You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
modelReasoningEfforts: [] makes routed models unusable as subagents (catalog supported_reasoning_levels empty) #883
When a routed model (e.g. cursor/composer-2.5) has modelReasoningEfforts: [] (explicit empty array) in config, ocx sync writes supported_reasoning_levels: [] into the Codex catalog. The Codex app then rejects the model for spawn_agent calls because [agents].default_subagent_reasoning_effort = "max" has no valid level to match against the empty list. The request never reaches the proxy, so the model works in the main thread but is unusable as a subagent.
constmodelEfforts=modelRecordValue(provider.modelReasoningEfforts,modelId);if(modelEfforts!==undefined)return ... [];// [] is explicit "no reasoning" → catalog []
An explicit [] and an omitted key behave differently. [] produces an empty supported_reasoning_levels in the catalog; an omitted key lets the default ladder fill in. A model that genuinely takes no reasoning effort becomes invisible to the subagent roster.
Expected
Either the [] → empty-list semantics should be documented/preserved consistently, or ocx sync should detect models in the subagent roster whose catalog supported_reasoning_levels is empty and either warn or backfill a placeholder level, so spawn_agent with the global default effort is not silently blocked.
Local workaround (verified)
Remove the composer-2.5: [] entry from modelReasoningEfforts. After ocx sync, the catalog fills in the default ladder and subagent routing works (the cursor adapter already drops the effort suffix for composer at the wire, so the upstream is unaffected).
Thanks for taking the time to submit this issue. It was closed automatically because it was not opened with a recognized issue template (Bug report, Feature request, Documentation, or Provider compatibility). Freeform and API-opened issues cannot skip this check.
This issue does not use a recognized issue template.
Please edit the issue to add:
Open a new issue with the Bug report, Feature request, Documentation, or Provider compatibility template.
Or edit this issue so the body uses the template section headings for the kind of report you are filing.
Activity
github-actions commentedon Aug 2, 2026
Automated translation bookkeeping — detected language: unknown.
github-actions commentedon Aug 2, 2026
Issue closed: use an issue template
Thanks for taking the time to submit this issue. It was closed automatically because it was not opened with a recognized issue template (Bug report, Feature request, Documentation, or Provider compatibility). Freeform and API-opened issues cannot skip this check.
Please edit the issue to add:
See the Contributing guide for details.
Once the report passes the automated checks, it will be reopened automatically unless a maintainer has changed its state.
feat(models): per-custom-model reasoning effort in the Models dashboard
fix(models): enforce ladder invariants and constrain Pi effort levels
feat(models): per-custom-model reasoning effort in the Models dashboard
fix(models): enforce ladder invariants and constrain Pi effort levels
feat(models): per-custom-model reasoning effort in the Models dashboard
fix(models): enforce ladder invariants and constrain Pi effort levels