v1.17.9
OpenCodeSummary
This release focuses on stability and configuration adherence, ensuring agent step limits are honored and custom headers are correctly passed to Copilot requests. It also introduces new GLM-5.2 thinking variants (`high` and `max`) for more nuanced generation control.
Bug Fixes
- Fixed agent runs failing mid-run by ensuring a final text response is generated when configured step limits are reached.
- Fixed detection of the Devstral model when provider IDs used different casing.
- Fixed issue where configured custom headers were not being passed to Copilot model requests.
- Fixed TUI to only show the background subagent shortcut when the server actually supports it.
- Fixed command documentation within the customize-opencode skill.
Improvements
- Added `high` and `max` thinking variants for GLM-5.2 across supported providers.
- Improved prompt caching effectiveness by stopping the wrapping of follow-up user messages in a steering reminder.