Qt Translation Assistant Skill

Iron Laws

1. Never modify original TS files without backup - Always preserve original content
2. Validate AI translation quality - Verify translations are accurate and contextually appropriate
3. Maintain translation consistency - Use consistent terminology across all translations
4. Respect file encoding - Preserve UTF-8 encoding and special characters
5. Minimal changes principle - Only modify translation content, preserve XML structure

Red Flags

• User requests translation of non-TS files
• User asks to translate without proper AI configuration
• Requests to overwrite existing translations without verification
• Asks to translate to unsupported language codes

Rationalization Table

Excuse	Response
"Just translate everything quickly"	Quality matters in localization - proper AI configuration and validation required
"We don't need consistent terminology"	Inconsistent translations hurt user experience - consistency is critical
"Original files don't need backup"	Always preserve originals - translation errors can corrupt content
"Rewrite the whole file"	Only translation text should change - git diff will show other modifications

Quick Reference

Core Commands

# Translate entire directory of TS files
python translate.py /path/to/ts/files/

# Translate specific file
python translate.py /path/to/file.ts

# With custom batch size and workers
python translate.py /path/to/ts/files/ --batch-size 30 --max-workers 3

# Create configuration file
python translate.py --create-config

Configuration

{
  "api_url": "http://localhost:8080/v1/chat/completions",
  "api_key": "sk-uos-12345",
  "model": "qwen3-coder-flash",
  "temperature": 0.3
}

Common Mistakes & Fixes

Mistake: AI provider not configured properly

Fix: Create qt_translation_config.json with valid API credentials using --create-config

Mistake: Large files causing API timeouts

Fix: Adjust --batch-size parameter (try 20-50) and --max-workers (try 2-5)

Mistake: Language codes not detected correctly

Fix: Ensure TS files follow standard naming convention (e.g., project_zh_CN.ts, project_de.ts)

Mistake: Translation quality issues

Fix: Adjust model selection and temperature settings in configuration file

Mistake: Git diff shows many unnecessary changes

Fix: The tool only modifies translation content - any other changes indicate a bug that needs fixing

Architecture

This skill uses a parallel processing architecture:

• TranslationWorker: Handles AI API calls with automatic retry and exponential backoff
• QtTranslationAssistant: Main orchestrator with parallel batch processing
• ThreadPoolExecutor: Manages concurrent translation workers (default 3)

Performance improvements over subagent architecture:

• Direct API calls without subprocess overhead (~5-10x faster)
• Larger batch sizes (default 30 vs previous 10)
• Parallel workers (3 concurrent API calls vs sequential)
• Error isolation (single batch failure doesn't affect others)

Key Features

• Smart parsing of TS files to identify incomplete translations
• Parallel batch processing with ThreadPoolExecutor
• Support for multiple AI providers (OpenAI, Anthropic, DeepSeek, local servers)
• Configurable batch size and worker count
• Automatic retries with exponential backoff
• Git diff-friendly modifications (only changes translation content)