title agent use not allowed param reasoning_effort: minimal #22796

Activity

1. SsparKluo

   added

   bugSomething isn't workingSomething isn't working

   on Apr 16, 2026
2. github-actions

   assigned

   rekram1-node

   on Apr 16, 2026
3. github-actions

   added

   coreAnything pertaining to core functionality of the application (opencode server stuff)Anything pertaining to core functionality of the application (opencode server stuff)

   on Apr 16, 2026
4. github-actions commented on Apr 16, 2026

   

   github-actionsBot

   on Apr 16, 2026 – with GitHub Actions

   Contributor

   More actions

   This issue might be a duplicate of existing issues. Please check:

   • [BUG] 'minimal' value is not support with the 'gpt-5.1' model #4451: Same error — minimal value for reasoning_effort is not supported by certain OpenAI-compatible models (gpt-5.1 in that case, gpt-5-mini in yours)
5. SsparKluo

   closed this as a duplicateof GitHub Copilot session auto-titling fails when enterprise policy disables gpt-5-mini #17531 on Apr 16, 2026
6. SsparKluo

   reopened this on Apr 16, 2026
7. SsparKluo

   changed the title [-]title agent use not allowed param `reasoning_effort: low`[/-] [+]title agent use not allowed param `reasoning_effort: minimal`[/+] on Apr 16, 2026
8. tiger0425 commented on Apr 16, 2026

   

   tiger0425

   on Apr 16, 2026

   More actions

   I think this should be fixed in the reasoning-effort normalization layer rather than only in the title agent.

   From the error here and the older duplicate-like report in #4451, the broader problem seems to be:

   • OpenCode can send reasoning_effort: minimal
   • but some OpenAI-compatible / GitHub Copilot models (like github-copilot/gpt-5-mini) only accept [low, medium, high]
   • so the title generation request fails and session rename breaks

   Suggested fix:

   1. Normalize reasoning_effort against the target model's supported values before building the request
   2. If minimal is requested but the model only supports [low, medium, high], coerce minimal -> low
   3. If the model exposes no compatible value at all, omit reasoning_effort instead of hard-failing a non-critical path like title generation

   Why this seems better than a title-agent-only patch:

   • title agent use not allowed param reasoning_effort: minimal #22796 is title-agent-visible, but [BUG] 'minimal' value is not support with the 'gpt-5.1' model #4451 suggests this is part of a broader reasoning-effort compatibility problem
   • a central normalization/fallback would prevent the same class of bug from showing up in other agent paths or future models

   Practical fallback policy could be something like:

   • support contains minimal -> send minimal
   • support contains low but not minimal -> send low
   • otherwise -> omit the field

   At minimum, applying that fallback for the title agent / small_model path would unblock gpt-5-mini users immediately.
9. rekram1-node commented on Apr 28, 2026

   

   rekram1-node

   on Apr 28, 2026

   Collaborator

   More actions

     if (model.api.id.includes("5.") || model.api.id.includes("5-mini")) {
       return { store: false, reasoningEffort: "low" }
     }
   

   It will set it to low
10. rekram1-node

    closed this as completedon Apr 28, 2026