[flamboh]

Fix for the redesign stack, targets layer 4 (#28842).

Note

🤖 GPT-6.1 Sol on behalf of Oliver

Layer 4 restored provider-update launch toasts that main removed in #23233. Remove the launch coordinator and its root mount so the redesign matches main: updates remain available through the sidebar and Settings without opening a toast on launch.

Keep the shared update helpers and their tests. Their redesign changes format failures from user-triggered updates, and still serve the sidebar and settings actions. Guided setup and dialog changes are untouched.

Validation

• vp run check:changed origin/maria/redesign/4-settings-setup: passed formatting, lint, both knip passes, affected typechecks, and the React compiler check.
• Verified in an isolated local dev server with no paid turns: the server reported Codex 0.166.0 → 0.167.0; after launch there is no toast, the sidebar still has “Update 1 provider,” and provider settings still offers “Update all.”
• Both captures temporarily disabled development Strict Mode effect replay to match production behavior. All temporary patches were reverted. The dev server was stopped by its tracked process group.
• Applies to web and the desktop web shell, independent of connection mode or provider. Mobile, contracts, server behavior, and update actions are unchanged.

Before / after

What to look for: before, an update card opens at the top right on launch; after, that area stays clear while the sidebar update control remains at the bottom left.

GPT-6.1 Sol · Codex harness

[macroscopeapp]

Approvability

Verdict: Needs Human Approval

Macroscope's review found this PR requires human approval — Provider-update launch toasts are removed while sidebar and Settings update controls remain. This changes the default notification experience; a human must confirm that updates should no longer be proactively announced on launch.

No code changes detected at 76c135b. Prior analysis still applies.

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

[maria-rcks]

Note

Written by claude-opus-5-5 on behalf of Maria

Merged the rebased maria/redesign/4-settings-setup into this branch (6c482b7, merge commit, no rebase). The conflicts in __root.tsx and WorktreeSetupCard.tsx came from the layer-4 rebase; both now match the base, so the diff against layer 4 is still only the 84-line removal of ProviderUpdateLaunchNotification and its root mount. I merged the base branch instead of main because main is 137 commits ahead of layer 4 and would pull unrelated work into this stacked PR.

vp run check:changed against layer 4 passed (format, lint, knip, typecheck, React compiler), and so did ProviderUpdateLaunchNotification.logic.test.ts (9 tests). CI is green on the new head.

[maria-rcks]

Note

Written by gpt-6.1-sol on behalf of Maria

Verified in a real dev environment

Verified unchanged head 5220641268fc4870a98fcde0c1b920024e34fe8d on a disposable Linux VM with repo Playwright and /usr/bin/google-chrome, one background dev server at a time, isolated app state, and real Codex gpt-6.1-sol turns through the local proxy. Projects and threads were created through the app. No mocks, fixtures, demo data, or Strict Mode changes.

• github/main (d1909d68f8) already has the intended behavior: no launch toast. It completed a real Codex turn. The regression reproduces on this PR's redesign base 3851bcfcbf: Codex 0.164.0 with the real 0.167.0 advisory opens a persistent update toast.
• On the exact PR head, the same thread, response, binary, and advisory produce no launch/reload toast. The sidebar still offers “Update 1 provider”; clicking it opens the machine/provider confirmation.
• Settings still offers “Update all.” Clicking it actually upgraded Codex 0.164.0 → 0.167.0 and displayed the user-triggered success notification. Both provider settings and the CLI confirmed 0.167.0.
• A new post-update Codex gpt-6.1-sol thread ran both commands below and completed its response. The proxy/provider recovered after five retries; the turn completed in 51 seconds.
• Blacksmith passed the existing 9 shared update-logic tests and vp run check:changed 3851bcfcbffc66b14337030dcd0799c2ee5db523 (format, lint, both knip passes, affected typechecks, React compiler, scoped checks). The current head also has 30 successful CI contexts, none failing or pending. No source changes were needed.

before: codex-cli 0.164.0
Settings > Providers > Update all
 after: codex-cli 0.167.0
new thread, Codex gpt-6.1-sol:
  pwd                         -> /home/user/t3code
  git rev-parse --show-toplevel -> /home/user/t3code
  response: updated provider verification ok.

Main baseline:

Before, redesign base:

After, exact PR head with matching conversation and update advisory:

Recording: sidebar confirmation, real Settings update and success, new Codex turn, and updated provider settings. Idle time between submitting the turn and its completion is cut. No speed claim is intended.

Not verified

Native Electron, iOS/Android, macOS/Windows, remote/relay/tunnel connections, and other provider update executions were not exercised on this Linux web client. The sidebar confirmation was exercised; the actual install was initiated from Settings. Main cannot reproduce this regression because it already removed launch toasts; the redesign base provides the before reproduction.

GPT-6.1 Sol · Codex harness