[bwachman]

Summary

• A bounded model catalog takes its request and response parser as inputs, so a catalog that authenticates or reads a different response shape reuses the snapshot, in-flight dedupe and bounded read. An OAuth projection chooses its models per profile. The live ChatGPT catalog builds on both.
• When a live catalog keeps its fallback, the log says why: an empty or malformed catalog warns, and a connection that drops or times out logs at debug.
• After an OpenCode Go or Zen prefetch settles, the rebuilt catalog keeps the live model selection, so a model switched during the rebuild gets its own context window.

Verification

• bun run check and bun run test pass.

Part of CL-11727

[linear-code]

CL-11727

[corbits-triage]

Triage: Pass

No changes needed from you. A maintainer will review this pull request.

Notes:

• added code matches a code pattern: src/tui/runner/catalog-refresh.ts:27-57 promise-chain: Promise chain; use async/await with try/catch., src/tui/runner/catalog-refresh.ts:27-56 promise-chain: Promise chain; use async/await with try/catch., src/tui/runner/settings.ts:289-387 iife: Immediately invoked function; name the function., src/tui/runner/settings.ts:289-387 promise-chain: Promise chain; use async/await with try/catch.

[bwachman]

Stack merge order: #1580 → #1581 → #1579. This repo keeps merged branches, so GitHub does not retarget the next PR on its own:

1. After fix(auth): drop the models retired for ChatGPT sign-in #1580 merges: gh pr edit 1581 --base main
2. After refactor(provider): prepare live model catalogs for per-account lists #1581 merges: gh pr edit 1579 --base main

[TheGreatAxios]

Review · Approve

This change gets model lists ready for one list per signed-in account, instead of a single shared list.

Fetching a provider's models and reading the response are now arguments, so a later ChatGPT list can reuse the same cache instead of copying it. Each signed-in profile picks its own models. If a fetch fails, the log says whether the catalog was empty or the connection dropped. Refreshing an OpenCode Go or Zen list no longer forgets the model you just switched to, so that model keeps the right context window.

Checked against the parent branch (#1580) only. No problems in this slice.

[TheGreatAxios]

Review · Approve

A bounded catalog takes its request and parser as inputs, an OAuth projection chooses models per profile, and a settled prefetch keeps the live model selection.

Findings

No findings.

[TheGreatAxios]

Review · Comment

Planner: the injected request and parser, the per-profile models callback, and the shared prefetch refresh fit a later authenticated, per-account catalog without widening those seams.

Findings

No findings.