This repository was archived by the owner on Oct 22, 2026. It is now read-only.
Repository navigation
feat: route to GPT-6 and Opus 5.5, add doctor --fix - #9
Merged
Merged
Conversation
- Claude complex tier uses Opus 5.5. Codex routes GPT-6 Luna, GPT-6 Sol (balanced and strong profiles), and GPT-6 Astra; policy id default-v0.4. - GPT-5.6 Luna/Terra/Sol and Opus 5 stay in the catalog; codex-terra keeps Terra selectable while Codex offers it. - Jev effort questions describe the capability role instead of naming the model; code still binds each answer to its model. - doctor --fix compares the policy with the installed Codex model list and proposes confirmed, backed-up policy edits; launch failures point to it. - doctor shows the effective lineup; eligibleFamilies derives from routing. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude Code has no local model list, so doctor --fix asks whether the plan can use Fable. Some subscriptions require usage credits for it; if so, the highest Claude tier routes to Opus 5.5, and a later run offers to switch back. Cancelling --fix now reports that nothing was written. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ruban-24
changed the title
ruban-24
marked this pull request as ready for review
September 25, 2026 02:39
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 27658625ea
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Opus 5.5 needs Claude Code 2.1.280 or newer; older releases reject every request to it, including the uncertain fallback. doctor --fix now reads `claude --version` locally and, when it is too old, defaults to Opus 5 for those tiers until the user updates. A later run offers Opus 5.5 again. On a Codex without GPT-6 Luna or Sol, excluding both left GPT-6 Astra as the only tier. doctor --fix now defaults to the previous generation (GPT-5.6 Luna, GPT-5.6 Sol) through a profile override, and the exclude option names the model its tiers would move to. A later run offers the current model again, and defaults to it once Codex drops the old one. Questions are re-planned after each answer, so later options describe the already-changed policy. Tested versions are now Claude Code 2.1.283 and Codex 0.157.1. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A profile override could keep another group on a model Codex dropped, so restoring the overridden routes reported success while the next launch still failed. doctor --fix now applies the restore to a copy and offers it only when no route uses the missing model afterwards and every restored route runs on an available model. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
Route to the current model generation and add
switchboard doctor --fix.default-v0.4): the Claude complex tier uses Opus 5.5. Codex routes GPT-6 Luna, GPT-6 Sol (balanced and strong profiles), and GPT-6 Astra. GPT-5.6 Luna/Terra/Sol and Opus 5 stay in the catalog, andcodex-terrakeeps Terra selectable while Codex offers it.doctor --fixcompares the policy with the installed Codex model list and proposes confirmed, backed-up policy edits. Because Claude Code has no local model list, it asks whether the plan can use Fable; if it needs usage credits, the top Claude tier routes to Opus 5.5 and a later run offers to switch back. Cancelling reports that nothing was written. Launch failures point todoctor --fix.doctor --fixreadsclaude --versionlocally and defaults to Opus 5 for those tiers when Claude Code is too old. For a Codex without GPT-6 Luna or Sol it defaults to GPT-5.6 Luna/Sol through a profile override, instead of excluding them (which would push every Codex conversation to GPT-6 Astra). The exclude option now names the model its tiers would move to. A later run offers the current models again. Questions are re-planned after each answer, and a route restore is offered only when it clears the missing model from every route.doctorshows the effective lineup;eligibleFamiliesderives from routing.This is the base of a two-PR stack; the interactive settings menu builds on it.
Validation: all 290 tests passed;
npm run test:packagepassed type checking, build, isolated global installation, npx execution, initialization, and policy validation.Native adapter checks (combined build of #9, #10, #12, isolated
SWITCHBOARD_HOME):workspace-writeblocked a write outside the workspace.doctor --fix, a complex Claude task ran on Opus 5 and Codex launched and routed to GPT-5.6.🤖 Generated with Claude Code