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
Keep declared default effort visible for every chat agent - #36455
When you haven't picked an effort level in a chat, the effort control should still tell you what the model will use by default. Only Codex chats did this. For every other agent the effort control showed nothing once the chat was sitting idle after a reply, even when the agent had told Orca what its default is.
Two agents also told Orca the wrong default:
Grok reports its default effort, but Orca ignored that value.
OpenCode doesn't report a default at all. Orca guessed by taking the first effort option in the list, which isn't what OpenCode actually uses.
What this changes
Every agent, not just Codex, now shows its model's default effort in the effort control when you haven't picked one. It appears in the same "default" style Codex already uses.
Grok shows the default effort it actually reports.
OpenCode shows no default, rather than a guess, until it reports one.
Nothing you send changes. Every chat starts with, and sends with its first message, exactly the effort it did before. Only what the control displays is different.
How it works
Each model in an agent's model list can now carry one extra, optional value: the default effort that agent says it uses. Orca reads that value only to label the effort control. It is kept separate from the existing value Orca uses to decide what effort to send when it launches a chat or sends a first message. That separation is what keeps this display-only: an earlier version reused the existing value and, for Grok, changed the effort a new chat started with. Review caught and fixed that.
Anything you picked, saved, or that the agent reports for the running chat still wins over this default label.
Older versions
Nothing is saved to disk, and no new feature flag or minimum version is needed. An older host doesn't send the new value, so on an older host non-Codex agents simply keep showing no default, as they do today. Codex behaves as before everywhere.
Testing
Tests feed in Grok's and OpenCode's real recorded output and check what the control shows on desktop and phone. They also check that what every agent sends at launch and with its first message is identical to main. All CI checks pass.
It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 3 remain after this review.
Walkthrough
Catalog adapters now report declared reasoning effort separately from menu defaults. The native catalog merges declarations from discovery, live listings, and configured choices, while excluding transient declarations from persisted listings and listing keys. Session option views display declared effort when it matches an available choice. Tests and model fixtures cover these behaviors across catalog sources and session consumers.
Merge Risk: Moderate · up to 5aec0
A session’s selected effort can appear as the model’s declared default. Resolve that distinction before merging; the previously failing mobile assertion has been fixed.
Pre-merge checks: 3 · 2
Failed checks: 2 warnings
Check name
Status
Explanation
Resolution
Docstring Coverage
Docstring coverage is 41.18% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documenta…
Write docstrings for the functions missing them to satisfy the coverage threshold.
Description check
The description explains the problem, the change, the mechanism, older versions, and testing. It omits required template sections and evidence. It does not include ELI5, a valid issue link, Visual Pro…
Update the PR description. Add an ELI5 section. Add a valid issue reference in the required format, such as "Fixes #"; "tracked in STA-13018" alone is insufficient. Add BEFORE and AFTER visual captures, preferably video, for each af…
Passed checks: 3 passed
Check name
Status
Explanation
Title check
The title clearly and concisely describes the main change: preserving declared default effort visibility for every chat agent.
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.
Full details: Docstring Coverage
Explanation
Docstring coverage is 41.18% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentation-only edits and deletions within surviving functions are included; unchanged diff context and fully deleted functions are excluded. Analyzed 17 functions across 33 files.
Resolution
Write docstrings for the functions missing them to satisfy the coverage threshold.
Full details: Description check
Explanation
The description explains the problem, the change, the mechanism, older versions, and testing. It omits required template sections and evidence. It does not include ELI5, a valid issue link, Visual Proof, AI Disclosure, Review, Agent skill upstream boundary, or the completed checklist. Testing does not identify platforms or confirm manual testing. The change affects displayed effort defaults on desktop and phone, so visual proof is required.
Resolution
Update the PR description. Add an ELI5 section. Add a valid issue reference in the required format, such as "Fixes #<number>"; "tracked in STA-13018" alone is insufficient. Add BEFORE and AFTER visual captures, preferably video, for each affected agent on desktop and phone. If visual proof is not applicable, write "N/A" and explain why. Identify the tested platforms: macOS, Linux, Windows, and SSH. If manual testing was completed, tick the manual-testing checkbox and list the steps. Complete AI Disclosure by naming the AI model used or confirming that the author is a StablyAI team member. Complete the Review and Agent skill upstream boundary checkboxes. Complete the checklist and confirm that pnpm lint, typecheck, test, and build pass locally, or state that CI covers these checks.
Finishing Touches: 💡 2
📝 Generate docstrings: 💡
Commit to this branch
Create a new PR
🛠️ Fix failing CI checks: 💡
Commit to this branch
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 3 remain after this review.
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 0 remain after this review.
Use different values for current and declared effort.
The fixture sets both values to high. A regression that copies the session's current effort into declaredDefaultEffort would pass. Keep the current effort at high, but declare low in the model metadata and assert declaredDefaultEffort: 'low'.
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 4 remain after this review.
Native chats could lose their effort label after a reply even when the model declared a default. The original display restriction was agent-specific, while some catalog values also doubled as launch defaults; correcting those values directly could change a later chat’s first request.
Desktop and phone now show a source-backed declared default when no current effort exists. Grok uses its reported declaration; OpenCode does not infer one from variant order. Declaration evidence is separate from action defaults, and explicit unknown evidence defeats an older claim. Current/saved choices and every launch/first-send input stay unchanged. Older hosts and restored catalogs supply no new declaration proof until fresh evidence arrives; existing Codex behavior is retained.
The full review fixed the launch-policy leak, old-host provenance gap, actual live Grok result with missing membership IDs, and the typed desktop/phone test boundaries. No runtime finding or deferred in-scope bug remains. The final main sync was 9025d69; the later 5aec069 commit combines only duplicate test imports.
Validation: CI is green on 5aec0691be8f9f9af24843c4ef73cd0caf298d84: PR checks, Mobile, and WSL. Fedora needed one failed-only retry after PowerShell initialization failed; the retry passed and successful jobs were preserved. Focused recorded-source→real-host→desktop/phone contracts passed; pinned-main replay preserved High/High/Low first selections and old/new peer inputs for eleven agents. Source/freshness/persistence and N/10N operation contracts passed. Actual Mobile CI ran the phone hook and its typecheck ratchet. All changed flows are tests rows; no layout, interaction, lifecycle, process or login change required a running-app round or screenshot. No real-agent run is claimed.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
What was wrong
When you haven't picked an effort level in a chat, the effort control should still tell you what the model will use by default. Only Codex chats did this. For every other agent the effort control showed nothing once the chat was sitting idle after a reply, even when the agent had told Orca what its default is.
Two agents also told Orca the wrong default:
What this changes
How it works
Each model in an agent's model list can now carry one extra, optional value: the default effort that agent says it uses. Orca reads that value only to label the effort control. It is kept separate from the existing value Orca uses to decide what effort to send when it launches a chat or sends a first message. That separation is what keeps this display-only: an earlier version reused the existing value and, for Grok, changed the effort a new chat started with. Review caught and fixed that.
Anything you picked, saved, or that the agent reports for the running chat still wins over this default label.
Older versions
Nothing is saved to disk, and no new feature flag or minimum version is needed. An older host doesn't send the new value, so on an older host non-Codex agents simply keep showing no default, as they do today. Codex behaves as before everywhere.
Testing
Tests feed in Grok's and OpenCode's real recorded output and check what the control shows on desktop and phone. They also check that what every agent sends at launch and with its first message is identical to main. All CI checks pass.
tracked in STA-13018