[davisonio]

App-owned subagents launched without an explicit effort hid that effort in their hover card. Codex also forced medium with T3 MCP tools, so simply displaying the model default could misreport the run.

The shared Lineage/timeline tooltip now uses the matching model's default when no effort was saved. Explicit effort wins; provider-native missing effort stays unknown. Codex resolves the same provider-instance catalog default for each turn, sends it in the request and collaboration settings, and uses it in runtime context. This also resets a previous explicit effort on resumed threads. Both local CLI and managed ChatGPT providers use their existing in-memory snapshot. Bare custom models keep borrowed effort choices but no longer inherit another model’s default; authored custom capabilities remain intact.

Thanks @davisonio for the original reproduction, tooltip implementation and browser coverage in #20981. The takeover preserves those commits, merges current main, and reduces duplicate test scenarios. The adapter correction is included so the displayed default agrees with execution.

Before: main	After: this branch

Captured in headless Linux Chromium at 1280×800 from exact main and candidate 7a9d9e5993 using matching synthetic stopped subagents in isolated worktree state; cropped identically. No provider was launched for these screenshots. The fixture also confirms explicit high and no invented effort for a native subagent.

Validation: 209 tests pass across the Codex catalog, adapter, developer instructions and driver suites; all 24 Codex orchestrator replay cases pass. All 18 Lineage browser tests passed before the catalog-only repair; the primary repeated the integrated tooltip checks on this candidate. Regression mutation checks fail without the tooltip default, native guard, explicit precedence, or adapter catalog wiring. Scoped lint, formatting, type-aware checks and server typecheck pass. Across the three changed test suites, declarations are 189→190 (tooltip 18→18, adapter 169→169, catalog 2→3); tooltip scenarios reduced 6→4. The added catalog regression failed before the custom-default repair and passes after it. Two existing replay expectations now match omitted effort rather than forced medium.

Independent review: round 1 found an inherited custom-model default and two stale replay expectations; both are fixed and verified. Fresh Opus and Sol whole-PR round 2 audits found no blockers on this revision; Opus independently repeated all 209 focused tests and 24 replays, and Sol repeated the 172 catalog/adapter tests.

The bare-custom model correction is proven through the real provider-probe/dispatch test, rather than rendered-client evidence.

Not verified: live Codex inference, other Codex versions, macOS, Windows, iOS, Android, or the Electron desktop shell. Web/desktop share this tooltip; the server adapter correction applies to all clients and agent/scheduler turns. Native mobile has a separate UI. The tooltip fallback uses the current catalog, so later catalog changes can alter the label on historical tasks.

Takeover: GPT-6-Astra (tests), GPT-6.1-Sol (implementation and coordination), Claude Opus 5.5 and GPT-6-Sol (review), through Codex and Claude harnesses in T3 Code.

[macroscopeapp]

Approvability

Verdict: Approved at 2d48dbf

Macroscope's review found this PR approvable — This is a small, localized hover-card display fix that reads existing model defaults without changing model execution or product configuration. It preserves existing explicit selections and adds targeted browser coverage for the new fallback behavior.

^{You can add or adjust custom eligibility rules. Learn more.}

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

• Configuration used: Path: .coderabbit.config.ts
• Review profile: CHILL
• Plan: Advanced
• Run ID: 0b4500bb-5466-43d5-8f37-d2e506ab94ff

📥 Commits

Reviewing files that changed from the base of the PR and between 2d48dbf and 72dba0c.

📒 Files selected for processing: 2

• apps/web/src/components/chat/SubagentTooltipContent.tsx
• apps/web/src/components/chat/ThreadRelationshipsControl.agents.browser.test.tsx

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 8 remain after this review.

📝 Walkthrough

The model tooltip now uses a matching provider model’s selectable option value when no saved string option provides reasoning effort. Browser tests cover explicit values, descriptor defaults, current values, missing descriptors, and provider-native agents.

Changes: Reasoning effort display

Layer / File(s)	Summary
Tooltip fallback and test coverage apps/web/src/components/chat/SubagentTooltipContent.tsx, apps/web/src/components/chat/ThreadRelationshipsControl.agents.browser.test.tsx	The tooltip checks saved string options first. For app-owned tasks, it then checks matching selectable descriptors and uses the first string current value. Tests cover explicit selections, defaults, current values, missing descriptors, and provider-native agents.

Priority: ⬇️ Low

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Bug fix

Suggested reviewers: bil0000, juliusmarminge

Merge Risk: ⚪ Minimal · up to 72dba

Saved effort values remain authoritative, and eligible app-owned tasks can display a string descriptor value when no saved effort exists. Native and unmatched cases remain unchanged; no material product risk is apparent.

Pre-merge checks: 4

Check name	Status	Explanation
Linked Issues check		Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check		Check skipped because no linked issues were found for this pull request.
Title check		The title is concise, specific, and describes the default effort label fix. The Codex turns reference also matches the broader change described in the pull request.
Description check		The description covers the problem, implementation, scope, verification results, screenshots, and unverified environments. It is mostly complete, although explicit maintainer approval or a clearer exe…

Full details: Description check

Explanation

The description covers the problem, implementation, scope, verification results, screenshots, and unverified environments. It is mostly complete, although explicit maintainer approval or a clearer exemption rationale could be stated for the scope requirement.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}

[juliusmarminge]

Taking over this one as maintainer for batch scheduler-tasks-02. I will preserve your commits, rebuild the minimum fix against current main, and verify that the displayed default matches Codex execution before publishing. Thanks @davisonio for the focused reproduction and tooltip fix.

[juliusmarminge]

Thanks @davisonio for the reproduction, fix, and regression coverage. The maintainer compared both approaches and landed #20991 at ff9f41b65f for this problem. Your approach also displayed catalog-default effort in app-owned subagent tooltips and resolved it explicitly for Codex turns, while the selected change lets Codex resolve unset effort at runtime and preserves T3’s Astra policy. If this covered anything the landed change does not, please bring it forward as a follow-up on main.