Skip to content

chore: merge upstream main into the fork - #169

Merged
ScriptedAlchemy merged 386 commits into
mainfrom
t3/merge-upstream-main-20261119
Nov 20, 2026
Merged

ScriptedAlchemy merged 386 commits into
mainfrom
t3/merge-upstream-main-20261119

Conversation

@ScriptedAlchemy

@ScriptedAlchemy ScriptedAlchemy commented Nov 20, 2026 •

Copy link
Copy Markdown
Owner

Merge 383 upstream commits through pingdotgg/t3code@490a0d687b into the fork. Keep fork-specific provider model validation, LAN authentication, desktop attach, and cached PR reads alongside upstream's model discovery, relay, and CI changes. Preserve fork migration IDs 1–121 and reconcile upstream histories at 122–125.

The review found that delegation by driver rejected a valid workspace-only model until something else warmed the catalog. On a miss, delegation now scans eligible instances of that driver in the project and retries once. The patch also routes previously unmarked web and mobile copy through i18n, verifies SSH runtime compatibility before reusing an installation, and updates stale browser mocks and orchestration replay fixtures. Shorter instructions and tool descriptions bring all three agent-context measurements below the fork's main baselines without a budget override.

Validation at 8f1e08f1ca:

  • The fork's changed-file gate against origin/main passed, including formatting, lint, knip, affected typechecks (43 packages), and React Compiler checks. As in fork CI, tests were run separately.
  • Web shard 3/4: 1,526 passed. Server shard 2/8: 1,264 passed, 3 skipped. SSH package: 97 passed, 5 skipped. Cold-catalog delegation and the four MCP transcript cases passed locally.
  • All three GT source catalogs match. All nine web and mobile translations cover the new keys and preserve ICU arguments. The iOS dev client built and installed, paired to an isolated seeded server, and rendered Chats and Usage → Limits.
  • Model-facing eval request construction and deterministic tests ran; no paid live model evaluation was run.

Prepared with GPT-6.2 Sol and GPT-6.2 Astra via Codex in T3 Code.

juliusmarminge and others added 30 commits November 18, 2026 19:09
…browser bar (pingdotgg#38045)

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…roid (pingdotgg#38663)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…tgg#38746)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
…otgg#38831)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
…uth status (pingdotgg#38775)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
pingdotgg#38636)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Julius Marminge <julius0216@outlook.com>
…and runs, and Stop ends it (pingdotgg#38916)

agy ends its prompt while a command it backgrounded still runs, and replies by
itself when the command finishes. T3 Code ended the run at the prompt, so the
thread looked idle and Stop was not offered.

The run now stays open until the command ends. A daemon does not hold it. Stop
sends session/cancel, which ends the command in agy 1.5.11. A message sent
while the command runs does not, so the command keeps running. A held run
fails when the agent process exits.

The Servers panel no longer lists an agent CLI's own ports.

Takes over pingdotgg#35073 and pingdotgg#36585.

Co-authored-by: Julius Marminge <julius0216@outlook.com>
Co-authored-by: saphid <4596216+saphid@users.noreply.github.com>
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
…g#38873)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
…older page (pingdotgg#38657)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
pingdotgg#38856)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
…ad (pingdotgg#38890)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
mwolson and others added 23 commits November 19, 2026 13:42
…pload fails (pingdotgg#38395)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
…rs (pingdotgg#33568)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
…dotgg#39395)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
…el (pingdotgg#39415)

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
…elegating (pingdotgg#39413)

Co-authored-by: Simone <185146821+Lucenx9@users.noreply.github.com>
Co-authored-by: Julius Marminge <julius0216@outlook.com>
…ect loop for good (pingdotgg#39420)

Co-authored-by: Julius Marminge <julius0216@outlook.com>
@changeset-bot

changeset-bot Bot commented Nov 20, 2026 •

Copy link
Copy Markdown

⚠️ No Changeset found

Latest commit: 49d96a8

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

@ScriptedAlchemy ScriptedAlchemy left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I would keep this draft at db2f0b044b for two reasons:

  1. Driver-only delegation rejects a valid workspace model on a cold catalog. In apps/server/src/mcp/OrchestratorMcpService.ts:1584, a target specifying driverKind without providerInstanceId returns before the refresh path. I verified this with the existing OpenCode folder-model fixture, starting with no cached workspace snapshot: delegate_task for uni/medium-local by driver returned model_unavailable and never scanned the workspace. The same request by instance ID scanned and dispatched successfully. On a model miss, scan eligible instances of the requested driver and retry resolution once; cover this path without calling capabilities first.
  2. The fork's PR validation command still fails. check:changed reported 190 no-untranslated-copy errors at this head, and a fresh focused lint run reproduced the errors at apps/web/src/components/settings/ProviderSettingsPanel.tsx:1698. Catalog parity passes, but it does not check that source strings use i18n. Route the flagged strings through i18n and rerun the changed-file check before enabling auto-merge.

@ScriptedAlchemy

Copy link
Copy Markdown
Owner Author

Fixed the four failing hosted jobs from 8f1e08f in 49d96a80a1. The mobile model-discovery mock now returns an Effect; the group adapter fixture buffers events until its consumer subscribes; Kiro and Muse replay legacy instruction wording without losing the recorded model or effort; and MCP description checks match the shipped text while retaining permission assertions. Also fixed macOS-only temporary-path and mDNS fixtures. No production code, retries, timeouts, or skip settings changed in this commit.

Local validation: full mobile suite (3,799 passed), server shards 4/8 (1,732 passed, 3 skipped), 5/8 (1,153 passed, 2 skipped), 8/8 (1,138 passed, 5 skipped), and six provider replay suites (77 passed). The changed-file gate passed, including four affected typechecks and 150 tests. Keeping the PR in draft until hosted checks pass on this head.

@ScriptedAlchemy
ScriptedAlchemy marked this pull request as ready for review November 20, 2026 04:29
@ScriptedAlchemy
ScriptedAlchemy merged commit 2110fa8 into main Nov 20, 2026
20 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.