v3.5.0

🚀 Major Update

Model Transition

• Upgraded to Gemini 3.5 Flash: Transitioned from Gemini 2.5 Flash to the latest Gemini 3.5 Flash model as the new default
• Updated advanced thinking capabilities to support Gemini 3 models alongside Gemini 2.5
• Default batch size increased to 1000 for better efficiency

New Features

• Token Statistics Display: Added --token-stats flag to show real-time token usage during translation and transcription
  • Displays: Prompt Tokens, Thoughts Tokens, Output Tokens, and Total Tokens
• Context Preservation Control: Added --no-context flag to disable context preservation between batches (reduces token usage)
  • New preserve_context parameter (default: True) maintains context for improved continuity

Model Configuration Enhancements

• Improved thinking mode validation for Gemini 3 Pro models

📋 Parameter Changes

API Changes

• Default model_name: "gemini-2.5-flash" → "gemini-3.5-flash"
• Default batch_size: 300 → 1000

New Parameters

• token_stats: Show token usage information (default: False)
• preserve_context: Preserve context between batches (default: True)

🔧 Technical Improvements

Error Handling

• Improved error handling for empty API responses - now throws ValueError instead of continuing
• Reduced max consecutive errors from 5 to 3
• Better handling of JSON decode errors and line-related issues

Performance Optimizations

• Token validation model updated to gemini-3.1-flash-lite
• Reduced server overload retry wait time from 3 minutes to 60 seconds
• Streamlined streaming response processing for better efficiency

Code Quality

• Extensive refactoring of main.py for improved readability
• Better logging integration with token stats throughout the translation pipeline
• Enhanced thinking signature handling
• Improved conditional logic for model-specific constraints

📦 Dependencies

Updated minimum versions:

• google-genai: >= 1.56.0 → >= 2.6.0
• json-repair: >= 0.54.3 → >= 0.59.10
• Relaxed audioop-lts version constraint

Python Version Support

• Added support for Python 3.14

📚 Documentation

• Updated README examples to reflect new model and parameters
• Added documentation for token stats and context preservation features
• Updated CLI usage examples with new flags

---

Full Changelog: v3.0.1...v3.5.0