opencode 1.17.13: reasoning mode forced for OpenAI-compatible models, Copilot replay fix
Patch release. Reasoning mode is now forced for OpenAI-compatible reasoning models so reasoning settings apply on custom deployments, and stale GitHub Copilot response item IDs are no longer replayed into follow-up requests. Desktop gets a searchable model picker, session tab hover previews and clearer WSL setup, plus fixes that keep question prompts and page failures scoped instead of dismissing or breaking the app.
PUBLISHED2026-07-01
OBSERVED2026-08-11
AGE1mo
SOURCES1
Core
- Reasoning mode forced for OpenAI-compatible reasoning models so settings apply on custom deployments
- Stale GitHub Copilot response item IDs no longer replayed, avoiding follow-up request failures
Desktop
- Question prompts can be minimized and restored without being dismissed; pending questions stay open while under review
- Session page failures scoped to the affected tab instead of breaking the whole app
- Stale empty and old draft store files cleaned up
- Searchable v2 model picker and model management in the composer
- Session tab hover preview showing project, path, branch and server
- WSL server setup with clearer distro checks and install flows
- Tabs scoped per window so each window restores its own set
- Markdown and message layout alignment improved in the v2 session UI
COMMUNITY
No curated reactions recorded for this event. Facts and takes are kept in separate layers — community context is added by hand, never blended into the record above.