that loads matches for the next segments while you're thinking/typing, making Ctrl+Enter feel instant.
**How It Works:**
- When you stop typing for ~1 second (debounce), the app prefetches TM/glossary matches for the next 5 segments
- By the time you press Ctrl+Enter, matches are already cached
- Combines with existing "prefetch 20 on navigation" for comprehensive coverage
**The Problem:**
- User noticed lag when pressing Ctrl+Enter despite cache system existing
- Root cause: prefetch only started AFTER landing on a segment, not during editing time
- Fast workflow could outpace the prefetch worker
**The Solution:**
- Added `_trigger_idle_prefetch()` method called from `_handle_target_text_debounced_by_id()`
- Uses the 1-second typing debounce as trigger (user stopped to think = good time to prefetch)
- Prefetches next 5 segments (small enough to complete quickly, enough for fast workflow)
- Skips already-cached segments to avoid duplicate work
**Two-Tier Prefetch System:**
1. **On navigation** (existing): Prefetch next 20 segments when you land on a row
2. **On idle** (NEW): Prefetch next 5 segments when you stop typing for 1 second
**Files Modified:**
- `Supervertaler.py` - Added `idle_prefetch_timer` and `idle_prefetch_delay_ms` instance variables
- `Supervertaler.py` - Added `_trigger_idle_prefetch()` method
- `Supervertaler.py` - Called from `_handle_target_text_debounced_by_id()` after typing pause
---
### January 25, 2026 - TM Fuzzy Match Fix for Multi-TM Projects (v1.9.157)
**🔍 Fixed Missing TM Matches When Using Multiple Translation Memories**
Resolved regression where fuzzy TM matches were not being found in projects with multiple activated TMs.
**The Problem:**
- User reported segment 75 ("In een uitvoeringsvorm wordt Support Vector Machines (SVMs) gebruikt...") not finding expected 76% match
- Match existed in TM: "In een uitvoeringsvorm wordt Random Forests..."
- Issue appeared after Ctrl+Enter performance optimizations in v1.9.156
- Root cause: FTS5 candidate limit was too small for multi-TM search scenarios
**Root Cause Analysis:**
- FTS5 full-text search uses BM25 ranking which prioritizes entries matching MORE search terms
- When searching multiple TMs, the candidate pool was being filled with lower-quality matches from larger TMs
- The truly similar entry was at **position 102** in the candidate list - beyond the old 200-entry limit
- Old limit: `candidate_limit = max(200, max_results * 20)`
**The Solution:**
Increased FTS5 candidate pool in `_search_single_tm_fuzzy()`:
```python
# Increased from max(200, max_results * 20) to handle multi-TM scenarios better
# Position 102 finding of "Random Forests" entry proved 200 was too restrictive
candidate_limit = max(500, max_results * 50)
```
**Why This Fix Works:**
- Larger candidate pool ensures similar entries aren't pushed out by BM25 ranking
- SequenceMatcher then correctly scores all candidates by actual text similarity
- Negligible performance impact since FTS5 is already very fast
**Technical Details:**
- Key function: `_search_single_tm_fuzzy()` in `modules/database_manager.py`
- Line ~962-965: Increased candidate_limit calculation
- Each TM is searched separately with the larger pool, then results merged
**Files Modified:**
- `modules/database_manager.py` - Increased `candidate_limit` in `_search_single_tm_fuzzy()`
- `Supervertaler.py` - Added missing `enable_sound_effects` initialization (unrelated save bug fix)
---
### January 25, 2026 - Ctrl+Enter Performance Optimization (v1.9.156)
**⚡ Faster Segment Confirmation (Ctrl+Enter)**
Reduced verbose logging that was causing delays during segment confirmation and navigation.
**The Problem:**
- User reported Ctrl+Enter taking up to 3 seconds to confirm segment and move to next
- Analysis showed excessive logging overhead - 15-20 log calls per segment navigation
- DEBUG print statements in TM fuzzy search running on every search operation
- Cache was working but navigation still slow due to logging
**The Solution:**
1. **Removed DEBUG prints from TM fuzzy search** (`modules/database_manager.py`):
- Removed 5 `print(f"[DEBUG] search_fuzzy_matches: ...")` statements
- These ran on every TM search operation, adding overhead
- Kept only error logging for SQL failures
2. **Reduced DELAYED TM SEARCH logging** (`Supervertaler.py`):
- Consolidated 5-6 log calls into a single message
- Now only logs when matches are found: `"🔍 TM: Found X matches for segment Y"`
3. **Removed auto-insert debug logging** (`Supervertaler.py`):
- Removed 8+ verbose log calls in cache hit path
- Kept only the essential auto-insert success message
4. **Simplified _confirm_current_row_segment logging** (`Supervertaler.py`):
- Removed 8+ verbose debug log calls (source preview, target preview, object IDs, verification)
- Kept only: `"🔍 Ctrl+Enter: Row X, Segment ID Y"` and `"✅ Segment Y confirmed"`
**Performance Impact:**
- ~15-20 fewer log calls per segment confirmation
- ~5 fewer print statements per TM search
- More responsive feel when navigating segments
**Files Modified:**
- `modules/database_manager.py` - Removed DEBUG prints from `search_fuzzy_matches()`
- `Supervertaler.py` - Reduced logging in delayed TM search, auto-insert, and confirm functions
---
### January 25, 2026 - Preview Panel Performance & Navigation (v1.9.155)
**⚡ Preview Tab Performance Optimization**
Paused background TM/glossary lookups when Preview tab is open for better responsiveness.
**The Problem:**
- When "zipping around the preview" clicking different segments, the app felt sluggish
- Background TM searches, termbase lookups, and prefetch workers were running on every segment change
- These operations are unnecessary when user is just reading/reviewing in Preview mode
**The Solution:**
- Added `_is_preview_tab_active()` helper method to check if Preview tab is selected
- Added early return in `_on_cell_selected_full()` to skip heavy lookups when Preview is active
- Stored `_preview_tab_index` as instance variable for visibility checks
**What Still Runs in Preview Mode:**
- ✅ Preview scroll and segment highlighting
- ✅ Grid row highlighting (orange segment number)
- ✅ Toolbar segment info update
- ✅ Notes panel update
**What Is Skipped in Preview Mode:**
- ⏭️ TM searches
- ⏭️ Termbase/glossary lookups
- ⏭️ Termview updates
- ⏭️ MT/LLM match scheduling
- ⏭️ Prefetch workers for next segments
**Files Modified:**
- `Supervertaler.py` - Added `_is_preview_tab_active()`, stored `_preview_tab_index`, added early return in cell selection
---
### January 25, 2026 - Match Panel Consolidation (v1.9.154)
**🎯 Compare Panel Removed, Match Panel Enhanced**
Streamlined the right panel by removing the Compare Panel (redundant with Translation Results) and enhancing the Match Panel.
**The Changes:**
1. **Compare Panel Removed**:
- Deleted `_create_compare_panel()` tab creation
- Removed from View menu and Settings
- Was redundant with Translation Results panel
2. **Match Panel Now Shows TM Matches**:
- Combines Termview (top) + TM Source/Target boxes (bottom)
- TM boxes have green background (#d4edda) for easy identification
- TM navigation arrows still work (prev/next match)
- Match metadata (TM name, match %) displayed
3. **Zoom Shortcuts Redirected**:
- Ctrl+Alt+= (zoom in) and Ctrl+Alt+- (zoom out) now work on Match Panel
- View menu renamed from "Compare Panel" to "Match Panel"
- All zoom methods renamed: `match_panel_zoom_in/out/reset()`
- Settings key changed to `match_panel_font_size`
4. **Fixed TM Match Display**:
- Removed guards checking for `compare_panel_current_source`
- `set_compare_panel_matches()` now always updates Match Panel
- TM matches properly flow to Match Panel regardless of Compare Panel existence
**New Right Panel Tabs (5 total):**
1. Translation Results
2. Match Panel (Termview + TM boxes)
3. Preview
4. Segment Note
5. Session Log
**Files Modified:**
- `Supervertaler.py` - Compare Panel removal, Match Panel TM display, zoom shortcuts
---
### January 23, 2026 - Tab Layout Reorganization (v1.9.153)
**📐 Phase 1: Dual Termview & Right Panel Consolidation**
Implemented major UI reorganization to improve translation workflow by consolidating tabs and reducing visual clutter.
**The Changes:**
1. **Termview Under Grid** (Preserved):
- Original Termview remains below the grid
- User can collapse/hide it via splitter if desired
- Quick access to glossary terms while editing
2. **Second Termview in Right Panel** (NEW):
- Created duplicate TermviewWidget instance: `self.termview_widget_right`
- Both Termviews update simultaneously when segments change
- New helper method: `_update_both_termviews(source_text, termbase_list, nt_matches)`
- Replaced all 5 direct `termview_widget.update_with_matches()` calls with helper
3. **Segment Note & Session Log Moved**:
- Moved from bottom_tabs to right_tabs
- Now alongside Translation Results, Compare Panel, Preview
- Consolidated resources in one location
4. **Updated Tab Structure**:
- **Left panel**: Grid + Termview (single tab)
- **Right panel**: Translation Results, Compare Panel, Preview, Segment Note, Session Log, Termview (6 tabs)
5. **Ctrl+N Shortcut Updated**:
- Now searches right_tabs for "Segment note" tab by name
- Works regardless of which tabs are shown/hidden
- Still focuses notes editor for immediate typing
**Implementation Details:**
- Line 11494: `_update_both_termviews()` helper method
- Line 19686-19698: Bottom tabs now only contains Termview
- Line 19766-19795: Right tabs now includes Segment Note, Session Log, and second Termview
- Line 6609: `focus_segment_notes()` updated to search right_tabs
- 5 locations updated to use helper: quick-add term, on_cell_selected (2x), highlight_source_with_termbase
**Future Enhancement (Phase 2 - Not Yet Implemented):**
User requested advanced docking functionality:
- Drag tabs to dock vertically in right panel
- Any two tabs visible simultaneously while translating
- Requires QDockWidget architecture (similar to VS Code/Qt Creator)
- Complex implementation - would need to replace QTabWidget with QMainWindow + QDockWidgets
**Files Modified:**
- `Supervertaler.py` - All UI reorganization and Termview updates
---
### January 23, 2026 - Instant Glossary Updates (v1.9.152)
**⚡ Lightning-Fast Term Addition Performance**
Eliminated 5-6 second delays when adding glossary terms during translation. Users can now build glossaries rapidly during intensive patent translation workflows.
**The Problem:**
- Users experienced 5-6 second delays after adding terms with Alt+Shift+Up/Down shortcuts
- Long patent sentences with 50+ words triggered 50+ individual database queries
- `find_termbase_matches_in_source()` was searching for ALL words just to find the ONE term we added
- The workflow felt sluggish and interrupted translation flow
**Root Cause Analysis:**
After adding a term to the database, the code called `_refresh_termbase_display_for_current_segment()` which:
1. Cleared the termbase cache for the segment
2. Called `find_termbase_matches_in_source()` to search ALL words in the segment
3. Each word triggered a database query: `db_manager.search_termbases(word, ...)`
4. For a segment with 50 words, this meant 50+ database queries
5. The log showed 6-second gap between cache clear and search completion
**The Solution: Skip the Search, We Already Know What We Added!**
Instead of searching for what we just added, create the match entry directly:
```python
# OPTIMIZATION: Directly add the new term to cache and TermView instead of full search
# This avoids the 5-6 second delay from searching all words in long patent segments
new_match = {
'source': source_text,
'translation': target_text,
'priority': 99,
'ranking': glossary_rank,
'term_id': term_id,
'termbase_id': target_termbase['id'],
'termbase_name': target_termbase['name'],
# ... other fields
}
# Add to cache directly
with self.termbase_cache_lock:
if segment_id not in self.termbase_cache:
self.termbase_cache[segment_id] = {}
self.termbase_cache[segment_id][term_id] = new_match
# Update TermView with all cached matches (including new one)
self.termview_widget.update_with_matches(segment.source, termbase_list, nt_matches)
# Update source highlighting directly
self.highlight_source_with_termbase(current_row, segment.source, cached_matches)
```
**Benefits:**
- ✅ TermView updates instantly (< 0.1 seconds vs 5-6 seconds)
- ✅ Source highlighting updates instantly
- ✅ Zero database searches for operation we already know the result of
- ✅ Smooth, responsive workflow for building glossaries
- ✅ Perfect for intensive patent translation workflows with many term additions
**Implementation Details:**
- Modified `_quick_add_term_with_priority()` at line ~11975
- ~60 lines of new cache update and display logic
- Fallback to full refresh if anything goes wrong (safety net)
**Files Modified:**
- `Supervertaler.py` - Optimized glossary quick-add workflow
---
### January 23, 2026 - TM Pre-Translation Fixed (v1.9.151)
**🔧 Critical Fix: "Pre-translate from TM" Now Works**
User reported that "Pre-translate from TM" batch operation found no matches, despite visible 100% TM match in Compare Panel.
**Root Cause: SQLite Thread Safety**
The `PreTranslationWorker` ran in a background QThread, but SQLite connections created in the main thread cannot be used in other threads. This caused:
```
sqlite3.ProgrammingError: SQLite objects created in a thread can only be used in that same thread.
The object was created in thread id 28960 and this is thread id 15536.
```
**User's Key Insight:**
> "What I don't understand is you can't get Batch Translate to look up exact matches in the TM. However, all I need to do is move to the next segment, and if there is a match, it is shown in the Compare panel. Can't you just use the same system that the Compare panel uses to find these matches?"
**The Fix:**
Instead of trying to work around SQLite threading (creating thread-local connections), the TM pre-translation now runs **on the main thread** - exactly like the Compare Panel does.
**Implementation (Line ~39894-39996):**
```python
# TM PRE-TRANSLATION - Run on main thread (no SQLite threading issues)
# Uses the same database methods as the Compare Panel
if translation_provider_type == 'TM':
# Get activated TM IDs
tm_ids = self.tm_metadata_mgr.get_active_tm_ids(project_id)
# Create progress dialog for TM pre-translation
progress = QProgressDialog(...)
for idx, (row_index, segment) in enumerate(segments_needing_translation):
if tm_exact_only:
match = self.tm_database.get_exact_match(segment.source, tm_ids=tm_ids)
else:
matches = self.tm_database.search_all(segment.source, tm_ids=tm_ids, ...)
# Update segment and grid immediately
if segment.target:
target_widget.setPlainText(segment.target)
self.update_status_icon(row_index, segment.status)
QApplication.processEvents() # Keep UI responsive
return # TM pre-translation handled, exit
```
**Benefits of Main Thread Approach:**
- ✅ Uses exact same database connection as Compare Panel
- ✅ No SQLite threading errors
- ✅ Simpler code - no thread-local connection workarounds
- ✅ QProgressDialog with processEvents() keeps UI responsive
- ✅ Reliable and predictable behavior
**Files Modified:**
- `Supervertaler.py` - Rewrote TM pre-translation path (lines 39894-40004)
**Key Learning:**
When a background worker has SQLite threading issues, consider whether the operation really needs a background thread. For TM lookups (fast database queries), running on the main thread with periodic `processEvents()` is simpler and more reliable.
---
### January 22, 2026 - CRITICAL: Batch Translation Crash Investigation (v1.9.149-beta) - RESOLVED
**🚨 Multiple Crash Issues During Production Testing**
User discovered critical crashes when testing batch translation feature on 9-segment patent file. This session involved extensive debugging of three separate crash scenarios.
---
**CRASH #1: Empty TM IDs When No TMs Activated (FIXED)**
**Symptoms:**
- User imported new project → all resources auto-deactivated (by design)
- Started batch translate → immediate crash
- Error: Crash in TM database search with empty `tm_ids=[]`
**Root Cause:**
When no TMs are activated at all, `get_active_tm_ids()` returns empty list `[]`. The TM database search methods (`get_exact_match()`, `search_all()`) don't handle empty lists gracefully and crash.
**Fix Applied (Line ~39833):**
```python
# Skip TM pre-check if no TMs are activated
if tm_ids is None or (isinstance(tm_ids, list) and len(tm_ids) == 0):
self.log(f" Skipping TM pre-check (no activated TMs)")
segments_needing_translation = segments_to_translate
else:
# Check each segment against TM
for row_index, segment in segments_to_translate:
...
```
**Location:** Supervertaler.py, batch translate TM pre-check (line ~39807-39907)
---
**CRASH #2: User Had TMs Activated But Only "Write" Enabled (FIXED)**
**Symptoms:**
- User activated 2 TMs in Project Resources
- Started batch translate → still crashed
- Same error: empty `tm_ids=[]` despite TMs being activated
**User's Key Insight:**
> "wait, maybe they were set to write but not read"
**Root Cause Discovery:**
The TM activation system has TWO checkboxes:
- **"Read" checkbox**: Enables TM for searching/matching during translation
- **"Write" checkbox**: Enables saving new translations to TM
`get_active_tm_ids()` in `modules/tm_metadata_manager.py` **ONLY returns TMs where `ta.is_active = 1`** (Read checkbox enabled):
```python
# Line 366-399 in tm_metadata_manager.py
SELECT tm.tm_id FROM translation_memories tm
INNER JOIN tm_activation ta ON tm.id = ta.tm_id
WHERE ta.project_id = ? AND ta.is_active = 1 # <-- Read checkbox
```
When user had only "Write" enabled, method returned empty list `[]` → crash.
**Fix Applied (Line 41935-41941):**
Added same empty list validation to delayed TM search (segment navigation):
```python
# Skip TM search if no TMs are activated
if tm_ids is not None and isinstance(tm_ids, list) and len(tm_ids) == 0:
self.log(f"🚀 DELAYED TM SEARCH: Skipping (no TMs activated)")
all_tm_matches = []
else:
all_tm_matches = self.tm_database.search_all(...)
```
**Location:** Supervertaler.py, delayed TM search (line ~41920-41945)
---
**CRASH #3: Second Batch Translate Code Path Missing Fix (FIXED)**
**Symptoms:**
- After fixing crashes #1 and #2, user tested again
- Still crashed during batch translate
- Same error pattern but different code location
**Root Cause:**
There are **TWO separate batch translate code paths** in Supervertaler.py:
1. **Line ~39819**: First batch translate location (had the fix)
2. **Line ~40122**: Second batch translate location (MISSING the fix)
The second location was missing the empty `tm_ids` validation, causing crashes when it was the active code path.
**Fix Applied (Line 40140-40145):**
Added empty list validation to second batch translate location:
```python
# Skip TM pre-check if no TMs are activated
if tm_ids is None or (isinstance(tm_ids, list) and len(tm_ids) == 0):
self.log(f" Skipping TM pre-check (no activated TMs)")
segments_needing_translation = segments_to_translate
else:
# Check each segment against TM
for row_index, segment in segments_to_translate:
try:
if check_tm_exact_only:
...
```
**Indentation Issue:** Had to fix multiple indentation errors in the try/except blocks when adding the else clause.
**Location:** Supervertaler.py, second batch translate TM pre-check (line ~40122-40200)
---
**CRASH #4: Unknown Error During LLM Translation (CURRENT ISSUE - INVESTIGATING)**
**Symptoms:**
- After fixing all three crashes above, user tested with 2 TMs properly activated (Read checkbox enabled)
- TM pre-check completed successfully (found 0 matches, didn't crash)
- Batch translation started with Gemini API
- Crash during actual LLM translation inside PreTranslationWorker thread
- Console shows: "Unhandled Python exception" with no details
**Log Output Analysis:**
```
[LOG] ✓ Activated TM 42 for project 1947053659
[LOG] ✅ TM 42 set to readable
[LOG] ✓ Activated TM 52 for project 1947053659
[LOG] ✅ TM 52 set to readable
[LOG] 📊 Project has 8 segments → Will translate ALL of them
[LOG] 🚀 DELAYED TM SEARCH: Using TM IDs: ['patents', 'brants_carr_001_be_wo']
[Batch translate dialog appears, user clicks Confirm & Next]
Unhandled Python exception
```
**Problem:** QThread exceptions don't print to console by default - they're silently swallowed!
**Debug Logging Added (Line ~5416, ~5514, ~5291):**
Added comprehensive error logging to PreTranslationWorker:
1. **In `_translate_batch_with_llm()` method:**
```python
import traceback
print(f"🚀 _translate_batch_with_llm: Starting batch of {len(batch_segments)} segments")
print(f"🚀 Languages: {source_lang} → {target_lang}")
# ... at end of method ...
except Exception as e:
import traceback
error_details = traceback.format_exc()
print(f"❌ Batch LLM translation error: {e}")
print(f"❌ Full traceback:\n{error_details}")
```
2. **In `run()` method's batch processing:**
```python
except Exception as e:
import traceback
error_details = traceback.format_exc()
print(f"❌ Batch processing error: {e}")
print(f"❌ Full traceback:\n{error_details}")
```
**Next Steps:**
- User needs to run batch translate again with debug logging active
- Console will now show FULL Python traceback when crash occurs
- Once we see the exact error message and line number, we can fix the root cause
**Suspected Issues:**
- Gemini API authentication/key problem
- Missing attribute access in worker thread
- Threading race condition
- API response parsing error
---
**Summary of Fixes Applied:**
| Location | Issue | Status |
|----------|-------|--------|
| Line ~39833 | Empty tm_ids validation (first batch translate) | ✅ FIXED |
| Line 41935 | Empty tm_ids validation (delayed TM search) | ✅ FIXED |
| Line 40140 | Empty tm_ids validation (second batch translate) | ✅ FIXED |
| Line 5416 | Debug logging in _translate_batch_with_llm | ✅ ADDED |
| Line 5291 | Debug logging in run() batch processing | ✅ ADDED |
| Line 5316 | Wrong dict key 'target' → 'target_text' | ✅ FIXED |
| Line 38801 | Wrong dict key 'target' → 'target_text' | ✅ FIXED |
| Line 39860 | Wrong dict key 'target' → 'target_text' | ✅ FIXED |
| Line 40163 | Wrong dict key 'target' → 'target_text' | ✅ FIXED |
| Line 39761 | Store tm_exact_only for retry passes | ✅ FIXED |
| Line 39583 | Retrieve tm_exact_only in retry path | ✅ FIXED |
**Files Modified:**
- `Supervertaler.py` (47,639 lines total):
- Lines 39807-39907: First batch translate TM pre-check (added empty list check)
- Lines 40122-40200: Second batch translate TM pre-check (added empty list check + fixed indentation)
- Lines 41920-41945: Delayed TM search (added empty list check)
- Lines 5164-5550: PreTranslationWorker class (added debug logging)
- Line 39761: Store `_batch_tm_exact_only` for retry passes
- Line 39583: Retrieve `tm_exact_only` from stored setting in retry path
**Key Learnings:**
1. **TM Activation System:**
- "Read" checkbox: Enables TM for searching/matching
- "Write" checkbox: Enables saving translations to TM
- Users must enable "Read" to use TM during translation
- `get_active_tm_ids()` only returns "Read"-enabled TMs
2. **Empty List Handling:**
- Always validate `tm_ids` before calling TM database methods
- Empty list `[]` is different from `None`
- Need to check: `if tm_ids is None or (isinstance(tm_ids, list) and len(tm_ids) == 0):`
3. **QThread Exception Handling:**
- Exceptions in QThread don't print to console automatically
- Need explicit traceback logging in worker threads
- Use `import traceback` and `traceback.format_exc()` for full error details
4. **Multiple Code Paths:**
- Batch translate has TWO separate implementations
- When fixing bugs, search for ALL occurrences of similar patterns
- Use `grep_search` to find duplicate code paths
5. **Retry Pass Variable Scope:**
- When recursive calls skip dialog, locally-scoped variables from dialog are undefined
- Must store settings as instance variables (`self._batch_*`) before dialog
- Retry path must retrieve all needed settings with `getattr(self, '_batch_*', default)`
**Testing Checklist for Next Agent:**
- [ ] Run batch translate with no TMs activated → should log "Skipping TM pre-check"
- [ ] Run batch translate with TMs only "Write" enabled → should log "Skipping TM pre-check"
- [ ] Run batch translate with TMs "Read" enabled → should work normally
- [ ] Run batch translate → let retry trigger for empty segments → should not crash
- [ ] Check console for full error traceback when crash occurs
- [ ] Fix root cause based on traceback information
---
**CRASH #4: Wrong Dictionary Key in TM Exact Match (FIXED)**
**Symptoms:**
- User ran batch translate with 2 TMs properly activated (Read + Write enabled)
- TM pre-check crashed with: `'str' object has no attribute 'get'`
- Console showed repeated error for segments #3-9
- Error occurred when calling `exact_match.get('target', '')`
**Root Cause Discovery:**
The crash message was misleading - it wasn't that `exact_match` was a string, it was that `.get()` was being called on a string KEY!
**The Problem:**
- `get_exact_match()` returns raw database row as dict: `{'source_text': '...', 'target_text': '...', ...}`
- But the code was calling `exact_match.get('target', '')` - using wrong key name!
- Column name in database is **`target_text`**, not `target`
- When dict doesn't have 'target' key, `.get()` returns default empty string
- But the error message suggested the dict itself was the problem
**Database Column Names vs Return Keys:**
```python
# What get_exact_match() returns (raw database columns):
exact_match = {
'source_text': 'source text here', # ← Column name
'target_text': 'target text here', # ← Column name
'tm_id': 'patent_tm',
...
}
# What search_all() returns (transformed keys):
matches = [{
'source': 'source text here', # ← Transformed from source_text
'target': 'target text here', # ← Transformed from target_text
'match_pct': 100,
...
}]
```
**Why search_all() Worked:**
The `search_all()` method transforms the column names from `source_text`/`target_text` to `source`/`target` before returning. But `get_exact_match()` returns the raw database row unchanged!
**Fix Applied (4 locations):**
Changed `.get('target', '')` to `.get('target_text', '')` in:
1. Line 5316: PreTranslationWorker thread TM lookup
2. Line 39860: First batch translate TM pre-check
3. Line 40163: Second batch translate TM pre-check
4. Line 38801: Auto-confirm 100% matches (Ctrl+Enter navigation)
```python
# BEFORE (wrong key):
exact_match = self.tm_database.get_exact_match(segment.source, tm_ids=tm_ids)
if exact_match:
tm_match = exact_match.get('target', '') # ← Wrong key!
# AFTER (correct key):
exact_match = self.tm_database.get_exact_match(segment.source, tm_ids=tm_ids)
if exact_match:
tm_match = exact_match.get('target_text', '') # ← Correct database column name
```
**Location:** Supervertaler.py lines 5316, 38801, 39860, 40163
**Lesson Learned:**
Always check what keys are actually in the dictionary returned by database methods. Don't assume the keys match other methods' return formats!
---
**CRASH #5: Retry Pass Missing tm_exact_only Variable (FIXED)**
**Symptoms:**
- User ran batch translate, 7/8 segments translated successfully
- 1 segment remained empty, triggering retry mechanism
- Crash with: `UnboundLocalError: cannot access local variable 'tm_exact_only' where it is not associated with a value`
- Traceback pointed to line 39961: `tm_exact_only=tm_exact_only`
**Root Cause:**
When `handle_retry_needed()` calls `translate_batch()` recursively for retry:
1. The dialog is skipped (because `is_retry_pass = True`)
2. But `tm_exact_only` was only set from the dialog checkbox in the `else` block
3. The retry path retrieved other settings (`_batch_provider_type`, `_batch_model`) but NOT `tm_exact_only`
4. Result: Variable undefined when passed to `PreTranslationWorker`
**Fix Applied (2 locations):**
1. **Store setting** at line 39761: `self._batch_tm_exact_only = tm_exact_only`
2. **Retrieve setting** at line 39583: `tm_exact_only = getattr(self, '_batch_tm_exact_only', False)`
```python
# STORING (after dialog, line 39761):
self._batch_retry_enabled = retry_until_complete
self._batch_tm_exact_only = tm_exact_only # NEW: Store for retry passes
# RETRIEVING (retry path, line 39583):
if is_retry_pass:
translation_provider_type = getattr(self, '_batch_provider_type', 'LLM')
translation_provider_name = getattr(self, '_batch_provider_name', 'openai')
model = getattr(self, '_batch_model', 'gpt-4o')
tm_exact_only = getattr(self, '_batch_tm_exact_only', False) # NEW: Retrieve
```
**Result:** Retry mechanism now works correctly with stored settings from first pass.
---
**ARCHITECTURAL DECISION: TM Pre-Check REMOVED from Batch Translation (v1.9.150)**
After user testing confirmed batch AI translation and retry work correctly, but TM pre-check continued to show errors for every segment, the user made an architectural decision:
**User's Insight:**
> "pre-translating against a TM and translating using AI at the same time is asking for trouble"
**Decision:** Remove TM pre-check from AI batch translation entirely.
**Rationale:**
- Mixing TM lookup with AI translation in one operation adds complexity
- TM pre-translation is a separate workflow step
- User should run "Pre-translate from TM" first, THEN "Batch Translate" with AI
- Simpler, clearer, more maintainable
**Code Removed (~117 lines from `translate_batch()`):**
```python
# REMOVED - Was checking TM before API calls:
if translation_provider_type in ['LLM', 'MT']:
if check_tm_before_api and self.tm_database:
# ... TM lookup code for each segment ...
# REPLACED WITH simple pass-through:
segments_needing_translation = segments_to_translate
```
**Settings UI Updated (line ~15167):**
- Label changed: "Check TM before API call" → "Check TM before single-segment AI translation (Ctrl+T)"
- Clarifies the setting ONLY affects single-segment translation (Ctrl+T), not batch
**What Still Uses TM Check:**
- `translate_current_segment()` (Ctrl+T) - Single-segment translation can still check TM first
- Auto-confirm 100% matches during navigation (Ctrl+Enter) - Still uses TM lookup
**New Recommended Workflow for Users:**
1. **Import project** → All segments need translation
2. **Run "Pre-translate from TM"** (Edit → Batch Operations → Pre-translate from TM) → Fills in TM matches
3. **Run "Batch Translate"** with AI → Translates remaining empty segments
**Result:** Cleaner separation of concerns, simpler code, no more TM pre-check errors.
---
**Current State (v1.9.150):**
- ✅ All previous crash fixes in place
- ✅ TM pre-check REMOVED from batch translation
- ✅ Settings UI updated to clarify single-segment scope
- ✅ Syntax validated
- ✅ Ready for user testing
---
### January 21, 2026 - User-Choosable Data Folder (v1.9.148)
**📁 Your Data, Your Location**
Complete redesign of user data storage to give users full control over where their data lives.
**What Changed from v1.9.147:**
- v1.9.147 stored data in hidden system folders (AppData)
- v1.9.148 uses a visible folder in your home directory by default
- Users can now choose their own location on first run
**Default Data Locations:**
| Platform | Default Location |
|----------|-----------------|
| **Windows** | `C:\Users\Username\Supervertaler\` |
| **macOS** | `~/Supervertaler/` |
| **Linux** | `~/Supervertaler/` |
**Key Features:**
- **First-Run Dialog**: Choose your data folder when you first launch
- **Settings Integration**: Change location anytime via Settings → General → "📁 Data Folder Location"
- **Auto-Recovery**: If config pointer is deleted, app recovers by checking default location
- **Unified System**: Same behavior for pip users, EXE users, and developers
**How It Works:**
- Config pointer file stores your chosen path (in standard config location)
- Data stored in visible, easily-accessible folder
- Easy to backup - just copy the folder!
**Config Pointer Locations:**
- Windows: `%APPDATA%\Supervertaler\config.json`
- macOS: `~/Library/Application Support/Supervertaler/config.json`
- Linux: `~/.config/Supervertaler/config.json`
**Implementation:**
- Removed `platformdirs` dependency (no longer needed)
- New functions: `get_config_pointer_path()`, `get_default_user_data_path()`, `save_user_data_path()`, `load_user_data_path_from_config()`
- Rewrote `get_user_data_path()` with unified resolution logic
- Added `_show_data_location_dialog()` for first-run folder picker
- Added `_reinitialize_with_new_data_path()` for runtime location changes
**Files Modified:**
- `Supervertaler.py` - New data path system, first-run dialog, Settings integration
- `pyproject.toml` - Removed platformdirs dependency
- `CHANGELOG.md`, `README.md`, `docs/index.html`, `AGENTS.md` - Version updates
---
### January 21, 2026 - Persistent User Data Location (v1.9.147)
**📁 User Data Now Survives pip Upgrades**
Major enhancement to store user data in a platform-specific persistent location outside the pip package directory.
**The Problem:**
- pip installs wipe and replace the entire package directory on upgrade
- User data (API keys, TMs, glossaries, prompts) was stored inside that directory
- Every `pip install --upgrade supervertaler` wiped all user data
- Multiple user reports of lost API keys, TMs, and glossaries after updates
**The Solution (v1.9.147):**
User data stored in platform-specific hidden locations. Note: This was superseded by v1.9.148 which uses visible user-choosable locations instead.
**Implementation:**
- Added `platformdirs>=4.0.0` dependency for cross-platform paths
- Rewrote `get_user_data_path()` function with site-packages detection
- Added migration function with safety checks
**Files Modified:**
- `Supervertaler.py` - `get_user_data_path()`, `migrate_user_data_if_needed()`, init section
- `pyproject.toml` - Added platformdirs dependency
- `CHANGELOG.md`, `README.md`, `docs/index.html`, `AGENTS.md` - Version updates
---
### January 21, 2026 - Glossary Bug Fixes & Compare Panel Enhancement (v1.9.141-145)
**🐛 Glossary/Termview Bug Fixes**
Fixed chain of bugs related to glossary management after Alt+Down term addition:
| Version | Issue | Fix |
|---------|-------|-----|
| **v1.9.141** | Termview goes blank after adding term ("list index out of range") | Fixed dict key mismatch: `source_term`→`source`, `target_term`→`translation` in `_refresh_termbase_display_for_current_segment()` |
| **v1.9.142** | Edit glossary entry error ("on_cell_selected() missing 2 required positional arguments") | Simplified `_refresh_current_segment_matches()` to use `_refresh_termbase_display_for_current_segment()` instead of calling `on_cell_selected()` |
| **v1.9.143** | Delete glossary entry error ("'DatabaseManager' object has no attribute 'get_connection'") | Changed to use `self.termbase_mgr.delete_term(term_id)` instead of non-existent database method |
**Key Method - Data Structure:**
- `find_termbase_matches_in_source()` returns dict with keys: `source`, `translation`, `termbase_name`, `priority`, etc.
- NOT `source_term` / `target_term` as previously assumed
**🎨 Compare Panel memoQ-Style Diff (v1.9.144-145)**
Enhanced `_set_compare_panel_text_with_diff()` to show word-level differences like memoQ's "Track changes view":
| Element | Style |
|---------|-------|
| **Deletions** | Red text (#CC0000) with strikethrough |
| **Insertions** | Red text (#CC0000) with underline |
| **Unchanged** | Normal text |
**Implementation:**
- Uses `difflib.SequenceMatcher` for word-level diffing
- Words split by whitespace for readable diffs
- Applied to both TM Source and TM Target compare boxes
- Matches professional CAT tool behavior
**Files Modified:**
- `Supervertaler.py` - All bug fixes and Compare Panel enhancement
- `CHANGELOG.md` - Added v1.9.141-145 entries
---
### January 20, 2026 - Context Placeholders & Auto-Center Fix (v1.9.130)
**📝 Three Context Placeholders for QuickMenu**
Split `{{DOCUMENT_CONTEXT}}` into specialized variants:
| Placeholder | Mode | Use Case |
|-------------|------|----------|
| `{{SOURCE+TARGET_CONTEXT}}` | `mode="both"` | Proofreading (needs both to verify) |
| `{{SOURCE_CONTEXT}}` | `mode="source"` | Translation questions (source only) |
| `{{TARGET_CONTEXT}}` | `mode="target"` | Consistency/style analysis |
**Implementation:**
- `_build_quickmenu_document_context(mode)` now accepts "both", "source", or "target"
- `_quickmenu_build_custom_prompt()` detects and handles all three placeholders
- `modules/unified_prompt_manager_qt.py` Placeholders tab updated
**🎯 Auto-Center Active Segment Fix**
Fixed "Keep Active Segment Centered" using Qt's built-in `scrollTo()` with `PositionAtCenter`:
```python
index = self.table.model().index(row, 0)
self.table.scrollTo(index, QAbstractItemView.ScrollHint.PositionAtCenter)
```
- Added `QAbstractItemView` to imports
- Replaced unreliable manual viewport calculations
- Works consistently across all screen sizes
**⌨️ Double-Tap Shift Context Menu (AutoHotkey)**
- `superlookup_hotkey.ahk` → `supervertaler_hotkeys.ahk`
- Added double-tap Shift detection for context menu (Supervertaler window only)
- Uses `Shift+F10` (Qt's native context menu trigger)
**Files Modified:**
- `Supervertaler.py` - Context placeholders, auto-center, AHK references
- `modules/unified_prompt_manager_qt.py` - Placeholders tab
- `modules/shortcut_manager.py` - Double-shift shortcut docs
- `supervertaler_hotkeys.ahk` - Combined hotkey script (renamed)
---
### January 19, 2026 - Prompt System Improvements (v1.9.126)
**🔄 Field Rename: `quickmenu_quickmenu` → `sv_quickmenu`**
Renamed redundant field throughout the codebase for cleaner API:
**What Changed:**
- All internal code now uses `sv_quickmenu` (Supervertaler QuickMenu)
- Updated in both `unified_prompt_library.py` and `unified_prompt_manager_qt.py`
- 18+ occurrences renamed across parsing, saving, display, and toggle methods
- Backward compatibility: Old .svprompt files still load correctly
- Legacy `quick_run` field kept in sync for compatibility
**Files Modified:**
- `modules/unified_prompt_library.py` - Parse, save, toggle methods
- `modules/unified_prompt_manager_qt.py` - Editor, creation, display code
- `Supervertaler.py` - Version bump to 1.9.126
- `CHANGELOG.md` - Added v1.9.126 entry
**📝 Placeholders Reference Tab**
Added dedicated reference tab in Prompt Manager showing all available placeholders:
**Features:**
- Table format: Placeholder | Description | Example
- All 5 placeholders documented:
* `{{SELECTION}}` - Selected text in grid
* `{{SOURCE_TEXT}}` - Full source segment
* `{{SOURCE_LANGUAGE}}` - Project source language
* `{{TARGET_LANGUAGE}}` - Project target language
* `{{DOCUMENT_CONTEXT}}` - Project segments (configurable %)
- Usage tips section with best practices
- Located after AI Assistant tab
**Implementation:**
- New `_create_placeholders_tab()` method (~120 lines)
- Added as Tab 3 in Prompt Manager sub-tabs
- Uses QTableWidget with 3 columns and 5 rows
- Monospace font for placeholders and examples
- Adjustable column widths for readability
**Files Modified:**
- `modules/unified_prompt_manager_qt.py` - New tab method, added to sub-tabs
---
### January 19, 2026 - QuickMenu Document Context (v1.9.124)
---
### January 19, 2026 - Prompt Manager UI Fixes (v1.9.127)
**🔧 Save Button Fix**
Fixed issue where Save button remained greyed out after creating new prompts:
- **Problem**: User creates new prompt → editor loads content → Save button stays disabled
- **Root Cause**: `_new_prompt_in_folder()` called `_load_prompt_in_editor()` which should enable button, but timing issue prevented it
- **Solution**: Added explicit `btn_save_prompt.setEnabled(True)` call immediately after loading prompt in editor
- **Result**: Save button now properly enabled for all new prompts
**📝 Label Rename for Clarity**
Renamed QuickMenu checkbox label:
- **Before**: "Show in QuickMenu"
- **After**: "Show in Supervertaler QuickMenu"
- **Reason**: Distinguishes app-level QuickMenu from Grid right-click QuickMenu
**Files Modified:**
- `modules/unified_prompt_manager_qt.py` - Save button enable, checkbox label
---
### January 19, 2026 - Prompt System Improvements (v1.9.126)
**🔄 Field Rename: `quickmenu_quickmenu` → `sv_quickmenu`**
Renamed redundant field throughout the codebase for cleaner API:
**What Changed:**
- All internal code now uses `sv_quickmenu` (Supervertaler QuickMenu)
- Updated in both `unified_prompt_library.py` and `unified_prompt_manager_qt.py`
- 18+ occurrences renamed across parsing, saving, display, and toggle methods
- Backward compatibility: Old .svprompt files still load correctly
- Legacy `quick_run` field kept in sync for compatibility
**Files Modified:**
- `modules/unified_prompt_library.py` - Parse, save, toggle methods
- `modules/unified_prompt_manager_qt.py` - Editor, creation, display code
- `Supervertaler.py` - Version bump to 1.9.126
- `CHANGELOG.md` - Added v1.9.126 entry
**📝 Placeholders Reference Tab**
Added dedicated reference tab in Prompt Manager showing all available placeholders:
**Features:**
- Table format: Placeholder | Description | Example
- All 5 placeholders documented:
* `{{SELECTION}}` - Selected text in grid
* `{{SOURCE_TEXT}}` - Full source segment
* `{{SOURCE_LANGUAGE}}` - Project source language
* `{{TARGET_LANGUAGE}}` - Project target language
* `{{DOCUMENT_CONTEXT}}` - Project segments (configurable %)
- Usage tips section with best practices
- Located after AI Assistant tab
**Implementation:**
- New `_create_placeholders_tab()` method (~120 lines)
- Added as Tab 3 in Prompt Manager sub-tabs
- Uses QTableWidget with 3 columns and 5 rows
- Monospace font for placeholders and examples
- Adjustable column widths for readability
**Files Modified:**
- `modules/unified_prompt_manager_qt.py` - New tab method, added to sub-tabs
---
### January 19, 2026 - QuickMenu Document Context (v1.9.124)
**📄 Context-Aware AI Suggestions**
Implemented major enhancement allowing QuickMenu prompts to access full project context for better AI suggestions:
**The Feature:**
- New `{{DOCUMENT_CONTEXT}}` placeholder for QuickMenu prompts
- Configurable percentage slider (0-100%, default 50%) in Settings → AI Settings
- Safety limit: Maximum 100 segments to prevent token overload
- Format: `[ID] source\n → target\n\n` with header showing segment count/percentage
**Implementation:**
- `_build_quickmenu_document_context()` - Builds formatted segment list
- Enhanced `_quickmenu_build_custom_prompt()` - Replaces placeholder with context
- Settings UI: Horizontal slider with dynamic value label
- Pattern reuse: Adapted from batch translation "surrounding segments" feature
**Example Use Case:**
```
{{DOCUMENT_CONTEXT}}
Suggest the best possible translation of "{{SELECTION}}" from {{SOURCE_LANGUAGE}} to {{TARGET_LANGUAGE}} within the context of the current project shown above.
```
**Benefits:**
- ✅ AI understands project domain and terminology
- ✅ Consistent translations across the document
- ✅ Better handling of ambiguous terms
- ✅ Context-aware suggestions for specialized fields
**Files Modified:**
- `Supervertaler.py` - New method, enhanced prompt builder, settings UI, save wiring
---
### January 19, 2026 - QuickMenu Generic AI Support (v1.9.123)
**🤖 Fixed QuickMenu Translation Mode Lock**
Fixed critical bug where QuickMenu prompts were being forced into translation mode:
**The Problem:**
- QuickMenu was calling `client.translate(text=input_text, ...)` which forced translation behavior
- Generic prompts like "Explain this", "Define the selection" would fail
- The AI would translate the prompt itself instead of executing it
**The Fix:**
- Changed to use generic AI completion: `client.translate(text="", custom_prompt=...)`
- Simplified prompt builder - removed translation-specific wrappers
- QuickMenu now supports ANY AI task (explain, define, suggest, analyze)
**Placeholders Available:**
- `{{SELECTION}}` - Selected text
- `{{SOURCE_TEXT}}` - Full source segment
- `{{SOURCE_LANGUAGE}}` - Project source language
- `{{TARGET_LANGUAGE}}` - Project target language
- `{{DOCUMENT_CONTEXT}}` - Full project context (v1.9.124)
**Files Modified:**
- `Supervertaler.py` - `run_grid_quickmenu_prompt()`, `_quickmenu_build_custom_prompt()`
---
### January 19, 2026 - Ctrl+N for Quick Notes (v1.9.122)
**⌨️ Shortcut Repurposed for Note-Taking**
Repurposed Ctrl+N from "New Project" (rarely used) to "Focus Segment Note tab":
**The Feature:**
- Press Ctrl+N to instantly jump to Segment Note tab
- Cursor placed in notes field, ready to type
- Perfect for quick proofreading notes, context reminders, translation decisions
**Implementation:**
- `modules/shortcut_manager.py` - Changed `file_new` default to "" (empty), added `editor_focus_notes` with Ctrl+N
- `Supervertaler.py` - New `focus_segment_notes()` method switches to tab index 1, focuses `bottom_notes_edit`
---
### January 19, 2026 - API Key Loading System Unified (v1.9.113)
**🔐 Unified API Key Loading with Dev-First Priority**
Consolidated the confusing multi-path API key loading system into a single, clear dual-path approach:
**The Problem:**
- Three different API key file locations existed (root, user_data, user_data_private)
- Two different loading mechanisms (`Supervertaler.load_api_keys()` vs `llm_clients.load_api_keys()`)
- Conflicting instructions in example files
- AI Assistant bug (#107): Keys worked for translation but failed for AI Assistant
**The Solution:**
- **Unified loading in main app**: `load_api_keys()` now checks TWO locations with clear priority
1. `user_data_private/api_keys.txt` (Dev mode - gitignored, never uploaded to GitHub)
2. `user_data/api_keys.txt` (User mode - ships with app)
- **AI Assistant fixed**: Now uses `parent_app.load_api_keys()` instead of module function
- **Example files updated**: Both example files now give consistent, clear instructions
**Developer Workflow:**
- Store keys in `user_data_private/api_keys.txt`
- Fully gitignored - safe from accidental commits
- All features find keys here (translation, AI Assistant, tests)
**User Workflow:**
- Keys go in `user_data/api_keys.txt`
- App auto-creates this location on first run
- Simple, single location
**Files Modified:**
- `Supervertaler.py` - `load_api_keys()` method now checks dev path first (line ~39407)
- `api_keys.example.txt` - Updated with dev/user instructions
- `user_data/api_keys.example.txt` - Updated with dev/user instructions
- `AGENTS.md` - Updated API Keys section with new dual-path documentation
**Result:**
- ✅ Developers: Keys safe in gitignored location
- ✅ Users: Simple single location
- ✅ AI Assistant: Now works with same keys as translation
- ✅ No more confusion about where to put keys
---
### January 19, 2026 - Version 1.9.112: Critical Bug Fixes
**🐛 Filter Pagination Bug Fixed**
Fixed critical bug where Filter Source/Target boxes only searched visible page instead of all segments:
- **The Problem**: When pagination was active (e.g., "50 per page"), filtering only searched through the currently visible rows in the table
- **Root Cause**: `apply_filters()` used `self.table.rowCount()` which could be limited by pagination state or previous filtering
- **The Fix**: Changed to `len(segments)` to always iterate through ALL segments in the project
- **User Impact**: Filtering now finds matches across the entire project regardless of which page you're viewing
**Technical Details:**
- Changed iteration from `for row in range(row_count)` where `row_count = self.table.rowCount()`
- To: `for row in range(total_segments)` where `total_segments = len(segments)`
- The filter builds a complete `matching_rows` set, then `_apply_pagination_to_grid()` handles combined visibility
**📝 Bilingual Table Export - Notes Column**
Fixed segment notes not being exported to Supervertaler Bilingual Table DOCX files:
- **The Problem**: Notes column was hardcoded to empty string: `cells[4].text = ''`
- **The Fix**: Now properly exports `seg.notes` from each segment
- **Formatting**: 8pt font to match Status column styling
- **Includes**: Proofreading notes (⚠️ PROOFREAD prefix), user notes, all segment annotations
**📏 Grid Column Width Optimization**
Reduced segment ID column width for more compact display:
- **Before**: 55px (unnecessarily wide even for 4-digit segment numbers)
- **After**: 40px (fits up to 3 digits comfortably, readable for 4+ digits)
- **Benefit**: More horizontal space available for Source/Target columns
**Files Modified:**
- `Supervertaler.py` - Fixed `apply_filters()` iteration logic, notes export in `_export_review_table()`, column width in grid initialization
---
### January 21, 2026 - Prompt Library Reorganization (v1.9.148-beta)
**📁 QuickMenu Prompt Organization & Title Case Folders**
Reorganized the prompt library structure for better menu appearance and user experience:
**Changes Made:**
1. **Moved QuickMenu prompts** from `quickmenu_prompts/` to `Translation Help/` folder
- Define.svprompt
- Explain (in general).svprompt
- Explain (within project context).svprompt
- Translate selection in context of current project.svprompt
2. **Deleted empty example folders**: Web Searches/, Text Processing/, quickmenu_prompts/
3. **Renamed to Title Case**: `proofreading/` → `Proofreading/`
**Final Structure** (all title case for menu appearance):
```
user_data_private/prompt_library/
├── Domain Expertise/
├── Proofreading/
├── Project Prompts/
├── Style Guides/
└── Translation Help/ ← QuickMenu prompts here
```
**Design Decision**: Folder names use title case because they will become menu names in the future QuickMenu system. Action-oriented naming (Translation Help, Proofreading, Style Guides) creates intuitive menu hierarchies.
**Files Modified:**
- `user_data_private/prompt_library/` - Folder reorganization (gitignored)
- `AGENTS.md` - Documentation update
---
### January 18, 2026 - TMX Language Pair Bug Fix (User Issue #105) - v1.9.109
**🔧 Fixed Critical TMX Import Language Reversal Bug**
**Issue ([#105](https://github.com/michaelbeijer/Supervertaler/issues/105)):** User reported that EN-GB → DE-DE TMX files were being imported as DE-DE → EN-GB, making it impossible to find matches for translated segments.
**Root Cause:**
- TMX import code incorrectly assumed the FIRST language in the TMX file was the source language and the SECOND was the target
- TMX files list languages in arbitrary order (often alphabetically), so this assumption was wrong
- Example: A file with languages listed as [de-DE, en-GB] would be imported as DE-DE → EN-GB even if the user wanted EN-GB → DE-DE
**Fix Implemented:**
- Added language pair selection dialog when importing TMX files
- User now explicitly selects which detected language should be source and which should be target
- Prevents accidental language reversal regardless of TMX file language order
- Applied to both "Create new TM from TMX" and "Add to existing TM" workflows
**User Workflow:**
1. Import TMX file via TM Manager
2. Dialog shows all detected languages (e.g., "de-DE, en-GB")
3. User selects: Source = en-GB, Target = de-DE
4. Import proceeds with correct language pair
5. TM matches now work correctly
**Implementation Details:**
- Added language selection dialog with two dropdowns (Source/Target)
- Validation ensures source ≠ target
- Dialog appears in two locations in `_import_tmx_as_tm()`:
- When creating new TM from TMX (line ~12852)
- When adding to existing TM with no languages set (line ~12957)
**Files Modified:**
- `Supervertaler.py` - Added language selection dialog in `_import_tmx_as_tm()` method (~110 lines added)
- `CHANGELOG.md` - Added v1.9.109 entry
- `README.md` - Version updated to v1.9.109
**GitHub:**
- Issue: https://github.com/michaelbeijer/Supervertaler/issues/105
- Response: `GitHub_Issue_105_Response.md`
---
### January 18, 2026 - memoQ XLIFF Import/Export Support (User Issue #106) - v1.9.108
**📥📤 Complete memoQ XLIFF (.mqxliff) Workflow**
Added full import/export support for memoQ XLIFF files - feature was implemented in module but never exposed in UI:
**User Issue:**
- User reported inability to import memoQ XLIFF files
- Slovak language support question (already supported but needed clarification)
- Handler module (`modules/mqxliff_handler.py`) existed but no menu items
- memoQ bilingual DOCX import had limited language detection (only 8 languages)
**Implementation:**
1. **Import Menu Item**: File → Import → memoQ XLIFF (.mqxliff)...
- Opens file dialog for `.mqxliff` files
- Automatically extracts source segments using `MQXLIFFHandler`
- Converts ISO language codes to full names (`sk` → `Slovak`)
- Stores handler and source path for round-trip export
2. **Export Menu Item**: File → Export → memoQ XLIFF - Translated (.mqxliff)...
- Updates target segments in original XLIFF structure
- Preserves formatting tags (bpt/ept pairs)
- Saves translated file with proper namespace handling
3. **Language Code Normalization**:
- New `_normalize_language_code()` method
- Converts ISO 639-1/639-2 codes to full language names
- Supports 30+ languages including Slovak (`sk`, `sk-SK`)
4. **memoQ Bilingual DOCX Language Detection**:
- Expanded `lang_map` in `import_memoq_bilingual()` from 8 to 24 languages
- Now includes Slovak, Czech, Hungarian, Romanian, Bulgarian, Greek, Russian, Ukrainian, Swedish, Danish, Finnish, Norwegian, Japanese, Chinese, Korean, Arabic, Turkish, Hebrew
- Fixes bug where Slovak would default to EN→NL instead of being detected
5. **Project Persistence**:
- Added `mqxliff_source_path` field to `Project` dataclass
- Source path saved in `.svproj` files
- Automatic handler restoration when loading projects
**Methods Added:**
- `import_memoq_xliff()` - Import XLIFF files (~90 lines)
- `export_memoq_xliff()` - Export with translations (~120 lines)
- `_normalize_language_code()` - ISO code → full name converter (~60 lines)
**Round-Trip Workflow:**
1. Export from memoQ as XLIFF
2. Import into Supervertaler
3. Translate segments
4. Export back to XLIFF
5. Import into memoQ
**Files Modified:**
- `Supervertaler.py` - Import/export menu items, methods, language normalization
- `Supervertaler.py` - Project dataclass: added `mqxliff_source_path` field
- `Supervertaler.py` - Project save/load: persist mqxliff_source_path
**GitHub Discussion:**
- https://github.com/michaelbeijer/Supervertaler/discussions/106
---
### January 15, 2026 - Version 1.9.107: Prompt Library & Superlookup Fixes
**🔧 Prompt Library Improvements**
Unified filename and Name field throughout the prompt library:
- **Tree Display**: Now shows full filename with `.svprompt` extension (e.g., "prompt.svprompt")
- **Editor Name Field**: Shows full filename including extension (was previously metadata-only)
- **File Operations**: Name field now represents the actual filename - editing it renames the file on disk
- **New Prompt Dialog**: Asks for "filename with extension" and auto-appends `.svprompt` if missing
- **Result**: One unified concept where filename = what you see = what you edit everywhere
**Files Modified:**
- `modules/unified_prompt_manager_qt.py` - Tree display, editor field, save logic, new prompt dialog
**🔍 Superlookup Navigation Fixes**
Fixed two critical navigation bugs after Supermemory removal:
1. **Ctrl+K AttributeError Fixed**:
- Removed orphaned Supermemory code from SuperlookupTab
- `search_supermemory()` now gracefully returns 0 instead of trying to access removed engine
- Removed unused methods: `create_supermemory_results_tab()`, `init_supermemory()`, `on_supermemory_result_double_click()`, `copy_selected_supermemory_target()`
- ~140 lines of dead code removed
2. **Ctrl+K Tab Navigation Fixed**:
- Updated `_go_to_superlookup()` to use correct tab index
- After Prompt Manager tab was added, indices shifted: Tools moved from index 2 → 3
- Ctrl+K was opening Prompt Manager instead of Superlookup
- Now correctly navigates to Tools tab (index 3)
**Tab Order:**
- 0 = Grid
- 1 = Resources
- 2 = Prompt Manager
- 3 = Tools (Superlookup is here)
- 4 = Settings
**Files Modified:**
- `Supervertaler.py` - Removed Supermemory code, fixed tab navigation
---
### January 15, 2026 - 🗑️ Supermemory Removed (v1.9.105)
**Strategic Refactoring: Remove Supermemory Entirely**
After extensive attempts to package Supermemory (vector-indexed semantic search) in Windows EXE builds, made the decision to remove it entirely from the project:
**Problems Encountered:**
- PyTorch/sentence-transformers has complex native dependencies (CUDA libs, Intel MKL, OpenMP, etc.)
- Multiple Windows DLL loading failures (`vcomp142.dll`, `c10.dll`, etc.) even after comprehensive packaging
- 7+ packaging iterations failed to produce a working frozen build
- ~600 MB footprint for a feature that didn't work reliably
**Strategic Decision:**
- Focus development effort on the **SQLite-based Translation Memory** system instead
- SQLite TM is:
- Faster and more reliable for professional translation workflows
- Works perfectly in frozen builds
- Easier to maintain and improve
- Better suited for large TMX imports (thousands of entries)
**Removed:**
- `modules/supermemory.py` (2100+ lines deleted)
- Supermemory tab from Resources
- Auto-init and cleanup code in main application
- Dependencies: `sentence-transformers`, `chromadb`, `tokenizers`
- Pip extra: `supervertaler[supermemory]`
- Feature manager entry
- Build spec references
**Files Modified:**
- `Supervertaler.py` - Removed create_supermemory_tab(), _auto_init_supermemory(), _cleanup_supermemory(), UI tab
- `modules/feature_manager.py` - Removed supermemory from FEATURE_MODULES
- `pyproject.toml` - Removed supermemory extra and heavy dependencies
- `Supervertaler.core.spec` / `Supervertaler.full.spec` - Updated comments (FULL now just for Local Whisper)
- `CHANGELOG.md` - Documented removal in v1.9.105
**Result:**
- Cleaner codebase
- ~600 MB smaller default installation
- Focus on what works: SQLite TM, LLMs, termbases, voice commands
- No more Windows EXE packaging headaches with PyTorch
---
### January 18, 2026 - Version 1.9.111: Clean Slate Project Imports
**🔒 Automatic Resource Deactivation on New Project Import**
Implemented automatic deactivation of all resources (TMs, glossaries, NT lists) when importing new projects:
**The Problem:**
- When importing a new project, all previously activated resources remained active
- Users could end up with dozens of unrelated TMs/glossaries active across projects
- No clean separation between project contexts
- Difficult to tell which resources were actually relevant to current project
**The Solution:**
- Added `_deactivate_all_resources_for_new_project()` method
- Automatically called after import in all 5 handlers: DOCX, TXT, memoQ, CafeTran, Trados
- Deactivates all TMs via `tm_metadata_mgr.deactivate_tm()`
- Deactivates all glossaries via `termbase_mgr.deactivate_termbase()`
- Deactivates all NT lists via `nt_manager.set_list_active(False)`
**User Workflow:**
1. Import new project → All resources automatically deactivated
2. Go to Project Resources tab → Activate only needed TMs/glossaries
3. Work on project with clean, relevant resource set
4. Import next project → Clean slate again
**Benefits:**
- ✅ Fresh start for every project
- ✅ Prevents unintended resource pollution
- ✅ Users explicitly choose what's relevant
- ✅ Clearer project context separation
- ✅ Consistent behavior across all import types
**Files Modified:**
- `Supervertaler.py` - Added `_deactivate_all_resources_for_new_project()`, called in 5 import handlers
---
### January 15, 2026 - 🏗️ Windows EXE: CORE + FULL release pipeline (v1.9.104)
- Added/used dual build flavors: CORE (no heavy ML stack) and FULL (bundles Supermemory + heavy deps).
- Important runtime rule (PyInstaller one-folder): users must run `Supervertaler.exe` from the extracted folder next to `_internal/` (do not move the EXE).
- FULL launch crash mitigation notes: avoid importing heavy deps during startup/feature detection; disable UPX for FULL; exclude problematic `VCOMP140.DLL` from the bundle when it causes `0xc0000409` startup crashes.
- Supermemory-in-FULL packaging note: do not exclude `torch.cuda` Python modules (CPU use can still touch those imports); keep `triton` excluded if needed.
- **Critical:** Do NOT exclude `unittest` - it's a standard library module needed by sentence-transformers/tokenizers. The spec initially had it in excludes list which caused "No module named 'unittest'" error.
- **Critical:** Add `tiktoken` to the `collect_submodules()` package list - sentence-transformers needs `tiktoken140.dll` which won't be bundled otherwise.
- **Critical:** Use `collect_dynamic_libs()` for packages with native extensions - added explicit DLL collection for `tokenizers`, `safetensors`, and `torch` to ensure all `.dll`/`.pyd` files are bundled (sentence-transformers depends on native code from all three).
- **Critical:** Use `collect_data_files("torch")` to bundle ALL torch lib files, not just DLLs - PyTorch has complex directory structure with many dependent files.
- **Exit crash fix:** Added `_cleanup_supermemory()` method in closeEvent to explicitly shut down ChromaDB client and clear PyTorch references before exit, preventing "Python has stopped working" crash on Windows.
- Artifact hygiene: keep only the two intended release zips in `dist/` (`Supervertaler-v1.9.104-Windows-CORE.zip` + `...-FULL.zip`) to avoid uploading/testing stale builds.
### January 14, 2026 - 📦 Packaging: Lighter Default Install (v1.9.104)
- Made Supermemory an optional install extra again so `pip install supervertaler` no longer pulls the heavy ML stack by default.
- Install on demand with `pip install supervertaler[supermemory]`.
### January 14, 2026 - ✅ Filtered Ctrl+Enter + Website Screenshots (v1.9.103)
- Fixed Ctrl+Enter under active Filter Source/Target: confirm + advance stays within the filtered set (no more grid “unfiltering” while the filter text remains).
- Website: added new screenshots for Compare Panel + Termview and updated the Prompt Manager screenshot.
### January 14, 2026 - ⚡ QuickMenu (Prompt actions in Grid)
- Renamed main tab label: "📝 Project editor" → "📝 Grid"
- Introduced "QuickMenu" prompt metadata (replacing "Quick run menu" terminology):
- `quickmenu_label` (menu display label)
- `quickmenu_grid` (show in Grid right-click QuickMenu)
- `quickmenu_quickmenu` (show in future app-level QuickMenu; legacy-synced with `quick_run`)
- Prompt Manager UI now exposes QuickMenu label + both inclusion toggles
- Grid right-click menus (Source + Target) now include a "⚡ QuickMenu" submenu:
- Run selected prompt and show response dialog
- Run selected prompt and replace selection/target
### January 13, 2026 - 📝 Update Checker Investigation (Work in Progress)
**Goal:** Add a Help → “🔄 Check for Updates…” action that tells users whether they’re on the latest release.
**Implementation Attempt (PyQt6/QtNetwork):**
- Implemented an async update-check flow against GitHub:
- API: `https://api.github.com/repos/michaelbeijer/Supervertaler/releases/latest`
- Fallback: `https://github.com/michaelbeijer/Supervertaler/releases/latest`
- Hardened against common Qt pitfalls:
- Avoided `QProgressDialog` auto-close/auto-reset surprises
- Added request-id guarding to ignore stale replies
- Prevented double-cleanup (double `finish()` calls)
- Added logging for network/SSL errors
**What we observed on the developer machine:**
- QtNetwork consistently returns `Unknown error` (no usable error string) for both API and fallback URLs.
- The dialog could hang far longer than intended due to overlapping timeout/fallback paths.
- This appears to be an environment-dependent Qt SSL/TLS backend issue (Windows).
**Temporary Decision (to avoid UX pain):**
- `Help → Check for Updates…` currently just opens the GitHub Releases page:
- `https://github.com/michaelbeijer/Supervertaler/releases/latest`
- This avoids hangs/crashes while we revisit a robust implementation later.
**Future direction (when revisiting):**
- Prefer a Python HTTPS-based check (`urllib`/`requests`) in a worker thread, or a tiny lightweight update endpoint.
- If keeping QtNetwork, validate Qt SSL backend availability and provide clear guidance when SSL is missing.
### January 13, 2026 - 🌐 New Website Planning: `michaelbeijer.co.uk` (NEW repo)
**Context / Goal:**
Build a new professional personal website for Michael Beijer (Dutch/English patent + technical translator & terminologist). Current site is MediaWiki-based at `https://michaelbeijer.co.uk/`; the goal is to replace it with a modern static site once complete.
**Brand / Domain strategy (agreed):**
- Keep the *personal* brand on `michaelbeijer.co.uk` (and keep email on that domain: `info@michaelbeijer.co.uk`).
- Keep product/project brands on their own domains:
- `https://supervertaler.com` (CAT tool)
- `https://beijerterm.com` (open-source multilingual terminology database)
- Optional: `beijer.uk` can exist as a vanity domain and 301-redirect to `michaelbeijer.co.uk` (recommended rather than migrating the main domain).
**Aesthetic / Inspiration:**
- The user wants a crisp, black-and-white, content-first aesthetic similar to Pagefind docs:
- Reference: `https://pagefind.app/docs/multilingual/`
- Strongly avoid garish colors; typography + whitespace + hairline borders.
**Key identity & contact details (provided by user):**
- Michael Beijer — Dutch/English translator & terminologist — Hastings, United Kingdom
- Tel: `+44 7475 771720`
- Email: `info@michaelbeijer.co.uk`
- Websites:
- Main: `michaelbeijer.co.uk`
- Terminology DB: `beijerterm.com`
- CAT tool: `supervertaler.com`
- Profiles:
- ProZ: profile + testimonials source (Feedback Card): `https://www.proz.com/feedback-card/652138`
- LinkedIn: `michaelbeijer`
**Single-page vs multi-page decision:**
- Recommended: a minimal multi-page site (still “docs-like” and simple), because Michael has multiple distinct offerings (patents/translation, copywriting, CAT consulting) + products (Supervertaler, Beijerterm).
**Information Architecture (proposed):**
- Home (positioning + proof + CTA)
- Patents (primary specialization: process, QA, confidentiality, what clients should send)
- Services (technical translation, editing/proofreading, copywriting, CAT/terminology consulting)
- Work (case studies)
- Testimonials (ProZ + selected quotes)
- Tools (side projects): Supervertaler, Beijerterm
- Blog (wanted; current blog exists at `https://michaelbeijer.co.uk/blog`)
- Contact (email-first CTA)
**CTA (agreed):**
- Primary CTA is email (not booking link): “Email me” with a short checklist (word count, deadline, source format, subject/domain).
**Case studies + testimonials: how to combine (agreed approach):**
- Use “Proof” as an umbrella concept:
- *Case studies* = structured “problem → approach → outcome” writeups.
- *Testimonials* = short curated quotes (primarily from ProZ) + link back to ProZ.
- Patents often involve NDAs; case studies should generally be anonymized:
- No client names/logos unless explicit permission.
- Avoid application/publication identifiers unless already public and acceptable.
- Keep focus on domain, constraints, workflow, QA strategy, measurable outcomes.
**ProZ testimonials handling (important):**
- Do NOT auto-scrape ProZ.
- Curate a small selection manually (short excerpts) and include a prominent link to the ProZ feedback card.
- Store testimonials as local data (YAML/JSON) so they can be reused on Home + Testimonials pages.
**Recommended tech stack for the new repo:**
- Astro (static site generation, component-based)
- Tailwind CSS (consistent type/spacing, minimal custom CSS)
- Markdown/MDX for Blog + Case studies
- Optional later: Pagefind for fully static site search (fits the desired “docs” vibe)
**Repo plan:**
- New GitHub repo name: `michaelbeijer.co.uk`.
- During development, the site can be “parked” locally/in a separate folder; once finished, point `michaelbeijer.co.uk` DNS to GitHub Pages (or another static host).
- Prefer keeping this separate from the Supervertaler repo and separate from the `beijerterm/` submodule.
**Design rules-of-thumb (to match the desired vibe):**
- Grayscale palette; at most one subtle accent for links.
- Narrow reading column (target ~70–80 characters) + generous whitespace.
- Hairline borders instead of heavy cards/shadows.
- Minimal nav, fast scanning; let headings + spacing do the work.
**Open questions (next agent can confirm if needed):**
- Initial blog migration: import which posts from the current MediaWiki blog, and in what format (manual copy vs export).
- Whether any named clients/published patents can be referenced explicitly (if yes, add a “Selected clients / published work” section).
### January 13, 2026 - ✅ FIXED: Ctrl+Return Not Working in Source Cell
**Issue:** Ctrl+Enter (main keyboard Return key) does not trigger `confirm_selected_or_next()` when cursor is in source cell (`ReadOnlyGridTextEditor`).
**Key Finding:** Ctrl+Enter on **numpad** (Key_Enter) WORKS, but main keyboard Return (Key_Return) does NOT.
**What We Observed:**
1. ✅ Global QShortcut IS registered: `key=Ctrl+Enter, enabled=True, context=ApplicationShortcut`
2. ❌ Widget's `keyPressEvent()` is NEVER called for main keyboard Return
3. ❌ Widget's `event()` override is NEVER called for main keyboard Return
4. ❌ Debug print statements produce NO output when pressing Ctrl+Return
**Likely Root Cause:**
On Windows/Qt, `Ctrl+Return` (Key_Return) can be swallowed before it reaches the cell widget, and the `QShortcut(QKeySequence("Ctrl+Return"))` path appears to effectively only catch numpad Enter (Key_Enter) in this scenario.
**Suspected Areas:**
- QTableWidget default key handling (line 18352: `setEditTriggers(DoubleClicked)`)
- Possible event filter on table or viewport
- Qt's internal focus/key event routing for QTextEdit inside QTableWidget cells
- Possible conflict with recent "Auto Resize Rows" feature
**Fix Implemented:**
- Added an application-level event filter that catches `Ctrl+Return` and `Ctrl+Enter` *before* Qt/table routing can swallow it, but only when focus is inside the grid.
- The filter calls `confirm_selected_or_next()` and returns `True` to stop further handling.
- Expanded the same handling so it also works when focus is in the **Filter Source** / **Filter Target** boxes.
**Implementation Notes:**
- `Supervertaler.py`: new `_CtrlReturnEventFilter(QObject)` near the grid editor classes.
- `Supervertaler.py`: installed via `QApplication.instance().installEventFilter(...)` in `setup_global_shortcuts()`.
- Kept `QShortcut` binding for `editor_save_and_next`, but the event filter ensures main-keyboard Return also works.
**Workaround (no longer needed):** Previously: use numpad Enter key instead of main keyboard Return.
---
### January 12, 2026 - Version 1.9.99: Compare Panel Shortcuts + Sound Effects
- **Compare Panel quick insert**: Added `Alt+0` (MT) and `Alt+0,0` (TM Target) insertion; replaces entire target segment in one undo step.
- **Compare Panel navigation**: Added `Ctrl+Alt+Left/Right` for MT and `Ctrl+Alt+Up/Down` for TM match navigation.
- **Context-aware shortcuts**: Match navigation/insertion now respects active panel (Compare Panel vs Translation Results).
- **Minimalist sound effects**: Per-event Windows sound mapping (beeps or Windows `.wav`) with sound effects OFF by default.
- **Glossary add feedback**: Status-bar “information bar” messages for add/duplicate/error outcomes.
- **Less TM log spam**: Collapsed repeated “Saved segment to TM(s)” messages into a single debounced `(xN)` log line.
### January 11, 2026 - Docs: Contributing Guide & Code of Conduct
- Added `CONTRIBUTING.md` with guidance for bug reports, feature requests, PRs, dev setup, and coding standards
- Added `CODE_OF_CONDUCT.md` using Contributor Covenant v2.1 (with required CC BY attribution)
- Updated `README.md` to link the new docs and clarify MIT vs CC BY scope
### January 11, 2026 - Version 1.9.98: Glossary Notes in Tooltips
**📝 Fixed Glossary Entry Notes Not Appearing in Tooltips**
Fixed bug where glossary entry notes were not appearing in tooltips despite being saved correctly to the database.
**The Problem:**
- Notes were saved correctly to `termbase_terms.notes` database column
- When converting termbase matches from dict to list format for display, `notes` field was dropped in 5 places
- TermView and source cell tooltips never received notes data
**The Fix:**
Fixed 5 locations where glossary notes were being lost:
1. Cached termbase matches conversion (~lines 26753-26768)
2. Fresh termbase matches conversion (~lines 26835-26850)
3. Refresh current segment conversion (~lines 30420-30438)
4. TranslationMatch metadata (~lines 30469-30485)
**Also in this release:**
- **WebEngineView cleanup**: Fixed "Release of profile requested" terminal warnings
- **FAQ update**: Added embedded browser password/cookie security documentation
**Files Modified:**
- `Supervertaler.py` - Fixed notes field in 4 dict-to-list conversions + 1 TranslationMatch metadata
- `docs/superdocs/reference/faq.md` - Added FAQ about embedded browser security
---
### January 11, 2026 - Version 1.9.97: All MT Providers in Translation Results
**🌐 Multiple Machine Translation Providers Now Displayed**
The Translation Results panel now shows translations from **all configured MT providers**, not just Google Translate.
**Previously:**
- Only Google Translate was called when navigating to a segment
- DeepL, Amazon Translate, and MyMemory were only available in Batch Translate
**Now:**
- All enabled MT providers are called and displayed progressively
- Each provider's translation appears as it completes
- Provider codes: GT (Google), DL (DeepL), AT (Amazon), MM (MyMemory)
**Supported Providers:**
| Provider | Key Name | Notes |
|----------|----------|-------|
| Google Translate | `google_translate` | Requires API key |
| DeepL | `deepl` | Requires API key |
| Amazon Translate | `amazon_translate` + `amazon_translate_secret` | AWS credentials |
| MyMemory | `mymemory` | Free; email as key for higher limits |
**Files Modified:**
- `Supervertaler.py` - Expanded `_add_mt_and_llm_matches_progressive()` to call all MT providers
---
### January 11, 2026 - Version 1.9.96: Thread-Safe Logging
**🛡️ Fixed Crash When Adding Terms via Alt+Down**
Fixed critical crash caused by Qt threading violation when background workers called the `log()` method.
**The Problem:**
- `log()` method was called from background threads (termbase batch processor)
- Qt widgets (`status_bar`, `session_log_text`) were being accessed from non-main threads
- This caused: `QObject::killTimer: Timers cannot be stopped from another thread`
**The Fix:**
- Added `_log_signal = pyqtSignal(str)` to `SupervertalerQt` class
- Background threads now emit signal instead of directly updating UI
- Signal automatically queues to main thread's event loop
- New `_log_to_ui()` internal method handles actual widget updates
**Files Modified:**
- `Supervertaler.py` - Added `_log_signal`, `_log_to_ui()`, made `log()` thread-safe
---
### January 11, 2026 - Version 1.9.95: TM Fuzzy Matching Fix
**🔍 Improved Translation Memory Fuzzy Matching for Long Segments**
Fixed critical issue where highly similar TM entries were not being found for long segments (especially in patent/technical documents).
**The Problem:**
- FTS5 full-text search uses BM25 ranking which prioritizes entries matching MORE search terms
- For long segments with many technical compound words, BM25 pushed truly similar entries below the candidate limit
- Example: Two sentences 92% similar weren't matching because other entries matched more individual words
**The Fix:**
- Increased FTS5 candidate pool from 100 to 500 entries
- Changed `max(50, max_results * 10)` to `max(500, max_results * 50)` in `search_fuzzy_matches()`
**Technical Details:**
- FTS5 BM25 is great for keyword relevance but needs a larger pool for similarity-based reranking
- SequenceMatcher then correctly scores the candidates by actual text similarity
- More candidates = better chance of finding truly similar matches
**Files Modified:**
- `modules/database_manager.py` - `search_fuzzy_matches()` candidate limit increase
---
### January 11, 2026 - Version 1.9.94: TermView Quick-Insert Shortcuts
**🎯 TermView Quick-Insert Shortcuts**
**Note (updated in v1.9.99):** `Alt+0` / `Alt+0,0` are reserved for the Compare Panel. TermView shortcuts start at `Alt+1`.
Implemented a novel 20-term quick-insert system using Alt key shortcuts — a feature unique to Supervertaler:
- **Alt+1 through Alt+9**: Insert terms 1-9 instantly (no delay)
- **Double-tap Alt+N,N**: Insert terms 10-18 (displayed as 11, 22, ..., 99)
- **Smart double-tap detection**: First tap inserts immediately; if same key pressed within 300ms, undoes and inserts the double-digit term
- **Visual badges**: Blue circular badges (14px for single digit, 20px for double) show shortcut numbers
**Badge Mapping:**
| Badge | Shortcut | Internal Index |
|-------|----------|----------------|
| 1-9 | Alt+1 to Alt+9 | 0-8 |
| 11, 22, ..., 99 | Alt+1,1 to Alt+9,9 | 9-17 |
**Visual Improvements:**
- Background color now extends across entire term block (text + badge unified)
- Hover effect applies to whole container
- Rounded corners (3px border-radius) on term blocks
- Tooltips show exact shortcut instructions
**Implementation Details:**
*modules/termview_widget.py:*
- `TermBlock.__init__()`: Accepts `shortcut_number` parameter (0-19)
- `TermBlock.init_ui()`: Creates badge with dynamic width and proper styling
- `update_with_matches()`: Assigns shortcut numbers 0-19 to first 20 unique terms
- `insert_term_by_number()`: Inserts term by internal index, logs with badge text
- `shortcut_terms` dict: Maps internal index to target text
*Supervertaler.py:*
- `_termview_last_key` / `_termview_last_time`: Track double-tap state
- `_handle_termview_shortcut()`: Double-tap detection with undo + replace logic
- `_get_current_target_widget()`: Helper to get current target cell for undo
- `insert_termview_term_by_number()`: Delegates to TermView widget
*modules/shortcut_manager.py:*
- Added `termview_insert_0` through `termview_insert_9` definitions
- Category: "TermView Insertion"
- Descriptions explain double-tap behavior
**Files Modified:**
- `modules/termview_widget.py` - Badge display, shortcut tracking, insertion logic
- `Supervertaler.py` - Double-tap handler, shortcuts setup
- `modules/shortcut_manager.py` - Shortcut definitions for Alt+0-9
---
### January 11, 2026 - Version 1.9.93: Keyboard Shortcuts & Quick Glossary Add
**⚡ Quick Add to Priority Glossary**
- **Alt+Shift+Up**: Add selected term pair to glossary with Priority #1
- **Alt+Shift+Down**: Add selected term pair to glossary with Priority #2
- Uses database `termbase_activation` table for priority lookup
- No dialog required — instant term addition
**🔧 Shortcut Enable/Disable Feature**
- New "Enabled" checkbox column in Settings → Keyboard Shortcuts
- Disabled shortcuts fully release their key combinations
- Settings persist between sessions
**Bug Fixes:**
- Fixed Ctrl+Enter not working in target editor cells
- Fixed "Save & Next" button not confirming segments
- Renamed button to "✓ Confirm & Next"
---
### January 9, 2026 - Version 1.9.87: Auto-Confirm & Tab Layout
**⚡ Auto-Confirm 100% TM Matches Feature**
Implemented intelligent auto-confirmation system for perfect TM matches during Ctrl+Enter navigation:
- **Core Feature**: Automatically inserts, confirms, and skips segments with 100% TM matches
- **Recursive Logic**: Continues processing multiple consecutive 100% matches until finding segment needing manual work
- **Safety Check**: Only auto-confirms segments with empty targets (won't overwrite existing translations)
- **Hash-Based Lookup**: Uses `get_exact_match()` for instant O(1) MD5 hash-based matching
- **TM Integration**: Auto-confirmed segments automatically saved to activated Translation Memories
- **Pagination Handling**: Correctly switches pages when auto-skipping across page boundaries
**Implementation Journey** (5 iterations):
1. Initial implementation with fuzzy search
2. Fixed AttributeError: get_activated_tms() doesn't exist → use tm_settings
3. Fixed TypeError: limit parameter → changed to max_results
4. Switched from fuzzy to exact match (fuzzy returned 0 for 100% matches)
5. Added empty target check for safety
**Code Locations:**
- `Supervertaler.py` line 4747: Instance variable initialization
- `Supervertaler.py` lines 13468-13477: UI checkbox in General Settings
- `Supervertaler.py` lines 16243-16245: Save/load settings
- `Supervertaler.py` lines 31250-31438: Complete navigation logic with recursion
**📐 Tab Layout Customization Feature**
Added optional setting to move Termview and Session Log tabs above or below the grid:
- **New Setting**: "Show Termview/Session Log tabs above grid" checkbox in View Settings → Tab Layout
- **Persistence**: Setting saved to general_settings.json and restored on app start
- **Layout Logic**: Conditional widget ordering in `create_main_layout()` based on `self.tabs_above_grid` flag
- **Splitter Sizes**: Adjusted proportions for both layouts (200/600 for tabs above, 600/200 for tabs below)
**Code Locations:**
- `Supervertaler.py` line 4751: Instance variable `self.tabs_above_grid`
- `Supervertaler.py` lines 14287-14306: UI checkbox in View Settings
- `Supervertaler.py` lines 16243-16245, 24693-24694: Save/load
- `Supervertaler.py` lines 17016-17027: Conditional layout creation
**🎨 UI Polish & Bug Fixes**
- **Segment Column Width**: Increased from 35px to 55px (fits 4-digit segment numbers)
- **Auto-Center Persistence**: Fixed "Keep Active Segment Centered" not saving between sessions
- **Badge Text Color**: Changed from black to dark gray (#333333) for better appearance
- **Color Customization**: Added badge text color picker with 8 preset colors
- **Settings Rename**: "View/Display" → "View Settings" for clarity
**Files Modified:**
- `Supervertaler.py` - All implementations above
---
### January 9, 2026 - Version 1.9.89: Critical Bug Fixes
**🔧 Translation Results Zoom Persistence Fix**
Fixed critical bug where Translation Results pane font size settings were not being restored when loading projects:
- **Root Cause**: Method name typo in `load_project()` at line 18492: `set_compare_font_size()` instead of `set_compare_box_font_size()`
- **Impact**: Users complained "zoom settings are constantly being forgotten during a project and across projects"
- **Solution**: Changed method call to correct name `set_compare_box_font_size()`
- **Result**: Font sizes now properly persist across projects and sessions
**🎨 Border Thickness Spinbox Fix**
Fixed Target Cell Focus Border thickness control arrows not appearing:
- **Root Cause**: Complex stylesheet was hiding the spinbox arrow buttons completely
- **Previous State**: Users saw no arrows, couldn't click to adjust border thickness
- **Solution**:
- Removed problematic stylesheet that hid buttons
- Added explicit `setButtonSymbols(QSpinBox.ButtonSymbols.UpDownArrows)`
- Increased maximum thickness from 5px to 10px
- Made spinbox wider (90px) for better visibility
- **Result**: Arrows now visible and fully functional
**🌍 Language Pair Memory Fix**
Fixed DOCX import always defaulting to EN→NL instead of remembering user's language pair:
- **User Report**: "Whatever I do before importing a NL-ES table, the language is detected always as EN-NL!!!"
- **Root Cause**: Import dialog hardcoded to `source_combo.setCurrentIndex(0)` (English) and `target_combo.setCurrentIndex(1)` (Dutch)
- **Solution**:
- Added `last_import_source_lang` and `last_import_target_lang` to general_settings.json
- Load last used languages from settings on dialog open
- Fall back to current project languages if available
- Save selected languages to settings when user clicks Import
- **Result**: Language pair now persists across sessions and imports
**Code Locations:**
- `Supervertaler.py` line 18492: Fixed `set_compare_font_size` → `set_compare_box_font_size`
- `Supervertaler.py` lines 14363-14375: Fixed border thickness spinbox
- `Supervertaler.py` lines 19589-19625: Load last language pair from settings
- `Supervertaler.py` lines 19670-19677: Save language pair to settings
**Files Modified:**
- `Supervertaler.py` - All bug fixes implemented
---
### January 9, 2026 - Version 1.9.86: Glossary Duplicate Prevention & Priority Filtering
**🚫 Duplicate Term Prevention**
Implemented comprehensive duplicate prevention when adding terms to glossaries:
- **Case-Insensitive Check**: Added duplicate detection in `termbase_manager.py` before inserting terms
- **User Feedback**: Clear warning dialog when attempting to add duplicate: "This term already exists in [glossary]. Duplicate terms are not allowed."
- **Graceful Handling**: `add_term()` returns `None` if duplicate found (no exception raised)
- **Database Query**: Uses `LOWER()` SQL function for case-insensitive comparison
**🎯 Priority-Based Duplicate Filtering**
Enhanced glossary match display to show only highest priority match when duplicates exist:
- **Filtering at Source**: `find_termbase_matches_in_source()` filters duplicates before caching
- **Priority Logic**: When same source→target exists in multiple glossaries, keeps only lowest ranking number (highest priority)
- **Comprehensive Coverage**: Filtering applies to grid highlighting, Translation Results panel, TermView, and Superlookup
- **Performance**: Duplicate filtering adds negligible overhead (single pass through matches)
**⚖️ TermView Font Normalization**
Fixed font size inconsistency in TermView widget:
- **Before**: Source text 10pt, target text 8pt (2pt smaller)
- **After**: Both source and target use same font size
- **Benefit**: Improved readability and visual consistency
**Implementation Details:**
- `modules/termbase_manager.py` (lines 620-680): Added duplicate check query before INSERT
- `Supervertaler.py` (lines 27680-27707): Added duplicate filtering logic with seen_pairs dict
- `Supervertaler.py` (lines 38420-38430): Show warning dialog when duplicate detected
- `modules/termview_widget.py` (line 219): Changed target font size from `self.font_size - 2` to `self.font_size`
**Files Modified:**
- `modules/termbase_manager.py` - Duplicate check in `add_term()`
- `Supervertaler.py` - Duplicate filtering in `find_termbase_matches_in_source()`, warning dialog in `show_add_term_dialog()`
- `modules/termview_widget.py` - Font size normalization in `TermBlock.init_ui()`
---
### January 9, 2026 - Bug Fix: QTextCursor Scope Error
**🐛 Fixed Navigation Crash**
Resolved critical UnboundLocalError that occurred during segment navigation (Alt+Up/Alt+Down):
- **Root Cause**: `QTextCursor` imported inside if block but used in else block (classic Python scope error)
- **Solution**: Moved import to top of both `go_to_next_segment()` and `go_to_previous_segment()` methods
- **Impact**: Prevented crashes during normal workflow navigation
**Files Modified:**
- `Supervertaler.py` - Fixed QTextCursor import scope in navigation methods
---
### January 9, 2026 - Beijerterm: Splitpen Image Resize
**🖼️ Image Size Optimization**
Reduced oversized first image on splitpen term page:
- **Before**: Full-width display (~1200px)
- **After**: 300px width using HTML `<img>` tag
- **Conversion**: Changed from Markdown `` to `<img src="url" width="300">`
**Files Modified:**
- `beijerterm/content/terms/splitpen.md` - First image now uses HTML with width attribute
---
### January 7, 2026 - Version 1.9.85: AI Proofreading System
**✅ Intelligent Translation Quality Verification**
Complete proofreading system for AI-powered translation quality checking:
- **Batch Proofreading Feature**: LLM analyzes translations for errors, inconsistencies, and quality issues
- Processes segments in batches of 20 for efficient API usage
- Issues stored in Notes field with `⚠️ PROOFREAD:` prefix for easy identification
- Real-time progress dialog shows statistics during operation
- **Results Management**:
- Proofreading Results dialog shows all segments with issues in a table
- Double-click any result to navigate directly to that segment
- Orange highlight on status icons for segments with proofreading notes
- **Filter Integration**: New "Has proofreading issues" option in Advanced Filters
- Quickly isolate segments that need attention
- Works alongside existing status/text filters
- **Clear Operations**:
- Bulk clear: Remove all proofreading notes from entire project
- Individual clear: Right-click segment to clear proofreading notes
**Access Points:**
- Edit → Batch Operations → ✅ Proofread Translation...
- View → ✅ Proofreading Results...
- Right-click → ✅ Clear Proofreading Notes
**Implementation Details:**
- `Supervertaler.py` (lines ~25732-26131): Proofreading methods
- `show_proofread_dialog()`: Configuration and execution
- `_run_proofreading()`: Batch processing with LLM
- `show_proofreading_results_dialog()`: Results table UI
- `_clear_all_proofreading_notes()`: Bulk clear operation
- Orange status icon highlighting when `segment.notes` contains proofreading data
- Advanced Filters updated with proofreading checkbox
**Files Modified:**
- `Supervertaler.py` - Proofreading system implementation
---
### January 7, 2026 - Version 1.9.84: Subscript & Superscript Support
**📐 New Formatting Tags**
Added support for subscript and superscript formatting throughout the entire pipeline:
- **Import**: DOCX files with subscript/superscript text now preserve formatting as `<sub>` and `<sup>` tags
- **Display**: Tags shown in grid cells (e.g., `P<sub>totaal</sub>`, `m<sup>2</sup>`)
- **Export**: Tags converted back to real Word subscript/superscript formatting
- **Preview**: Document Preview renders actual subscript/superscript positioning
**Technical Implementation:**
- `modules/tag_manager.py`:
- `FormattingRun` dataclass: Added `subscript: bool` and `superscript: bool` fields
- `TAG_PATTERN`: Extended to `r'<(/?)([biu]|bi|li|sub|sup)>'`
- `tag_colors`: Added `'sub': '#666600'` and `'sup': '#006666'`
- `extract_runs()`: Now captures `run.font.subscript` and `run.font.superscript`
- `runs_to_tagged_text()`: Generates `<sub>` and `<sup>` tags
- `tagged_text_to_runs()`: Parses `<sub>` and `<sup>` tags
- `modules/docx_handler.py`:
- Formatting check includes `<sub>` and `<sup>` tag detection
- Fallback mode strips `<sub>` and `<sup>` tags
- `_replace_paragraph_with_formatting()`: Applies `run.font.subscript = True` and `run.font.superscript = True`
- `Supervertaler.py`:
- `_render_formatted_text()`: Handles `<sub>` and `<sup>` tags using `QTextCharFormat.VerticalAlignment.AlignSubScript/AlignSuperScript`
**Files Modified:**
- `modules/tag_manager.py` - FormattingRun, TAG_PATTERN, tag_colors, extract_runs, runs_to_tagged_text, tagged_text_to_runs
- `modules/docx_handler.py` - Format detection, tag stripping, formatting application
- `Supervertaler.py` - Preview rendering with actual sub/superscript
---
### January 6, 2026 - Version 1.9.83: Notes Tab & Status Indicator
**📝 Notes Tab in Translation Results Panel**
Translation Results panel now has tabbed interface for TM Info and Notes:
- **TM Info Tab**: Shows TM match details when a match is selected
- **Notes Tab**: Add/edit notes for each segment with auto-save
- **Persistence**: Notes saved to .svproj project file
- **Removed Comments Tab**: Redundant tab under grid removed (Notes tab replaces it)
**🟠 Notes Indicator on Status Icon**
Clean visual indicator for segments with notes:
- **Orange Background**: Status icon (✓/✗) gets subtle orange highlight when segment has notes
- **Compact Design**: No separate icon cluttering the status cell
- **Tooltip**: Hover over status cell to see notes preview
- **Narrower Column**: Status column reduced from 120px to 70px
**Files Modified:**
- `Supervertaler.py` - Notes saving/loading, status icon highlighting, removed Comments tab
- `modules/translation_results_panel.py` - Added TM Info + Notes tabs interface
---
### January 5, 2026 - Version 1.9.82: Export for AI
**🤖 AI-Readable Export Format**
New export option for segments in a format optimized for AI systems:
- **Menu Location**: File → Export → 🤖 AI-Readable Format (TXT)...
- **Output Format**: `[SEGMENT 0001]` with language-labeled source/target
- **Auto Language Codes**: Detects project languages and converts to short codes (NL, EN, DE, etc.)
**Configurable Options:**
- **Language Codes**: Editable source/target codes (auto-filled from project)
- **Numbering**: Start number and zero padding (1-8 digits)
- **Content Modes**: Bilingual (source+target), Source only, Target only
- **Segment Filters**: All, Untranslated only, Translated only
- **Live Preview**: Shows format preview before export
**Use Cases:**
- Export source-only for AI translation (ChatGPT, Claude, Gemini)
- Export bilingual for AI quality review
- Simple format for automated processing and re-import
**Files Modified:**
- `Supervertaler.py` - Added `export_for_ai()` method, menu item in Export submenu
---
### January 4, 2026 - Version 1.9.81: Superlookup UX Improvements
**🔍 Search History Dropdown**
Superlookup search box now remembers last 20 searches:
- **HistoryComboBox**: Replaced QTextEdit with editable combo box showing search history
- **Persistent**: Saved to `user_data/superlookup_history.json`
- **Most Recent First**: New searches added to top of dropdown
**↔️ Resizable Sidebar**
Web Resources sidebar now resizable:
- **QSplitter**: Allows resizing between 120-250px
- **No Text Cutoff**: Resource buttons properly visible at all widths
**🎨 UI Polish**
- **Focus Rectangles Removed**: Global stylesheet removes ugly focus outlines from all buttons
- **Styled Radio Buttons**: 5 plain QRadioButton instances replaced with CheckmarkRadioButton
- **External Mode Fix**: External browser mode now correctly triggers web search
**Files Modified:**
- `Supervertaler.py` - HistoryComboBox, QSplitter, global stylesheet, radio buttons, external mode
---
### January 4, 2026 - Version 1.9.80: GitHub Code Search (Beijerterm)
**💻 GitHub Code Search for Beijerterm**
Added dedicated GitHub Code search for Beijerterm repository:
- **New Resource**: "💻 GitHub Code (Beijerterm)" button in Superlookup's Web Resources
- **Search URL**: `https://github.com/search?q={query}+repo:michaelbeijer/beijerterm&type=code`
- **Search Source Files**: Query YAML glossary files and Markdown documentation
- **Renamed**: "GitHub Code" → "GitHub Code (all)" for clarity
**Files Modified:**
- `Supervertaler.py` - Added `github_beijerterm` resource, renamed `github_code` to "GitHub Code (all)"
---
### January 4, 2026 - Version 1.9.79: Beijerterm Integration in Superlookup
**📚 Beijerterm Web Resource**
Added Beijerterm to Superlookup's Web Resources tab:
- **Updated Resource**: Replaced old `michaelbeijer.co.uk` wiki with new Beijerterm static site
- **Search URL**: `https://michaelbeijer.github.io/beijerterm/?q={query}`
- **500k+ Terms**: Dutch-English terminology database with 583,000+ term entries
- **URL Search**: Beijerterm now supports `?q=searchterm` for programmatic search integration
**Files Modified:**
- `Supervertaler.py` - Updated `beijerterm` resource in `self.web_resources` list
---
### January 9, 2026 - Version 1.9.88: Context Menu Enhancement
**🔍 Superlookup Integration in Context Menus**
Implemented quick concordance search via right-click context menu:
- **New Context Menu Item**: "🔍 Search in Superlookup (Ctrl+K)" appears when text is selected
- **Both Editor Types**: Added to ReadOnlyGridTextEditor (source) and EditableGridTextEditor (target)
- **Smart Workflow**: Select text → Right-click → Instant search in Superlookup
- **Auto-Navigation**: Opens Superlookup tab and triggers search automatically
- **Language-Aware**: Passes project source/target language pair to search
- **Vertical View**: Uses traditional concordance list layout
- **Unified Search**: Searches TM, glossaries, Supermemory, MT, and web resources
**Implementation Details:**
- `_handle_superlookup_search()` method added to both editor classes
- Gets selected text from `textCursor().selectedText()`
- Calls `_go_to_superlookup()` for navigation
- Calls `lookup_tab.search_with_query()` with full parameters
- Menu item positioned after Paste, before Add to Glossary
**Code Locations:**
- `Supervertaler.py` lines 1885-1895: ReadOnlyGridTextEditor context menu
- `Supervertaler.py` lines 1670-1702: ReadOnlyGridTextEditor handler method
- `Supervertaler.py` lines 2412-2422: EditableGridTextEditor context menu
- `Supervertaler.py` lines 2360-2392: EditableGridTextEditor handler method
**Files Modified:**
- `Supervertaler.py` - Added menu items and handler methods to both editor classes
---
### January 4, 2026 - Version 1.9.78: Find & Replace History & Batch Sets
**🔍 Find & Replace History**
Enhanced Find & Replace dialog (Ctrl+F / Ctrl+H) with history dropdowns and batch operations:
- **History Dropdowns**: Find and Replace fields now have dropdown arrows showing last 20 searches
- **Persistent History**: Search/replace terms saved to `user_data/find_replace_history.json`
- **F&R Sets Panel**: Collapsible panel for creating and managing batch replace operations
- **Batch Operations**: Run multiple find/replace operations with a single click
- **Import/Export**: Save F&R sets as `.svfr` files for sharing or backup
**New Module: `modules/find_replace_qt.py`**
- `FindReplaceHistory` - Manages and persists recent search/replace terms
- `FindReplaceOperation` - Dataclass for single F&R operation
- `FindReplaceSet` - Collection of operations that can be saved/loaded
- `FindReplaceSetsManager` - QWidget UI for managing F&R sets
- `HistoryComboBox` - Editable combo box with history dropdown
**UI Changes:**
- Find/Replace inputs changed from QLineEdit to HistoryComboBox
- Added "F&R Sets (Batch Operations)" collapsible panel below main dialog
- Added "+ Add to Set" button to save current operation to selected set
- Added "▶ Run All" button to execute all enabled operations in a set
- Double-click operation in table loads it into the dialog fields
**Files Created:**
- `modules/find_replace_qt.py` - New F&R support module (~400 lines)
**Files Modified:**
- `Supervertaler.py` - Added import, `fr_history` instance, rewrote `show_find_replace_dialog()`, added helper methods
---
### January 4, 2026 - Version 1.9.77: GitHub Code Search in Superlookup
**💻 GitHub Code Search**
Added GitHub Code Search to Superlookup's Web Resources tab:
- **New Resource**: "💻 GitHub Code" button in the web resources sidebar
- **Search URL**: `https://github.com/search?q={query}&type=code`
- **Use Case**: Search for terms/code across all public GitHub repositories
- **Great for**: Finding how technical terms are used in real code, locating terminology in open-source projects
**Files Modified:**
- `Supervertaler.py` - Added `github_code` resource to `self.web_resources` list
---
### January 3, 2025 - Version 1.9.76: Onboarding, Spellcheck & Project Info
**🎉 First-Run Welcome for New Users**
New users now get a welcoming introduction to the modular architecture:
- **Welcome Dialog**: Shows on first launch explaining modular pip extras
- **Auto-Navigate**: Opens Settings → Features tab automatically
- **Don't Show Again**: Checkbox to skip on future launches (uses CheckmarkCheckBox)
- **Bug Fix**: First-run flag now saves to correct file (ui_preferences.json)
**💰 Free vs Paid LLM Info Box**
Added clear pricing information to Settings → AI Settings:
- **Google Gemini**: FREE tier (15 req/min, 1M tokens/day)
- **Ollama**: 100% FREE (runs locally)
- **OpenAI/Claude**: Paid API only
- **Important Note**: Clarifies that ChatGPT Plus and Claude Pro web subscriptions do NOT include API access
**🔤 Spylls Spellcheck (Windows Fix)**
Fixed Hunspell spellcheck on Windows by switching to `spylls`:
- **Problem**: `cyhunspell` fails to compile on Python 3.12+ / Windows
- **Solution**: Replaced with `spylls` - pure Python Hunspell implementation
- **Benefit**: Supports regional variants (en-US vs en-GB distinguish "colour/color")
- **Backend Priority**: hunspell → spylls → pyspellchecker (graceful degradation)
**🌍 Language Variant Support**
Spellcheck now supports regional language variants:
- **Dropdown shows variants**: "English (US)", "English (GB)", "Portuguese (BR)", etc.
- **Subdirectory search**: Finds dictionaries in `dictionaries/en/en_GB.dic` subfolders
- **Added en_ZA**: South African English variant mapping
- **Regional spelling works**: "colour" correct in en_GB, incorrect in en_US
**📋 Improved Spellcheck Info Dialog**
Enhanced Spellcheck Info dialog with better documentation:
- **Three backend display**: Shows Hunspell, Spylls, pyspellchecker status separately
- **Active backend highlighting**: Current backend highlighted in green
- **Bundled dict info**: Explains Spylls includes EN, RU, SV dictionaries
- **Project links section**: Links to pyspellchecker, spylls, Hunspell GitHub pages
- **Add More Dictionaries**: Renamed section with clearer instructions
**📋 Project Info Dialog (NEW)**
New dialog to view comprehensive project information:
- **Menu**: File → 📋 Project Info...
- **Overview section**: Name, file path, languages, created/modified dates, project ID
- **Statistics section**: Segment counts, word counts, character counts, progress %
- **Source Files section**: Shows original DOCX, memoQ, CafeTran, Trados paths
- **Resources section**: Active prompt, TMs, glossaries, spellcheck settings
**Files Modified:**
- `Supervertaler.py` - Added `_show_first_run_welcome()`, `show_project_info_dialog()`, free/paid info box, spellcheck dialog improvements
- `modules/feature_manager.py` - Changed hunspell check_import from `cyhunspell` to `spylls`
- `modules/spellcheck_manager.py` - Added spylls backend, language variants, subdirectory search, CODE_TO_DISPLAY mapping
- `pyproject.toml` - Replaced `cyhunspell>=2.0.0` with `spylls>=0.1.7`
- `requirements.txt` - Updated hunspell dependency to spylls
---
### January 2, 2025 - Version 1.9.75: Modular Architecture
**📦 Modular Installation System**
Major new feature allowing users to install only the features they need, significantly reducing disk space requirements:
- **Feature Manager Module**: New `modules/feature_manager.py` provides centralized feature detection and management
- **pip Extras**: Install specific features with `pip install supervertaler[supermemory,voice,web]`
- **Settings → Features Tab**: New UI showing installed/missing features with size estimates and install commands
- **Lazy Loading Helpers**: Functions like `lazy_import_supermemory()` for conditional imports
**Installation Options:**
- `pip install supervertaler` - Core only (~300 MB)
- `pip install supervertaler[supermemory]` - Add semantic search (+600 MB)
- `pip install supervertaler[voice]` - Add voice commands (+150 MB)
- `pip install supervertaler[all]` - Everything (~1.2 GB)
**Feature Modules:**
| Module | pip Extra | Size |
|--------|-----------|------|
| Supermemory | `supermemory` | ~600 MB |
| Supervoice | `voice` | ~150 MB |
| Web Browser | `web` | ~100 MB |
| PDF Rescue | `pdf` | ~30 MB |
| MT Providers | `mt` | ~30 MB |
| Hunspell | `hunspell` | ~20 MB |
| AutoFingers | `windows` | ~10 MB |
**Files Created:**
- `modules/feature_manager.py` - FeatureManager class, FEATURE_MODULES dict, lazy import helpers
**Files Modified:**
- `Supervertaler.py` - Added `_create_features_settings_tab()`, version bump to 1.9.75
- `pyproject.toml` - Reorganized dependencies into core vs optional extras
- `requirements.txt` - Updated with modular structure documentation
- `AGENTS.md` - Added modular installation documentation
---
### January 2, 2026 - Docs: Superdocs link sanity sweep
- Fixed a broken internal Superdocs link in the TMX Editor page (Related → Translation memory).
- Re-ran SUMMARY.md and internal markdown link checks to confirm clean navigation.
**Files Modified:**
- `docs/superdocs/tools/tmx-editor.md`
### January 1, 2026 - Docs: Superdocs FAQ integration & Sv website icon
- Integrated the full legacy repository FAQ into Superdocs (GitBook source) so the FAQ lives inside the online help center.
- Updated the in-app Help → FAQ link to open the Superdocs FAQ page.
- Updated the website navbar to use the Sv icon in the top-left brand slot.
**Files Modified:**
- `docs/superdocs/reference/faq.md`
- `Supervertaler.py`
- `docs/index.html`
- `docs/sv-icon.svg`
- `docs/superdocs/.gitbook/assets/sv-icon.svg`
### January 1, 2026 - Docs: Tools pages expanded
- Rewrote the Superdocs Tools pages for accuracy against the current PyQt6 UI.
- Expanded documentation for TMX Editor, AutoFingers, Supervoice (voice commands), and Image Extractor.
**Files Modified:**
- `docs/superdocs/tools/tmx-editor.md`
- `docs/superdocs/tools/autofingers.md`
- `docs/superdocs/tools/voice-commands.md`
- `docs/superdocs/tools/image-extractor.md`
### January 1, 2026 - Docs: Superlookup pages expanded
- Expanded the Superdocs Superlookup pages (TM Search, Glossary Search, Machine Translation, Web Resources) to match current UI and behavior.
- Updated the Superlookup overview page to align entry points (Ctrl+K / Ctrl+Alt+L), language/direction filters, and the actual Web Resources list.
- Fixed Superlookup copy/insert behavior for TM and Glossary results (table items were missing under rich cell widgets).
**Files Modified:**
- `docs/superdocs/superlookup/overview.md`
- `docs/superdocs/superlookup/tm-search.md`
- `docs/superdocs/superlookup/glossary-search.md`
- `docs/superdocs/superlookup/mt.md`
- `docs/superdocs/superlookup/web-resources.md`
- `Supervertaler.py`
### December 31, 2025 - Version 1.9.73: External Prompt Editor Display
**📝 External Prompts Now Display in Editor**
When loading an external prompt file (not in the library), it now displays in the Prompt Editor panel:
- **Editor display**: External prompts show in the editor with name, description, and content fields
- **Read/write support**: External prompts can be edited and saved back to the original file
- **Format detection**: `.svprompt` files parsed as JSON to extract name, description, and content
- **Plain text support**: `.txt` and `.md` files displayed as raw content
- **Visual indicator**: Editor label shows "📁 External: {name}" to distinguish from library prompts
- **Save functionality**: Changes can be saved back to external files (JSON for .svprompt, plain text for others)
- **Project load fix**: Prompts stored in project files now display in editor when project is loaded (both external and library prompts)
**Files Modified:**
- `modules/unified_prompt_manager_qt.py` - Added `_display_external_prompt_in_editor()`, `_save_external_prompt()`, updated `_load_external_primary_prompt()`, `_save_current_prompt()`, and `_set_primary_prompt()`
- `Supervertaler.py` - Added call to `_display_external_prompt_in_editor()` in project load code
---
### December 30, 2025 - Version 1.9.67: memoQ Tag Highlighting & Batch Retry Fix
**🏷️ memoQ Content Tag Highlighting**
Fixed and improved tag highlighting for memoQ bilingual DOCX files:
- **Mixed bracket tags now highlighted**: `[uicontrol id="..."}`, `{uicontrol]`, `[image cid="..." href="..."]`
- **Closing tags**: `{tagname}` like `{MQ}`, `{tspan}` now properly pink
- **No false positives**: `[Company]`, `[Bedrijf]` placeholders stay black (only tags with attributes or mixed brackets highlighted)
- **Regex patterns fixed**: Patterns now exclude unwanted characters to prevent greedy matching across multiple tags
**Final Tag Patterns:**
```python
r'\[[^}\]]+\}' # memoQ mixed: [anything}
r'\{[^\[\]]+\]' # memoQ mixed: {anything]
r'\[[a-zA-Z][^}\]]*\s[^}\]]*\]' # memoQ content: [tag attr...]
r'\{[a-zA-Z][a-zA-Z0-9_-]*\}' # memoQ closing: {tagname}
```
**🔄 Batch Translate Retry Fix**
Fixed `UnboundLocalError: cannot access local variable 'provider_dialog'` when retry pass starts:
- **Root Cause**: Retry passes skipped the dialog but code still tried to use dialog variables
- **Fix**: Properly wrapped entire dialog section in `if not is_retry_pass:` block (~200 lines)
- **Retry passes**: Now correctly use stored settings and skip directly to progress dialog
**📄 Pagination After Clear Filters**
Fixed pagination resetting when clearing filters:
- **Previously**: Clear Filters showed ALL segments, ignoring per-page setting
- **Now**: After clearing filters, `_apply_pagination_to_grid()` is called to maintain pagination
**🎨 Tag Color Setting Visibility**
Improved discoverability of tag color setting:
- **Group renamed**: "Translation Results Pane Font Size" → "Translation Results Pane & Tag Colors"
- **Tooltip updated**: Now explains the color applies to "CAT tool tags in the grid and results pane"
**Files Modified:**
- `Supervertaler.py` - TagHighlighter patterns, batch translate retry logic, clear_filters pagination, view settings labels
---
### December 30, 2025 - Version 1.9.68: memoQ Tag Color as Default
**🎨 Default Tag Color Changed to memoQ Dark Red**
Changed the default tag highlight color from light pink (`#FFB6C1`) to memoQ's actual dark red (`#7f0001`):
- **Color picked from memoQ**: The exact color memoQ uses for inline tags (`#7f0001` / RGB 127,0,1)
- **Updated everywhere**: Grid cells, Translation Results panel, Settings defaults
- **Preset colors**: Added 8 preset colors to the color picker (memoQ red, memoQ orange, Trados blue/purple, etc.)
- **Reset button**: Now resets to memoQ red with updated tooltip
**Tag Color Behavior:**
- **memoQ exports**: Preserve original memoQ colors from source file ✅
- **Trados exports**: Preserve original Trados colors from source file ✅
- **Phrase exports**: Preserve original Phrase colors from source file ✅
- **CafeTran exports**: Preserve original CafeTran colors from source file ✅
- **Supervertaler Bilingual Table**: Uses memoQ red (`RGBColor(127, 0, 1)`) ✅
**Files Modified:**
- `Supervertaler.py` - Default tag colors, preset colors, reset button, bilingual table export
- `modules/translation_results_panel.py` - Default tag color
---
### December 30, 2025 - Version 1.9.69: Page Up/Down Pagination Navigation
**📄 Page Up/Down Shortcuts for Pagination**
Added Page Up and Page Down keyboard shortcuts for navigating through pagination pages:
- **Page Up**: Go to previous page in pagination
- **Page Down**: Go to next page in pagination
- **Added to Settings**: Shortcuts now appear in Settings → Keyboard Shortcuts under "Grid Navigation" category
**Files Modified:**
- `Supervertaler.py` - Added `shortcut_page_up` and `shortcut_page_down` QShortcuts
- `modules/shortcut_manager.py` - Added `page_prev` and `page_next` shortcut definitions
---
### December 30, 2025 - Version 1.9.71: Go to Segment Dialog (Ctrl+G)
**⌨️ Improved Go to Segment Dialog**
Enhanced the Go to Segment feature with a streamlined dialog:
- **Global shortcut**: Ctrl+G now works from anywhere in the application (added as QShortcut)
- **Minimal dialog**: Small, focused dialog with just a text field
- **Type and Enter**: Just type the segment number and press Enter - no need to click buttons
- **Input validation**: Only accepts valid segment numbers within range
- **Shows current position**: Placeholder shows current segment number
- **Pagination-aware**: Automatically switches to the correct page when jumping to segments on other pages
- **Cursor placement**: Target cell is focused and cursor placed at end, ready to edit
- **Shortcut conflict fix**: Removed duplicate Ctrl+G assignment (was on both QShortcut and menu action)
**Already in Settings**: The shortcut was already listed in Settings → Keyboard Shortcuts under "Edit" category
**Files Modified:**
- `Supervertaler.py` - Added `shortcut_goto` QShortcut, redesigned `show_goto_dialog()` with pagination support, cursor focus, removed duplicate shortcut
---
### December 30, 2025 - Version 1.9.67: memoQ Tag Highlighting & Batch Retry Fix
**📚 Documentation Consolidation**
Merged two separate guides into a single comprehensive manual:
- **Previous**: `QUICK_START.md` + `CAT_WORKFLOW.md` (separate files)
- **Now**: `MANUAL.md` - Supervertaler Manual (consolidated into online Superdocs)
- **Contents**: Installation, API setup, CAT tool workflows (memoQ, Trados, CafeTran, Phrase), keyboard shortcuts, formatting, troubleshooting
- **Location**: `docs/guides/MANUAL.md` (archived); primary docs now hosted at https://supervertaler.gitbook.io/superdocs/
**⬆️⬇️ memoQ-Style Arrow Key Navigation**
Implemented intuitive segment navigation using Up/Down arrow keys:
- **Up Arrow at top line**: Moves to previous segment, positions cursor at last line
- **Down Arrow at bottom line**: Moves to next segment, positions cursor at first line
- **Cursor column preserved**: When moving between segments, cursor stays at same column position (or end of line if shorter)
- **Works in both Source and Target cells**: Navigation works whether you're in the read-only source cell or editable target cell
- **Normal behavior within cell**: Arrow keys work normally when cursor is NOT at first/last line
**New Methods in `EditableGridTextEditor`:**
- `_position_cursor_at_end_of_segment()` - Position cursor at last line of cell
- `_position_cursor_at_start_of_segment()` - Position cursor at first line of cell
**Also Added to `ReadOnlyGridTextEditor`:**
- Same Up/Down arrow key navigation
- Tab key now cycles to target cell (column 3)
**🔢 Grammar Fix**
Fixed "Batch Translate 1 Segments" → "Batch Translate 1 Segment" (proper singular/plural)
**Files Modified:**
- `Supervertaler.py` - Arrow key navigation in both editor classes, grammar fix
- `docs/guides/MANUAL.md` - New consolidated manual (now archived; primary docs live in Superdocs)
- Deleted: `docs/guides/QUICK_START.md`, `docs/guides/CAT_WORKFLOW.md`
---
### December 29, 2025 - Version 1.9.64: Grid Pagination & Batch Translate Retry
**📄 Working Grid Pagination**
The pagination controls now actually filter the displayed segments:
- **Previously**: Pagination UI existed but all segments were always shown
- **Now**: When you select "50 per page", only 50 segments are displayed at a time
- **Navigation**: First/Prev/Next/Last buttons or type a page number
- **Efficient**: Uses show/hide rows approach without reloading the grid
**New Pagination Methods:**
- `_get_total_pages()` - Calculate total pages based on segments and page size
- `_update_pagination_ui()` - Update labels and enable/disable buttons
- `_apply_pagination_to_grid()` - Show/hide rows for current page
- Updated `go_to_first_page()`, `go_to_prev_page()`, `go_to_next_page()`, `go_to_last_page()`, `go_to_page()`, `on_page_size_changed()`
**🔄 Batch Translate Retry Until Complete**
New option to automatically retry translating empty segments:
- **Checkbox**: "🔄 Retry until all segments are translated (recommended)" in batch translate dialog
- **Default**: Enabled by default
- **Behavior**: After first pass completes, checks for segments that are still empty. If any found, automatically starts another pass with just those segments.
- **Max Retries**: 5 passes maximum to prevent infinite loops
- **LLM Only**: Only works with LLM provider (not TM or MT modes)
**Instance Variables for Retry:**
- `_batch_retry_enabled` - Whether retry is enabled
- `_batch_retry_pass` - Current retry pass number (0 = first pass)
- `_batch_provider_type`, `_batch_provider_name`, `_batch_model` - Stored settings for retry passes
**🤖 Prompt Manager Tab Rename**
- "Prompts" tab renamed to "Prompt manager" in Project resources
**📁 External Prompt Restoration Fix**
Fixed external prompts not being restored when loading a project:
- External prompts are saved with `[EXTERNAL] /path/to/file.svprompt` prefix
- On load, detects `[EXTERNAL]` prefix and calls `library.set_external_primary_prompt()`
- Updates UI label with 📁 icon and prompt name
- Shows warning if external file no longer exists
**Files Modified:**
- `Supervertaler.py` - Pagination methods, batch translate retry, prompt restoration, tab rename
- `modules/unified_prompt_library.py` - (no changes, existing method used)
---
### December 30, 2025 - Version 1.9.66: Performance Boost & Cache Fix
**⚡ Termbase Cache Fix**
Fixed critical bug where termbase cache wasn't working - same segments were being re-searched repeatedly:
- **Root Cause**: Cache check was looking at whether `stored_matches` was empty, not whether the segment was in the cache. Empty results (`{}`) were never cached, causing repeated slow searches for segments with no termbase matches.
- **Fix**: Changed to proper membership check using `segment_id in self.termbase_cache`:
- Uses `cache_checked` boolean flag to track if cache was consulted
- Stores results in cache EVEN IF EMPTY so segments without matches are remembered
- Removed duplicate cache-checking code blocks
**🔇 Reduced Logging Overhead**
Removed verbose logging that was contributing to navigation slowness:
- Removed per-word termbase search logging (was logging each of 50+ words per segment)
- Removed per-match termbase result logging
- Removed prefetch worker progress logging (every 10 segments)
- Removed MT/LLM scheduling/execution debug logging
- Removed TM search debug logging (project ID, activated TMs, match counts)
- Removed termbase highlighting debug logging
- Kept only error logs and essential status messages
**Result**: Navigation between segments should feel significantly faster, especially for segments with no termbase matches (which were being re-searched every time).
**Files Modified:**
- `Supervertaler.py` - Cache logic fix (~lines 24260-24290), removed verbose logging throughout
---
### December 29, 2025 - Version 1.9.63: Linux Memory Access Violation Fix
**🐧 Linux Stability Improvements**
Fixed memory access violations (segfaults) that could occur on Linux when clicking in the grid after importing a Trados package:
**Root Cause**: Native code libraries (Hunspell spellcheck, ChromaDB, Sentence-Transformers) can crash with segfaults on Linux, especially when:
- Hunspell dictionaries are not properly installed for the target language
- ChromaDB's Rust backend has thread safety issues
- Tokenizers library uses parallel processing
**Fixes Applied:**
1. **Safer Hunspell Initialization** (`modules/spellcheck_manager.py`):
- Added test spell check call during initialization to catch crashes early
- Added `_crash_detected` flag to disable spellcheck permanently if it crashes
- If spellcheck crashes mid-session, it's automatically disabled
2. **Protected Spellcheck Highlighting** (`Supervertaler.py`):
- Added safety checks in `TagHighlighter._highlight_misspelled_words()`
- Wrapped spellcheck loop in try/except to catch any errors
- If errors occur, spellcheck is disabled for the session
3. **Skip AutoHotkey on Linux/Mac** (`Supervertaler.py`):
- AutoHotkey registration now skipped entirely on non-Windows platforms
- No more "AutoHotkey not found" warnings on Linux/Mac
- Settings and menu items already hidden via `os.name == 'nt'` checks
**User Guidance for Linux:**
- If crashes persist, disable spellcheck in Settings → View Settings
- Ensure Hunspell dictionaries are installed: `sudo apt install hunspell-pl` (for Polish)
- Disable Supermemory auto-init if using ChromaDB causes issues
**Files Modified:**
- `modules/spellcheck_manager.py` - `_try_hunspell()`, `check_word()`, added `_crash_detected` flag
- `Supervertaler.py` - `_highlight_misspelled_words()`, `main()` Linux env vars
---
### December 22, 2025 - Version 1.9.61: DOCX Export Tag Fix & TagHighlighter Accuracy
**📄 DOCX Export List Tag Stripping**
Fixed `<li-o>` and `<li-b>` tags appearing in exported Word documents:
- Added `re.sub(r'</?li-[ob]>', '', text)` to both `_replace_paragraph_text()` and `_replace_paragraph_with_formatting()` in docx_handler.py
- These list item tags are now stripped during DOCX export, not passed through to the document
**Files Modified:**
- `modules/docx_handler.py` - `_replace_paragraph_text()`, `_replace_paragraph_with_formatting()`
**🎨 TagHighlighter - Fixed False Positives for Bracket Text**
Fixed `[Bedrijf]`, `[Company]` and other square-bracketed placeholder text being incorrectly highlighted in pink (tag color):
- **Root Cause**: The TagHighlighter patterns were too broad:
- `r'\[[a-zA-Z][^}\]]*\]'` matched ANY `[text]` starting with a letter
- `r'\{[a-zA-Z][^}\]]*\}'` matched ANY `{text}` starting with a letter
- **Fix**: Removed these overly broad patterns. Now only highlights:
- HTML/XML tags: `<tag>`, `</tag>`, `<tag/>`
- Trados numeric: `<1>`, `</1>`
- memoQ numeric: `[1}`, `{1]`, `[1]`, `{1}` (numbers only, not arbitrary text)
- **Result**: `[Bedrijf]`, `[Company]`, `{placeholder}` etc. are no longer colored pink - only actual CAT tool tags get highlighted
**Files Modified:**
- `Supervertaler.py` - `TagHighlighter.highlightBlock()` - removed overly broad bracket patterns
---
### December 23, 2025 - Version 1.9.62: Dead Code Cleanup & Code Quality
**🧹 Dead Code Removal (~230+ lines)**
Removed deprecated and unused methods from the codebase:
- `toggle_sidebar()` - Quick Access sidebar removed long ago
- `update_sidebar_recent_files()` - Quick Access sidebar removed
- `handle_ribbon_action()` - Ribbon UI removed, replaced by menu bar
- `create_toolbar()` - Toolbar removed, replaced by ribbon (then menu)
- `_create_llm_settings_tab()` - Deprecated redirect, never called
- `_render_paragraph()` - Explicitly marked "no longer used"
- `_handle_target_text_debounced()` - Superseded by `_by_id` version
- `update_for_segment()` (termview) - Deprecated, use `update_with_matches()`
- `tokenize_source()` (termview) - Deprecated, use `tokenize_with_multiword_terms()`
**📋 Project Dataclass Cleanup**
- Added missing `spellcheck_settings: Dict[str, Any]` field to Project dataclass
- Added initialization in `__post_init__`
- Removed unnecessary `hasattr()` checks throughout codebase (fields now properly declared)
**🔇 Debug Logging Cleanup**
- Removed verbose TERMVIEW debug print statements from `termview_widget.py`
- Removed "💾💾💾 FINAL DEBUG" logging from project save
- Removed excessive "🔍 TERMVIEW:" log messages from main app
**🤖 AutoFingers UI Simplification**
- Removed unnecessary QTabWidget wrapper (was single tab "🎮 Control Panel")
- Control panel now displays directly without tab overhead
**Files Modified:**
- `Supervertaler.py` - Removed deprecated methods, added dataclass field, cleaned debug logs
- `modules/termview_widget.py` - Removed deprecated methods and debug prints
---
### December 22, 2025 - Version 1.9.60: Tag-Aware TM Matching & AutoFingers Cleanup
**🔍 Tag-Aware TM Matching**
Translation Memory fuzzy matching now works regardless of whether segments contain formatting tags:
- **Dual Search Strategy**: Searches both with original text (including tags) AND with tags stripped
- `<b><u>Technisch mankement</u></b>` generates search terms: `['Technisch', 'mankement']`
- Also includes any tag-derived terms (in case TM was indexed with tags)
- Single FTS query with combined terms - negligible performance impact
- **Tag-Stripped Similarity Calculation**: `calculate_similarity()` now strips HTML/XML tags before comparing
- `<b>Hello</b>` vs `Hello` now gives 100% match instead of ~70%
- More accurate fuzzy match percentages
- **Benefits**:
- TMs imported with tags now match segments without tags (and vice versa)
- Similarity scores based on actual text content, not tag presence
- Debug output shows combined search terms
**Files Modified:**
- `modules/database_manager.py` - `search_fuzzy_matches()` dual search, `calculate_similarity()` tag stripping
**🧹 TMX Tag Cleaner - Additional Tags**
- Added `<li-b></li-b>` (bullet list item) and `<li-o></li-o>` (ordered list item) tags to Formatting category
- Both enabled by default
**Files Modified:**
- `modules/tmx_editor_qt.py` - Added list item tags to `DEFAULT_TAG_PATTERNS`
**🤖 AutoFingers Cleanup**
- Removed TMX Manager tab (unused features)
- Added "📥 Import from TM" button to Control Panel's TMX File group
- Simplified TMX status display: now shows "✓ X TUs loaded" instead of verbose multi-line stats
- Removed unused `create_tmx_tab()` method
**Files Modified:**
- `Supervertaler.py` - AutoFingers UI cleanup, removed tab, added button
---
### December 22, 2025 - Version 1.9.59: TMX Tag Cleaner & UX Improvements
**🧹 TMX Tag Cleaner - Complete Implementation**
Tag cleaning functionality in both TMX Editor and main application:
- **TMX Editor Access**:
- Toolbar button: "🧹 Clean Tags"
- Edit menu: Edit → Bulk Operations → 🧹 Clean Tags...
- Tools menu: Tools → 🧹 Clean Tags...
- **Main App Access**:
- Edit menu: Edit → Bulk Operations → 🧹 Clean Tags...
- Cleans tags from currently loaded project segments
- **Tag Selection Dialog**:
- User can select/deselect individual tag patterns to clean
- Tags grouped by category: Formatting, TMX/XLIFF, memoQ, Trados, Generic
- Quick select buttons: "Select All", "Select None", "Select Formatting Only"
- **Supported Tag Types**:
- **Formatting**: `<b>`, `<i>`, `<u>`, `<bi>`, `<sub>`, `<sup>`
- **TMX/XLIFF**: `<bpt>`, `<ept>`, `<ph>`, `<it>`, `<hi>`, `<ut>`
- **memoQ**: `{1}`, `[2}`, `{3]` index tags
- **Trados**: `<1>`, `</1>` numbered tags, `{1}`, `{/1}` curly tags
- **Generic**: All XML-style tags
- **Replacement Options**:
- Remove tags completely (no replacement)
- Replace tags with a space
- **Scope Options**:
- Clean both source and target segments
- Clean source segments only
- Clean target segments only
- **Implementation**:
- Works in both RAM mode and database mode
- Progress dialog with cancel option
- Shows statistics on completion (TUs modified, tags removed)
- Automatically refreshes grid after cleaning
**New Class:**
- `TmxTagCleanerDialog` - Configuration dialog with checkboxes for each tag pattern
**Files Modified:**
- `modules/tmx_editor_qt.py` - Added `TmxTagCleanerDialog`, `show_clean_tags_dialog()`, `clean_tags()`, toolbar button, menu items
---
### December 22, 2025 - Version 1.9.57: Flattened Tab Structure
**🏠 Simplified Main Navigation**
Reorganized the main tab structure from nested to flat hierarchy:
- **Before**: Workspace (containing Editor + Resources subtabs) → Tools → Settings
- **After**: Project editor → Project resources → Tools → Settings
**Changes Made:**
- Removed nested `project_home_tabs` QTabWidget
- Grid widget now added directly to `main_tabs` as "📝 Project editor"
- Resources tab added directly to `main_tabs` as "🗂️ Project resources"
- Updated all tab index references throughout codebase (Tools is now index 2, Settings is now index 3)
- Updated View → Navigate To menu items
- Updated keyboard shortcut action mappings
- Updated all navigation methods: `_go_to_superlookup()`, `_open_superdocs_tab()`, `show_autofingers()`, `show_image_extractor_from_tools()`, `_open_mt_settings()`, `_navigate_to_termbase_entry()`
**Capitalization Style:**
- Tab names use lowercase for subtabs: "Project editor", "Project resources" (NOT "Project Editor", "Project Resources")
- Tools and Settings remain capitalized as they are top-level concepts
**Files Modified:**
- `Supervertaler.py` - `create_main_layout()`, navigation menu, shortcut actions, all navigation methods
---
### December 22, 2025 - Version 1.9.56: Glossary Renaming Feature
**✏️ Glossary Renaming via Right-Click**
Fixed the glossary (termbase) renaming functionality:
- **Problem**: Editing the glossary name directly in the table appeared to work but changes were never saved to the database
- **Root Cause**: `QTableWidgetItem` in the Name column was editable by default, but no handler saved the edited value
- **Solution Implemented**:
- Added right-click context menu to glossary table with "✏️ Rename Glossary" and "🗑️ Delete Glossary" options
- New `rename_termbase()` method in `TermbaseManager` class updates the name in the database
- New `_show_termbase_context_menu()` method shows context menu at right-click position
- New `_rename_termbase_dialog()` method shows QInputDialog for entering new name
- Disabled inline editing with `setEditTriggers(QTableWidget.EditTrigger.NoEditTriggers)`
**Files Modified:**
- `modules/termbase_manager.py` - Added `rename_termbase()` method
- `Supervertaler.py` - Added context menu, rename dialog, disabled inline editing
---
### December 22, 2025 - Version 1.9.54: Superlookup UX Improvements & Terminology Rename
**🔍 Superlookup Enhancements**
Several usability fixes for the Superlookup unified lookup window:
- **Enter Key Behavior**: Pressing Enter in search box now triggers search (Shift+Enter for newline)
- **Edit in Glossary Navigation**: Right-click "Edit in Glossary" now correctly selects the glossary in the Resources tab
- **Fuzzy Search Filter**: Minimum term length filter prevents single-letter terms from matching long search queries (e.g., "A" no longer matches "Machine Fingerprint")
- **TM Source Column**: Added new "TM" column to TM results showing which Translation Memory the match came from
**📝 User-Facing Terminology Rename: Termbase → Glossary**
Comprehensive rename of user-facing terminology throughout the application:
- **Tab Names**: "Termbase Matches" → "Glossaries", "TM Matches" → "TMs"
- **Dialog Titles**: "Create New Termbase" → "Create New Glossary", "Import Termbase" → "Import Glossary"
- **Settings Labels**: "TM/Termbase lookup delay" → "TM/Glossary lookup delay", "Termbase Highlight Style" → "Glossary Highlight Style"
- **Button Text**: "Add to Termbase" → "Add to Glossary", "TM/Termbase ON" → "TM/Glossary ON"
- **Tooltips**: All termbase-related tooltips updated to use "glossary"
- **Context Menus**: "Add to Termbase" → "Add to Glossary", "Edit in Termbase" → "Edit in Glossary"
- **Warning Messages**: All QMessageBox dialogs updated to use "glossary" terminology
- **Checkboxes**: "Enable TM & Termbase Matching" → "Enable TM & Glossary Matching"
**Technical Notes:**
- Internal code (variable names, method names, database columns) remains unchanged for backward compatibility
- Only user-visible strings were modified
- Project files (.svproj) still use `termbase_settings` key for compatibility
**Files Modified:**
- `Supervertaler.py` - Comprehensive UI string updates throughout
---
### December 21, 2025 - Version 1.9.55: Lightning-Fast Filtering
**⚡ Optimized Filter Performance**
Major performance optimization for filter operations (Ctrl+Shift+F):
- **Before**: ~12 seconds to apply or clear a filter
- **After**: Instant (typically <200ms)
**Root Cause**: Filter operations were calling `load_segments_to_grid()` which recreates all QTextEdit widgets for every segment. With 500+ segments, this took ~12 seconds.
**Solution Implemented**:
- `apply_filters()` now uses `setUpdatesEnabled(False)` to batch UI changes, only shows/hides rows and applies yellow highlights without recreating widgets
- `clear_filters()` now clears only yellow filter highlights (preserves termbase/tag formatting) and unhides rows in place
- New `_clear_filter_highlights_in_widget()` method surgically removes only yellow (#FFFF00) highlights while preserving green termbase highlights and pink tag colors
**🔄 Ctrl+Shift+F Toggle Behavior**
The filter shortcut now works as a true toggle:
- **First press** (with text selected): Filters on selected text
- **Second press** (with filter active): Clears the filter immediately
**📋 Keyboard Shortcuts Update**
Added "Clear filter" entry in keyboard shortcuts (same Ctrl+Shift+F shortcut) for discoverability. Updated shortcut description to reflect toggle behavior.
**Files Modified:**
- `Supervertaler.py` - `apply_filters()`, `clear_filters()`, `_highlight_text_in_widget()`, `_clear_filter_highlights_in_widget()` (new), `filter_on_selected_text()`
- `modules/shortcut_manager.py` - Added `clear_filter` shortcut entry, updated descriptions
---
### December 21, 2025 - Version 1.9.53: Superlookup Termbase Enhancements
**📋 Improved Termbase Matches Tab**
Enhanced the Superlookup Termbase Matches tab with additional metadata columns:
- **Renamed Column Headers**: "Term (Source)" → "Source", "Translation (Target)" → "Target"
- **New Columns Added**: Termbase (source name), Domain, Notes
- **Full Metadata in Results**: Each termbase match now includes domain, notes, priority, project, client, forbidden status
- **Tooltips**: Hover over cells to see full content (especially useful for long notes)
**📥 Termbase Import Progress Dialog**
New real-time progress dialog when importing termbases from TSV files:
- **Visual Progress Bar**: Shows percentage and entry count (e.g., "245/500 (49%)")
- **Live Statistics**: Running counters for imported (✅), skipped (⏭️), and errors (❌)
- **Scrolling Log Window**: Dark-themed console showing each entry as it's processed
- ✅ Imported: source → target
- ⏭️ Skipped duplicate: source
- 🔄 Updated: source → target (when update mode selected)
- ❌ Line X: error message
- **Auto-Refresh**: Termbase table automatically updates term counts after import
**Files Modified:**
- `Supervertaler.py` - `create_termbase_results_tab()`, `display_termbase_results()`, `search_termbases()`, `_import_termbase()`, `_refresh_termbase_table()`
- `modules/termbase_import_export.py` - `import_tsv()` now accepts `progress_callback` parameter
---
### December 20, 2025 - Version 1.9.52: Superlookup Web Resources
**🌐 Expanded Web Resources Tab**
Complete overhaul of the Superlookup Web Resources tab with 14 reference sites and persistent sessions:
- **New Web Resources (6 added)**:
- Juremy (ISO 639-3 language codes)
- michaelbeijer.co.uk (translator's personal site)
- AcronymFinder (uppercase language codes)
- BabelNet (multilingual knowledge graph)
- Wiktionary Source & Target (separate dictionary for each language)
- **Persistent Login Sessions**:
- QWebEngineProfile with `ForcePersistentCookies` policy
- Cookies stored in `user_data/web_cache/`
- Stay logged in to ProZ, Linguee, IATE, etc.
- **Auto Language Selection**:
- `set_project_languages()` method sets Superlookup dropdowns from project
- Called automatically on project load
- Language pair auto-fills when opening Superlookup
- **Compact UI Layout**:
- Single-line search bar with direction controls
- "Search" label replaced with 🔍 icon
- Simplified direction labels: "Both", "Source", "Target"
- Settings checkboxes control sidebar button visibility
- **Language Code Formats**:
- iso2, iso3, iso639_3, full_lower, iso2_upper
- `_get_web_lang_code()` handles all format conversions
- `_build_web_search_url()` with sl_upper/tl_upper placeholders
**Files Modified:**
- `Supervertaler.py` - Web resources tab, language handling, persistent profile, UI layout
---
### December 20, 2025 - Version 1.9.51: Superlookup MT Integration
**🔍 Complete Machine Translation in Superlookup**
Full MT integration with multiple providers and improved error handling:
- **MT Provider Status Display**: New compact status panel in Machine Translation tab
- Shows active providers (✅), disabled providers (⏸️), and missing API keys (❌)
- "⚙️ Configure in Settings" link navigates directly to Settings → MT Settings
- **All MT Providers Now Working**:
- Google Translate, Amazon Translate, DeepL, Microsoft Translator, ModernMT, MyMemory
- Error messages now displayed in red with full details (no more silent failures)
- Successful translations shown in blue with copy button
- **Language Name Mapping Fix**: Critical fix for all MT providers
- App stores languages as full names ("Dutch", "English")
- MT APIs require ISO codes ("nl", "en")
- Added `lang_name_to_code` mapping dictionary to Google Translate, Amazon Translate, MyMemory
- Supports 24+ languages including European, Asian, and Middle Eastern
- **Dependencies**: Added `boto3` and `deepl` to requirements.txt
- **Termbases Tab Improvements**:
- Search filter now functional (filters termbase list as you type)
- New split-view with editable terms grid on right side
- All term columns visible: Source, Target, Priority, Domain, Notes, Project, Client, Forbidden
- **Cleanup**:
- Removed debug print spam (ROW COLOR DEBUG messages)
- Removed redundant MT sub-tab from Superlookup Settings (MT config now in main Settings → MT Settings)
**Files Modified:**
- `Supervertaler.py` - MT provider calls, language mapping, Superlookup MT tab, Termbases split view
- `requirements.txt` - Added boto3, deepl
---
### December 18, 2025 - Version 1.9.50: Voice Commands Complete
**🎤 Voice Commands System - Final Polish**
Complete hands-free translation system with all features working:
- **OpenAI Whisper API Integration**: Added dual recognition engine support
- OpenAI Whisper API (recommended): Fast, accurate, works great for short voice commands
- Local Whisper model (fallback): Works offline but less accurate for short clips
- Recognition engine dropdown in Supervoice settings
- **Grid Toolbar Button**: New 🎧 Voice ON/OFF toggle button in the grid toolbar
- Shows current state: 🎧 Voice ON (green), 🔴 REC (red), ⏳ Processing (orange)
- Click to toggle always-on listening without going to settings
- **Status Bar Indicator**: Always-on indicator in status bar (bottom-right)
- 🎤 VOICE ON / 🔴 REC / ⏳ ...
- Clickable to toggle on/off
- **Bug Fixes**:
- Fixed `copy_source_to_grid_target()` not working - was using wrong column (1 instead of 3) and wrong method (`table.item()` instead of `table.cellWidget()`)
- Fixed `clear_grid_target()` same issue
- Fixed `get_api_key()` AttributeError - changed to `load_api_keys()` which returns dict
**Files Modified:**
- `Supervertaler.py` - Grid toolbar button, status bar indicator, API key loading fix, copy/clear target fixes
- `modules/voice_commands.py` - OpenAI Whisper API transcription path
---
### December 18, 2025 - Version 1.9.49: Always-On Voice Listening
**🎧 Always-On Listening Mode**
New VAD-based (Voice Activity Detection) continuous listening mode that eliminates the need to press F9 twice:
- **How it works:**
1. Continuously monitors microphone audio levels
2. When speech detected (RMS above threshold) → starts recording
3. When silence detected (0.8s of quiet) → stops and transcribes
4. Processes result as command or dictation
5. Repeats automatically
- **Settings UI:** Settings → 🎤 Supervoice → "🎧 Always-On Listening Mode" section:
- Start/Stop toggle button with status indicator
- Visual feedback: 🟢 Listening → 🔴 Recording → ⏳ Processing
- Microphone sensitivity: Low/Medium/High (for different environments)
- **F9 Behavior Update:**
- If always-on is active: F9 stops it
- If always-on is inactive: F9 works as push-to-talk (original behavior)
- **Technical Details:**
- VAD uses amplitude-based speech detection (RMS threshold)
- Minimum speech duration: 0.3s (ignores short sounds)
- Maximum recording: 15s (prevents runaway recordings)
- Silence timeout: 0.8s (adjustable via sensitivity)
- Model loaded once, cached for fast subsequent transcriptions
**Classes Added/Modified:**
- `ContinuousVoiceListener` - Complete rewrite with VAD
- `_VADListenerThread` - New background thread with proper VAD loop
- Settings persistence for sensitivity level
**Files Modified:**
- `modules/voice_commands.py` - ContinuousVoiceListener, _VADListenerThread
- `Supervertaler.py` - Import, initialization, toggle function, UI, signal handlers
---
### December 18, 2025 - Version 1.9.48: Voice Commands System (Talon-style)
**🎤 Voice Command System**
New Talon-style voice command system that lets users control Supervertaler and other applications by voice:
- **3-Tier Architecture:**
- **Tier 1: Internal Commands** - Control Supervertaler (confirm segment, navigate, translate, etc.)
- **Tier 2: System Commands** - AutoHotkey integration for controlling other apps (memoQ, Trados, Word)
- **Tier 3: Dictation Fallback** - If no command matches, insert as text
- **Built-in Commands:**
- Navigation: "next segment", "previous", "first segment", "last segment"
- Editing: "confirm", "copy source", "clear target", "undo", "redo"
- Translation: "translate", "translate all"
- Lookup: "lookup", "concordance"
- memoQ: "glossary" (Alt+Down), "tag next" (multi-key sequence)
- Trados: "confirm trados" (Ctrl+Enter)
- **Settings UI:** Settings → 🎤 Supervoice tab now includes:
- Enable/disable voice commands toggle
- Voice commands table with all phrases, aliases, actions
- Add/Edit/Remove custom commands
- Reset to defaults button
- AutoHotkey status indicator
- **Custom Command Support:**
- Phrase + aliases (fuzzy matching)
- Action types: internal, keystroke, AHK inline code, AHK script file
- Categories for organization
- User commands stored in `user_data/voice_commands.json`
**New Files:**
- `modules/voice_commands.py` - VoiceCommandManager, VoiceCommand dataclass, fuzzy matching
- `voice_commands.ahk` - AutoHotkey v2 bridge script for system-level automation
- `user_data/voice_commands.json` - User command library (auto-created)
**Files Modified:**
- `Supervertaler.py` - Added VoiceCommandManager, VoiceCommandEditDialog, enhanced Supervoice settings
---
### December 18, 2025 - Version 1.9.46: Workspace UI Redesign
**🏠 New Tab Hierarchy**
Cleaner, more intuitive tab structure:
- **Main tabs**: 🏠 Workspace → 🛠️ Tools → ⚙️ Settings
- **Workspace subtabs**: 📝 Editor (the grid) + 🗂️ Resources (TM, Termbases, Prompts, etc.)
- Removed Document View (unused feature)
- Simplified View menu (removed Grid/Document view switcher)
**Naming Philosophy:**
- "Workspace" - generic term that works for translation, localization, and copywriting
- "Editor" - describes what you do (edit content), not the UI (grid)
- "Resources" - all project resources in one place
**🐛 Critical Bug Fix: Termbase Activation**
Fixed termbase matches showing terms from non-activated termbases:
- Added `AND (ta.is_active = 1 OR tb.is_project_termbase = 1)` filter to `search_termbases()` query
- Now only returns terms from termbases with Read checkbox enabled
**Files Modified:**
- `Supervertaler.py` - `create_main_layout()`, navigation menu, view menu
- `modules/database_manager.py` - `search_termbases()` activation filter
---
### December 18, 2025 - Version 1.9.47: Code Cleanup
**🧹 Dead Document View Code Removed**
Removed ~811 lines of unused Document View code. The Document View feature was never used in production - the Grid View (Editor) is the primary and only workflow.
**Removed Items:**
- `LayoutMode` class (GRID/DOCUMENT enum)
- `create_editor_widget()` - View switcher with Grid/List/Document buttons
- `create_home_tab()`, `create_projects_manager_tab()` - Deprecated wrappers
- `create_editor_tab()`, `switch_view_mode()` - View switching logic
- `switch_home_view_mode()` - Home tab view switching
- `create_document_view_widget()`, `create_document_view_widget_for_home()` - Document view creation
- `refresh_document_view()` - ~300 lines of document rendering code
- `_find_document_container()`, `_register_document_container()`, `_set_active_document_host()`, `_get_document_container()`, `_locate_document_container()` - Container helpers
- `_find_widget_by_object_name()` - Widget search helper
- Document View Methods section (on_doc_segment_clicked, on_doc_status_change, etc.)
- Document view state variables (doc_segment_widgets, doc_current_segment_id, document_containers, active_document_host, current_view_mode)
**Result:**
- File reduced from 35,249 to 34,438 lines
- Cleaner codebase, easier to maintain
- No functional changes - Grid View (Editor) is unaffected
---
### December 17, 2025 - Version 1.9.45: Termbase Highlight Styles & Spellcheck Auto-Language
**🏷️ Configurable Termbase Highlight Styles**
Three visual styles for termbase term highlighting in the translation grid:
- **Background (Default)**: Pastel green shades based on priority (existing behavior)
- **Dotted Underline**: DotLine underline with priority-based colors
- Priority 1: Red
- Priority 2-3: Gray shades
- Priority 4+: User-configurable color (default dark green)
- Reset button to restore default color
- **Semibold**: DemiBold font weight with tinted green foreground colors
**Settings UI:**
- New "🏷️ Termbase Highlight Style" section in Settings → View Settings
- Radio button selection for style type
- Color picker for dotted underline (Priority 4+)
- Reset button to restore default color
**🔤 Spellcheck Auto-Language Initialization**
Fixed spellcheck not using project's target language:
- **All Import Handlers**: DOCX, TXT, memoQ, CafeTran, Trados bilingual, SDLPPX, Phrase now initialize spellcheck for target language
- **Project Load**: Always sets spellcheck to project target language (ignores old saved language)
- **Short Code Support**: Added `SHORT_CODE_MAP` in spellcheck_manager.py to handle "nl" → "nl_NL", "de" → "de_DE", etc.
**Bug Fixes:**
- Fixed NT list crash (`lst.active` → `lst.is_active`)
- Fixed NonTranslatable attributes (`pattern` → `text`, `description` → `notes`)
- Fixed Ctrl+K going to Settings instead of Superlookup (tab index 3 → 2)
- Renamed Grid/Document tabs to "Grid View"/"Document View"
**Files Modified:**
- `Supervertaler.py` - `highlight_termbase_matches()`, `_create_view_settings_tab()`, `_save_view_settings_from_ui()`, all import handlers, `load_project()`
- `modules/spellcheck_manager.py` - Added `SHORT_CODE_MAP`, updated `set_language()` to handle short codes
---
### December 16, 2025 - Version 1.9.41: Dark Mode Complete Implementation
**🌙 Full Dark Theme Support**
Completed comprehensive dark mode implementation after extensive debugging session:
- **Compare Boxes Fixed**: Translation Results panel compare boxes (Current Source, TM Source, TM Target) now properly dark in dark mode
- **Termview Visibility**: All words in Termview pane now visible - non-matched words use light text color
- **Root Cause Discovery**: Qt doesn't reliably apply stylesheets to hidden widgets
- **Solution**: Added `_apply_compare_box_theme()` method called when compare frame becomes visible
**Technical Changes:**
- `modules/translation_results_panel.py` - Added `_apply_compare_box_theme()`, uses both stylesheet AND QPalette for reliability
- `modules/termview_widget.py` - Made `TermBlock` and `NTBlock` theme-aware with `theme_manager` parameter
**Key Lesson Learned:**
When Qt stylesheets aren't visually applying despite the code running correctly, consider:
1. Widget visibility state at time of styling
2. Using QPalette as an alternative/supplement to stylesheets
3. Re-applying styles when widget becomes visible
---
### December 17, 2025 - Version 1.9.42: Multi-File Project Support
**📁 Import Folder (Multiple Files)**
Major new feature allowing users to import entire folders of files as a single multi-file project:
- **New Menu Item**: File → Import → Folder (Multiple Files)...
- **Supported Formats**: DOCX and TXT files in selected folder
- **File Selection Dialog**: Preview files with size, select/deselect individual files
- **Language Pair Selection**: Set source/target language for all files
- **Progress Dialog**: Shows import progress with per-file segment counts
**🗂️ Per-File Progress Tracking**
- **File Progress Dialog**: View → File Progress... (or click on status bar)
- **Per-File Statistics**: Segments, words, translated, confirmed, progress bar for each file
- **Status Indicators**: ✅ Complete, 📝 Translated, 🔄 In Progress, ⬜ Not Started
- **Navigation**: Double-click file to jump to its first segment
**📊 Status Bar Enhancement**
- **Files Indicator**: Shows "📁 Files: X/Y" for multi-file projects (completed/total)
- **Clickable**: Click the files indicator to open File Progress dialog
- **Tooltip**: Hover for file count and completion summary
**🔍 File Filter Dropdown**
- **New Dropdown**: Appears in filter panel for multi-file projects
- **Filter by File**: Select a specific file to show only its segments
- **All Files**: Default option to show all segments from all files
**Data Model Changes:**
- `Segment` dataclass: Added `file_id` (int) and `file_name` (str) fields
- `Project` dataclass: Added `files` (list of file metadata) and `is_multifile` (bool) fields
- Backward compatible: Single-file projects work unchanged
**📤 Export Folder (Multiple Files)**
- **New Menu Item**: File → Export → Folder (Multiple Files)...
- **Format Options**: TXT (plain text), DOCX (formatted), Bilingual (Source/Target table)
- **File Preview**: Shows all files with format, segment count, and status
- **Progress Dialog**: Shows export progress with per-file updates
**New Methods in Supervertaler.py:**
- `import_folder_multifile()` - Folder import dialog
- `_import_multifile_project()` - Import multiple files into single project
- `export_folder_multifile()` - Folder export dialog
- `_export_multifile_to_folder()` - Export multiple files to folder
- `_export_file_as_txt()` - Export single file as plain text
- `_export_file_as_docx()` - Export single file as formatted DOCX
- `_export_file_as_bilingual()` - Export single file as bilingual DOCX table
- `show_file_progress_dialog()` - Per-file progress dialog
- `_add_overall_progress_section()` - Overall progress stats
- `_on_file_filter_changed()` - File filter dropdown handler
- `_update_file_filter_combo()` - Populate file filter dropdown
**🔧 Source File Backup & Recovery (added later)**
- **Automatic Backup**: Source files now copied to `_source_files/` folder inside project directory during import
- **Relocate Source Folder**: File → Relocate Source Folder... menu item to fix broken paths when source files are moved
- **Export Warning**: Shows helpful message when source files are missing, suggests using Relocate feature
**🔍 Superlookup Fixes**
- **Class Renamed**: `UniversalLookupTab` → `SuperlookupTab` for consistency
- **Theme Support**: Added `theme_manager` attribute for theme-aware search term highlighting
- **Fixed Error**: Resolved `'UniversalLookupTab' object has no attribute 'theme_manager'` error
**📋 Spellcheck Info Dialog Redesign**
- **Compact Layout**: Completely redesigned to fit on screen without scrolling off the bottom
- **Horizontal Top Row**: Status, language dropdown, and backend indicator on one line
- **Collapsible Troubleshooting**: Diagnostics section collapsed by default (click to expand)
- **Inline Links**: Download links for Hunspell dictionaries in horizontal row with separators
- **Max Height**: Dialog limited to 500px to prevent overflow
---
### December 12, 2025 - Trados Bilingual DOCX Workflow Documentation
**📚 Comprehensive Trados Workflow Documentation**
Added critical workflow documentation for Trados Bilingual Review DOCX format:
- **Import dialog warning**: `import_trados_bilingual()` now shows detailed ⚠️ CRITICAL WORKFLOW warning explaining the required preparation steps
- **CAT_WORKFLOW.md**: Updated Trados Studio section with both SDLPPX (recommended) and Bilingual DOCX (workaround) workflows
- **QUICK_START.md**: Added warning note directing users to full documentation
**The Critical Workflow:**
The Trados Bilingual Review format is designed for **review only**, not translation - it doesn't export empty target segments! ([RWS Community reference](https://community.rws.com/product-groups/trados-portfolio/trados-studio/f/studio/34874/export-for-bilingual-review-exports-only-source-text))
Required workaround:
1. In Trados: Copy source to target (fills empty segments)
2. Export as Bilingual Review DOCX
3. In Word: Delete all target text (cells exist but are empty)
4. Import into Supervertaler, translate, export
5. Reimport into Trados Studio
**Files Modified:**
- `Supervertaler.py` - `import_trados_bilingual()` warning dialog (line ~18562)
- `docs/guides/CAT_WORKFLOW.md` - Comprehensive Trados section with both workflows
- `docs/guides/QUICK_START.md` - Warning note with link to full documentation
---
### December 12, 2025 - Version 1.9.40: Superlookup Unified Concordance System
**🔍 Ctrl+K Now Opens Superlookup**
- Major consolidation: Concordance search now uses Superlookup instead of separate dialog
- All lookup resources in one place: TM, Termbase, Supermemory, MT, Web Resources
- Selected text in source/target automatically populates search field
- `show_concordance_search()` now calls `_go_to_superlookup()` and triggers search
**📊 Dual-View Toggle for TM Matches**
- Horizontal (Table): Source | Target columns side-by-side - compact, scannable
- Vertical (List): Dutch: ... / English: ... stacked - traditional concordance layout
- Radio button toggle in TM Matches tab
- Both views stay in sync with search results
- `toggle_tm_view_mode()` and `display_tm_results()` updated for dual display
**🗂️ Tab Reorganization**
- "Resources" renamed to "Project Resources"
- Tab order changed: Project Editor → Project Resources → Prompt Manager → Tools → Settings
- Removed "Concordance" and "Import/Export" tabs from Translation Memories (redundant)
- Source text box in Superlookup shrunk from 100px to 50px
- "Termbase Terms" renamed to "Termbase Matches"
**⚡ FTS5 Full-Text Search Optimization**
- `concordance_search()` in database_manager.py now uses FTS5 MATCH queries
- 100-1000x faster than previous LIKE queries on large databases
- Auto-sync: FTS5 index rebuilt on connect if out of sync
- New methods: `rebuild_fts_index()`, `check_fts_index()`
**🐛 ChromaDB Stability Fix**
- Removed all `collection.count()` calls that caused native Rust crashes
- Stats now use SQLite metadata count instead of ChromaDB collection queries
- ChromaDB 0.6.3 with tokenizers 0.22.0 (stable combination)
**Files Modified:**
- `Supervertaler.py` - show_concordance_search(), create_tm_results_tab(), display_tm_results(), tab ordering
- `modules/database_manager.py` - concordance_search() FTS5, rebuild_fts_index(), check_fts_index()
- `modules/supermemory.py` - Removed collection.count() calls, uses metadata for stats
---
### December 11, 2025 - Version 1.9.39: Superlookup Multilingual Search
**🔍 Multilingual Language Filtering**
- Added From/To language dropdown filters in Superlookup search bar
- Filter TM and termbase searches by source/target language pair
- Languages auto-populate from TMs and termbases on first tab view
- Alphabetically sorted with language family grouping (all Dutch variants together, etc.)
- Display format: "English (en)", "Dutch (nl-BE)" for clarity
**↔️ Search Direction Controls**
- Radio buttons: Both (bidirectional), Source only, Target only
- Concordance search respects direction setting
- Termbase search also respects direction for bidirectional term lookup
**🎨 UI Improvements**
- Yellow highlighting of search terms in TM/termbase results
- Compact results display with word wrap and 60px max row height
- Tooltips show full text on hover
- Hidden row numbers for cleaner display
- Removed Manual Capture button (redundant with paste)
- Removed Operating Modes dropdown (only Universal mode used)
**Files Modified:**
- `Supervertaler.py` - UniversalLookupTab UI, language dropdowns, direction radio buttons, display methods
- `modules/superlookup.py` - search_tm() accepts direction and language parameters
- `modules/translation_memory.py` - concordance_search() accepts language filters
- `modules/database_manager.py` - concordance_search() filters by source_lang/target_lang
---
### December 11, 2025 - Version 1.9.38: Project File & UX Improvements
**📁 Reorganized .svproj File Structure**
- Metadata now at top of file (name, languages, dates, ID)
- Settings next (prompts, TM, termbases, spellcheck)
- Source paths follow (DOCX, memoQ, Trados, CafeTran, SDLPPX)
- Segments moved to END of file for easier human inspection
**💡 Improved Batch Translate Warning**
- Added tip about using Select All + Clear Target from right-click menu
- Users no longer need to re-import memoQ files just to clear targets
**Files Modified:**
- `Supervertaler.py` - `Project.to_dict()`, batch translate warning message
---
### December 11, 2025 - Version 1.9.37: User-Configurable Grid Fonts
**🔤 Font Customization in Settings → View Settings**
- Added font family dropdown with 10 popular fonts (Calibri, Segoe UI, Arial, Consolas, Verdana, Times New Roman, Georgia, Courier New, Tahoma, Trebuchet MS)
- Added live preview panel showing sample source/target text with tags - updates in real-time as you adjust settings
- Font family now persists between sessions (previously only font size was saved)
- Fixed font size spinbox up/down arrows with improved styling and click targets
- Added friendly note: "If your favourite font is missing, contact the developer!"
**Files Modified:**
- `Supervertaler.py` - `_create_view_settings_tab()`, `_save_view_settings_from_ui()`, `load_font_sizes_from_preferences()`, `save_current_font_sizes()`
---
### December 10, 2025 - Version 1.9.35: memoQ Red Tag Color Fix + Universal Tag Coloring
**🔴 memoQ Inline Tag Color Preservation (Export)**
Fixed critical issue where red/magenta inline tags (e.g., `{1}`, `[2}`) in memoQ bilingual exports were appearing as black text in the target column.
**Root Cause Discovery:**
- memoQ stores tag colors in **character styles** (e.g., `mqInternal`), not directly on the run
- The `mqInternal` style defines color `800000` (dark red) at the style level
- `python-docx` API (`run.font.color.rgb`) returns `None` for style-based colors
**Solution Implemented:**
- Added 3-tier color extraction method in `export_memoq_bilingual()`:
1. Check `run.font.color.rgb` (direct color)
2. Check XML `w:color` element in `rPr` (inline XML color)
3. **NEW**: Check `w:rStyle` element, lookup character style, extract style's color
---
**🎨 Universal Tag Coloring in Grid (Display)**
Extended `TagHighlighter` to color ALL CAT tool tags with pink (`#FFB6C1`) in the translation grid:
| Format | Tag Examples | Status |
|--------|-------------|--------|
| HTML/XML | `<b>`, `<i>`, `<u>`, `<li-o>` | ✅ Already worked |
| memoQ | `{1}`, `[2}`, `{3]`, `[4]` | ✅ NEW |
| Trados | `<1>`, `</1>` | ✅ NEW |
| Phrase | `{1}`, `{2}` | ✅ NEW |
**CafeTran Pipe Fix:**
- Pipe symbols (`|`) now only highlighted red in **CafeTran projects**
- Previously, pipes were red in ALL project types (bug)
- Added `TagHighlighter._is_cafetran_project` class flag
**Files Modified:**
- `Supervertaler.py` - `TagHighlighter.highlightBlock()`, CafeTran import function
### December 10, 2025 - Version 1.9.34: UI Fixes
**🎨 Global UI Standardization**
- Replaced standard OS radio buttons with custom green `CheckmarkRadioButton`.
- Unified design across Filters, Import, AutoFingers, and Find/Replace dialogs.
- Replaced default radio buttons in Import TMX dialogue with standard `CheckmarkRadioButton` widgets
- Ensures consistency with application's green UI theme
---
### December 10, 2025 - Version 1.9.32: Trados SDLRPX Status Fix
**📦 Fixed Critical SDLRPX Export Bug**
- Fixed segments staying in "Draft" status instead of being updated to "Translated" in exported SDLRPX packages
- Trados Studio now correctly recognizes translated segments when client opens return package
- Added `_update_segment_status()` method to `modules/sdlppx_handler.py`
- Updates `conf` attribute in `sdl:seg-defs` section of SDLXLIFF files
**Root Cause:**
- The `_update_xliff_tree()` function was updating target text but not updating the `conf` attribute
- SDL/Trados uses `conf="Draft"` vs `conf="Translated"` in `<sdl:seg>` elements
---
### December 10, 2025 - Version 1.9.33: Spellcheck Update Fix
**🐛 Fixed Spellcheck Highlighting Bug**
- Fixed issue where adding/ignoring words only removed underline in the current cell
- Now triggers global refresh of all highlighters
- Modified `_add_to_dictionary` and `_ignore_word` in `EditableGridTextEditor`
---
### December 10, 2025 - Version 1.9.31: Spellcheck Language Fix
- Spellcheck now correctly uses the project's target language instead of defaulting to English
- Added language dropdown in Spellcheck Info dialog
---
### December 10, 2025 - Version 1.9.30: Critical LLM Fix
- Removed hardcoded debug file path that caused "No such file or directory" errors
---
### December 10, 2025 - Version 1.9.29: Spellcheck Integration
- Red wavy underlines for misspelled words
- Right-click context menu with suggestions
- Add to Dictionary, Ignore Word
- New module: `modules/spellcheck_manager.py`
---
### December 9, 2025 - Version 1.9.28: Phrase DOCX & Show Invisibles
- Phrase (Memsource) bilingual DOCX support
- Show Invisibles feature (spaces, tabs, line breaks)
---
### December 9, 2025 - Version 1.9.27: Simple Text Import/Export
- Import text files (each line = segment)
- Export translations as text
---
### December 8, 2025 - Version 1.9.26: Model Version Checker
- Auto-detect new LLM models from OpenAI, Anthropic, Google
---
### December 7, 2025 - Version 1.9.24: Smart Word Selection
- Selecting part of word expands to full word
---
### December 5, 2025 - Version 1.9.20: SDLPPX Persistence
- Package path saved in .svproj files
---
### December 4, 2025 - Version 1.9.19: Trados Package Support
- Import SDLPPX, export SDLRPX
- New module: `modules/sdlppx_handler.py`
---
### December 3-4, 2025 - Version 1.9.17-18: Supermemory
- Domain management, filtering, Superlookup integration
- Concordance Search with semantic tab
---
### December 2, 2025 - Supermemory: Vector TM
- ChromaDB + Sentence-Transformers
- Semantic search across TMs
- Module: `modules/supermemory.py` (2100+ lines)
---
### December 1, 2025 - Version 1.9.16: Ollama Support
- Local LLM translation with Ollama
---
### November 30, 2025 - Versions 1.9.13-15
- Document Preview tab
- Bilingual Table export/import
---
### November 27-28, 2025 - Versions 1.9.8-11
- CafeTran integration
- Navigation improvements
---
### November 8-9, 2025 - Unified Prompt System
- 2-layer prompt architecture
- AI Assistant for prompt generation
---
### November 7, 2025 - TagCleaner
- CAT tool tag removal module
---
### November 6, 2025 - LLM Integration Complete
- Multi-LLM support operational
---
### January 31, 2026 - Performance Optimization Session
**Problem:** Grid navigation was extremely slow compared to memoQ. User reported waiting times when clicking between segments.
**Root Causes Identified:**
1. **Destructive Cache Invalidation** - Every Ctrl+Enter (confirm segment) was calling `invalidate_translation_cache()` which cleared 20-45 future segments from the cache, destroying the batch worker's pre-populated data.
2. **Non-existent TM Method** - The prefetch worker was calling `db_manager.search_translation_memory()` which doesn't exist. This caused silent failures and 0 TM matches in prefetch.
3. **Empty Results Cached** - The prefetch worker was caching empty results (`{TM: [], Termbases: [], ...}`), which blocked future lookups from finding actual matches.
4. **Excessive Debug Prints** - Dozens of debug print statements firing on every navigation:
- `[HIGHLIGHT DEBUG]` prints in `highlight_termbase_matches`
- `[PROACTIVE DEBUG]` prints in `_trigger_idle_prefetch` and `_apply_proactive_highlighting`
- `[PROACTIVE NAV DEBUG]` prints in `on_cell_selected`
- `[DEBUG] Superlookup` prints in `perform_lookup`
- `[DEBUG] TMDatabase.search_all` prints (7 per TM search)
- `📋 Found X active termbases` logs firing constantly
- `🎯 TranslationResultsPanel.set_matches()` debug output
5. **No Debounce for Mouse Clicks** - Arrow key navigation had 75ms debounce, but mouse clicks went directly to full lookup.
**Changes Made:**
| File | Change |
|------|--------|
| `Supervertaler.py:22950-22955` | Removed `invalidate_translation_cache()` call on TM save |
| `Supervertaler.py:22755-22779` | Replaced broken TM lookup in prefetch with `pass` (skip TM in prefetch) |
| `Supervertaler.py:22686-22689` | Changed to only cache if `total_matches > 0` |
| `Supervertaler.py:31387` | Reduced debounce from 75ms to 10ms |
| `Supervertaler.py:31370-31396` | Applied debounce to ALL navigation (including mouse clicks) |
| `Supervertaler.py:1674-1835` | Removed `[HIGHLIGHT DEBUG]` prints |
| `Supervertaler.py:22568-22617` | Removed `[PROACTIVE DEBUG]` prints |
| `Supervertaler.py:31903-31944` | Removed `[PROACTIVE NAV DEBUG]` prints |
| `Supervertaler.py:34568-34633` | Cleaned up `_apply_proactive_highlighting` debug prints |
| `Supervertaler.py:10325-10330` | Removed SuperlookupTab initialization debug prints |
| `Supervertaler.py:47020-47072` | Removed Superlookup perform_lookup debug prints and file writes |
| `modules/termbase_manager.py:612` | Removed frequent `📋 Found X active termbases` log |
| `modules/translation_results_panel.py:1679-1684` | Removed `set_matches()` debug prints |
| `modules/translation_memory.py:208-253` | Removed all `[DEBUG] TMDatabase.search_all` prints |
**Remaining Issue:** ✅ **FIXED in v1.9.182** (see below)
---
### January 31, 2026 - In-Memory Termbase Index (v1.9.182) ⚡
**Problem:** Termbase lookup taking 52+ seconds per segment navigation. Ctrl+Enter felt frozen.
**Root Cause:** The `_search_termbases_thread_safe()` method was running a complex SQL UNION query **for every word** in the source text. A 50-word patent segment = 50+ database queries per segment.
For 349 segments × 50 words = **~17,500 database queries** taking 365 seconds!
**Solution:** Load ALL terms from activated termbases into memory ONCE, then do fast in-memory matching.
**New Architecture:**
1. **`_build_termbase_index()`** - Called on project load:
- Single SQL query loads all terms from activated termbases
- Pre-compiles regex patterns for word-boundary matching
- Sorts by term length (longest first) for proper phrase matching
- Stores in `self.termbase_index` (thread-safe with lock)
2. **`_search_termbase_in_memory()`** - Called for each segment lookup:
- Iterates through pre-loaded terms (typically ~1000 terms)
- Fast `in` substring check (implemented in C)
- Pre-compiled regex validation for word boundaries
- Returns matches in <1ms per segment
**Performance Improvement:**
| Metric | Before | After |
|--------|--------|-------|
| Per-segment lookup | 1-52 seconds | <1 millisecond |
| 349 segments batch | 365 seconds | <1 second |
| Ctrl+Enter response | 52+ seconds | Instant |
**Files Modified:**
- `Supervertaler.py` (v1.9.182):
- Added `self.termbase_index` and `self.termbase_index_lock`
- Added `_build_termbase_index()` method (~80 lines)
- Added `_search_termbase_in_memory()` method (~40 lines)
- Modified `_start_termbase_batch_worker()` to build index first
- Modified `_termbase_batch_worker_run()` to use in-memory search
- Modified `find_termbase_matches_in_source()` to use index when available
- Modified `on_read_toggle()` to rebuild index on termbase activation change
- Modified `_quick_add_term_with_priority()` to update index when terms added
**Key Code Locations:**
- Index building: `_build_termbase_index()` at ~line 22249
- In-memory search: `_search_termbase_in_memory()` at ~line 22346
- Index initialization: line ~6217 (`self.termbase_index = []`)
**Next Steps (if needed):**
1. Investigate why `_search_termbases_thread_safe` returns empty for most segments
2. Compare its query with `find_termbase_matches_in_source` (which works in Force Refresh)
3. Consider unified TM+Termbase lookup system (user's suggestion)
**Architecture Notes:**
Two separate caches exist:
- `termbase_cache` - Raw termbase matches per segment (populated by batch worker)
- `translation_matches_cache` - Combined TM+TB+MT+LLM matches (populated by prefetch worker)
The prefetch worker checks `termbase_cache` first, then falls back to `_search_termbases_thread_safe` if not found.
---
### February 23, 2026 — v1.9.311-313: Trados SDLXLIFF/SDLPPX Overhaul
**Context:** A user reported "Total segments: 0" when importing a Trados SDLPPX package. Investigation revealed multiple bugs in the SDLXLIFF handling and led to significant improvements across three versions.
**v1.9.311 — Self-closing mrk fix:**
- Fixed export silently skipping self-closing `<mrk mtype="seg" mid="N"/>` tags (empty/untranslated segments)
**v1.9.312 — Major SDLXLIFF improvements:**
- **Fix "Total segments: 0"**: Changed `.glob('*.sdlxliff')` to `.rglob('*.sdlxliff')` in `_load_xliff_files()` — files were in nested subdirectories
- **Lock TU xid handling**: New functions `_find_max_locked_id()`, `_remap_lock_xids()`, `_insert_lock_tus()` in `sdlppx_handler.py` ensure each target context gets unique lock TU references with fresh UUIDs
- **Already-translated segments**: Existing target lock xids are preserved (detected and re-attached after `_markers_to_xml()` strips them)
- **Self-closing mrk regex fix**: Negative lookbehind `(?<!/)` prevents cross-matching through `/>` of self-closing mrk elements
- **SDLPPX multi-file views**: Packages with multiple SDLXLIFF files now activate the multi-file views system (file filter dropdown, per-file navigation)
- **memoQ submenu**: Import/Export menu items grouped under "memoQ" submenu (matching Trados Studio pattern)
- **Import performance**: Locked/non-translatable segments filtered before UI loading (~42K to ~1.7K segments)
- **Unicode crash fix**: `→` character replaced with `->` in log messages for Windows cp1252 consoles
**v1.9.313 — Optional locked segment loading:**
- **SDLPPX info dialog**: Checkbox to include locked segments with count shown
- **Standalone SDLXLIFF/Folder imports**: Yes/No prompt when locked segments detected
- **Visual distinction**: Locked rows get muted tan/grey background + read-only target cells
- **`_map_sdlxliff_segment()`**: New `locked` parameter propagates to Segment dataclass
- Lock TU structural containers (`lockTU_*`) always excluded; only `translate="no"` segments optionally included
- Existing Advanced Filters "Locked Status" works automatically for post-import filtering
**Key Files Modified:**
- `modules/sdlppx_handler.py` — Lock TU remapping, rglob fix, Unicode fix
- `Supervertaler.py` — Multi-file views, memoQ submenu, locked segment toggle, import filtering, grid locked row styling
**Key Architecture Notes for SDLXLIFF:**
- Export uses **text-based regex replacement** (not ElementTree) to preserve BOM, XML formatting, namespaces
- Marker format: `<g id="N">` → `<N>`, `<x id="N"/>` → `<N/>` for display, reversed by `_markers_to_xml()`
- Lock elements: `<x id="lockedN" xid="lockTU_UUID"/>` reference separate `<trans-unit id="lockTU_...">` elements
- Each context (source, seg-source, target) MUST have unique lock TU xids
- `_replace_target_content()` handles 4 cases: existing target with mrk, self-closing mrk, missing target, unsegmented
---
## 🗄️ Database Schema (SQLite)
### Core Tables
- **translation_units** - TM entries
- **termbases** - Termbase definitions
- **termbase_terms** - Individual terms
- **termbase_activation** - Project termbase tracking
- **non_translatables** - Locked terms
- **projects** - Translation projects
### Naming Conventions
- ✅ Use "Termbase" (one word)
- ❌ Never use "Glossary" or "glossary_terms"
---
## � Planned Features
### ✅ AI Proofreading System (IMPLEMENTED - January 7, 2026)
> **STATUS:** Fully implemented in v1.9.85
**Overview:**
An intelligent proofreading system that uses LLMs to verify translation quality. Issues are stored in the Notes field with ⚠️ PROOFREAD prefix.
**Access Points:**
- Edit → Batch Operations → ✅ Proofread Translation...
- View → ✅ Proofreading Results...
- Advanced Filters → "Has proofreading issues"
- Right-click → ✅ Clear Proofreading Notes
**Features:**
- Batch processing (20 segments per API call)
- Progress dialog with real-time stats
- Results table with navigation to segments
- Bulk and individual clear operations
- Orange indicator on status icons
- Filter for segments with issues
**Files Modified:**
- `Supervertaler.py` - All implementation
---
## 📚 Additional Resources
| File | Purpose |
|------|---------|
| `CHANGELOG.md` | Complete version history |
| `README.md` | User-facing documentation |
| `FAQ.md` | Common questions |
| `docs/guides/` | User guides |
---
*This file replaces the previous CLAUDE.md and PROJECT_CONTEXT.md files.*
*Last updated: February 23, 2026 - v1.9.313 (Trados SDLXLIFF/SDLPPX overhaul + locked segment toggle)*
---
**💡 TIP FOR AI AGENTS:** When starting a new session, read the Quick Start section at the top, then jump directly to the most recent dated entry in "Recent Development History" for current context.
