GitAssistant 2.2.0 is out — safer Commit Insights and better AI commit error handling
Hey everyone,
GitAssistant 2.2.0 is now available.
This release mainly focuses on making Commit Insights safer to use and improving the AI commit generation experience when something goes wrong.
What’s new in 2.2.0
Commit Insights now supports on-demand analysis. Instead of automatically running a full Git analysis when a project opens, GitAssistant will ask for repository size confirmation before starting potentially expensive analysis runs.
This should make the feature much safer for large repositories.
I also added better error classification for AI commit generation. GitAssistant can now distinguish between different failure types, including:
- configuration issues
- connectivity problems
- model errors
- thinking / reasoning-related issues
- prompt-size problems
- empty responses
- unknown failures
The goal is to make failures easier to understand and recover from, instead of just showing a generic “AI failed” message.
Fixed
- Commit Insights no longer runs full Git analysis automatically when a project opens.
- AI provider verification and model refresh UI updates are now bound to the settings dialog lifecycle.
- Empty Git commit messages and “API returned an empty response.” issues with reasoning models were fixed in 2.1.0, and 2.2.0 builds on that work.
GitAssistant 2.2.0 is mostly a stability and UX release, but I think it makes the plugin feel much safer and more predictable in real-world projects.
Feedback is very welcome, especially from anyone using large repositories or OpenAI-compatible reasoning models.