feat(codex): hydrate account models from chatgpt backend

[agustif]

Summary

• hydrate the OpenAI Codex model catalog from the live ChatGPT backend for the logged-in account
• thread async model hydration through model registry, startup, model listing, interactive selection, RPC, and session restore flows
• add tests for live catalog filtering/merging and registry hydration behavior

Testing

• npm --workspace packages/ai test -- openai-codex-models.test.ts
• npm --workspace packages/coding-agent test -- model-registry.test.ts

Notes

• adds PI_OPENAI_CODEX_CLIENT_VERSION override for backend-gated model version negotiation
• did not use full packages/coding-agent build as the merge base had unrelated workspace/type-resolution issues outside this patch

[github-actions]

This PR was auto-closed. Only contributors approved with lgtm can open PRs. Open an issue first.

Maintainers review auto-closed issues daily. Issues that do not meet the quality bar in CONTRIBUTING.md will not be reopened or receive a reply.

If a maintainer replies lgtmi, your future issues will stay open. If a maintainer replies lgtm, your future issues and PRs will stay open.

See CONTRIBUTING.md.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0ce0472b90

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".

[chatgpt-codex-connector]

Preserve dynamically registered providers during hydration

hydrateAvailableModels() rebuilds from this.baseModels, which is only set in loadModels() and does not include models later added via registerProvider(). After any extension/runtime provider registration, a hydration call will overwrite this.models with that stale base list and drop the dynamically registered provider models. This regression is user-visible now that hydration is invoked in startup/session flows, so extension-provided models can disappear from selection and resolution.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Add a timeout to catalog fetch used in model hydration

The account catalog request is awaited without an abort signal or timeout, and hydration is now called on critical paths (startup/session/model resolution). If chatgpt.com/backend-api/codex/models stalls at the TCP/HTTP layer, the CLI can block for an extended network timeout before falling back, despite hydration being treated as best-effort. Adding a bounded timeout (for example via AbortSignal.timeout(...)) avoids long startup hangs.

Useful? React with 👍 / 👎.