What's Changed
Breaking Changes
- OpenAI provider now exclusively supports GPT-5 models - Migrated from GPT-4 to GPT-5 using the new Responses API
- Supported models:
gpt-5(default),gpt-5-mini,gpt-5-nano - GPT-4 models are no longer supported with OpenAI provider
- Users of GPT-4 should switch to GPT-5 or use alternative providers (Anthropic, Gemini, Local)
- Supported models:
New Features
- GPT-5 reasoning support - Added
--reasoningflag for GPT-5 models (high, medium, low, minimal) - Native reasoning for gpt-oss models - Local Ollama models with
gpt-ossprefix now support native reasoning - Configurable output tokens - Added
--max-predict-tokensflag (default: 200,000 tokens) - Enhanced configuration system - Project-specific settings via
.git-smart-squash.yml
Improvements
- Increased default max prediction tokens from 32,000 to 64,000
- OpenAI context limit set to 400,000 tokens for GPT-5
- Better error messages and migration guidance for users switching from GPT-4
- Improved code organization with new utility modules for better maintainability
- Enhanced test coverage for new features
Internal Changes
- Major refactoring of CLI module for better separation of concerns
- Added dependency validator for better error handling
- New commit applier strategy for improved reliability
- Enhanced display utilities and git diff handling
Full Changelog: v3.7.3...v4.0.0
Migration Guide
If you were using GPT-4 models with OpenAI, you need to either:
- Switch to GPT-5:
--ai-provider openai --model gpt-5 - Use an alternative provider:
- Anthropic:
--ai-provider anthropic --model claude-sonnet-4-20250514 - Gemini:
--ai-provider gemini --model gemini-2.5-pro - Local:
--ai-provider local --model devstral
- Anthropic: