Skip to content

test(server): ACP adapter tests no longer race the prompt settle - #14876

Closed
tris203 wants to merge 1 commit into
pingdotgg:mainfrom
tris203:fix/acp-adapter-test-settle-race
Closed

tris203 wants to merge 1 commit into
pingdotgg:mainfrom
tris203:fix/acp-adapter-test-settle-race

Conversation

@tris203

@tris203 tris203 commented Oct 2, 2026 •

Copy link
Copy Markdown
Contributor

Problem

Eleven tests in AcpAdapterV2.test.ts guess at when the adapter has settled a prompt, and ten of them can lose that race and fail intermittently in the Test Server 5 CI job, which then fails Check on PRs that do not touch this code.

Each test waits for the raw stopReason line on the wire, runs Effect.yieldNow twice, and assumes the adapter has marked the prompt as settled. The adapter only sets promptSettled several async hops later, inside the prompt completion callback. If Stop wins that race, interruptTurn takes the "send cancel, wait up to 10 seconds" path, and the test's own TestClock.adjust("10 seconds") fires that timeout immediately. The test fails in about 400ms with:

ProviderAdapterInterruptError: Failed to interrupt acp-test provider turn …
Caused by: ProviderAdapterProtocolError: … did not acknowledge cancellation before the interrupt timeout.

Production is not affected: on a real clock the late completion callback finalizes the turn as interrupted within milliseconds.

Change

The eleven tests now wait on the existing afterPromptSettledWithBackgroundWork test hook instead of the wire line plus two yields. Three other tests in the same file already wait this way. No production code changes.

Scope and approval

No prior issue. This is a focused reliability fix for an obvious test bug: a wait that guesses at scheduling, in a file where the deterministic signal already exists. It is test-only, touches one file, and changes no product behavior. All eleven edits are the same fix for the same guessed wait.

Verification

How the problem was established

Two unrelated PRs failed Test Server 5 on 2026-10-02 with the identical error, on two different tests:

  • run 37055609990, job 110999442239 (t3code/disabled-instance-badge): direct Stop on a deferred subagent hold terminalizes the subagent and does not carry it forward, failed in 411ms
  • run 37054999299, job 110997409196 (git-actions-menu-trigger-disabled-when-empty): user attach flushes a deferred terminal but leaves wake traffic for its continuation

Reproduction

Unmodified, the test passes locally (3 of 3 runs), so I widened the window instead. Temporarily add a 50ms real-time delay before the prompt completion callback in AcpAdapterV2.ts, just above the Effect.flatMap((result) => runRuntimeCallbackAtGeneration(promptGeneration, … in startTurn (not part of this PR):

Effect.tap(() => Effect.promise(() => new Promise((r) => setTimeout(r, 50)))),

Then, from apps/server on Linux:

vp test run src/orchestration-v2/Adapters/AcpAdapterV2.test.ts
Delay injected No delay
Before this change (cc1e634bfa) 10 failed, 104 passed not run for the whole file
After this change 114 passed 114 passed, 3 consecutive runs

The ten failures before the change include both tests that failed in CI. Nine fail in about 400ms with the exact CI error; projects a child terminal in the completed-root finalization window exactly once instead hangs to the 120s test timeout, because its TestClock.adjust("1 second") runs before the timer it is meant to fire is registered.

The eleventh changed test, preserveRuntimeOnSettledInterrupt keeps the process alive and carries subagents through a settled steering interrupt, used the same wait but passed with the delay injected, because that flavor's interrupt also accepts the wire-settled signal. It is converted so it no longer depends on scheduling either.

vp fmt --check on the file is clean. vp lint on the file reports three warnings, all on lines this PR does not touch.

Not checked

  • I never saw the failure locally without the injected delay; the CI logs are the only natural occurrence.
  • I did not run the rest of the server suite or any repo-wide checks.
  • Two tests that also wait on the stopReason wire line are unchanged because they passed with the delay injected: Direct Stop projects an interrupt-deferred terminal exactly once and projects completed-root carryover eagerly and drain cannot resurrect it.

Claude Opus 5.5, Claude Code harness in T3 Code.

@github-actions github-actions Bot added vouch:trusted Author is trusted by repo permissions or the VOUCHED list. size:L 100-499 changed lines (additions + deletions). labels Oct 2, 2026
macroscopeapp[bot]
macroscopeapp Bot previously approved these changes Oct 2, 2026
@macroscopeapp

macroscopeapp Bot commented Oct 2, 2026 •

Copy link
Copy Markdown
Contributor

Approvability

Verdict: Approved at db03ae5

Macroscope's review found this PR approvable — The PR is confined to an ignored test file and replaces scheduling guesses with deterministic prompt-settlement synchronization. Its impact is limited to test/CI reliability, with no production behavior, product defaults, or static-analysis configuration changes.

Notes:

  • Diff unchanged. Approvability was decided on eligibility alone.

You can add or adjust custom eligibility rules. Learn more.

@juliusmarminge juliusmarminge added the macroscope-review Opt PRs made by unvouched contributors in for Macroscope review. Vouched contributors auto-reviews label Oct 2, 2026 — with ChatGPT Codex Connector
@macroscopeapp
macroscopeapp Bot dismissed their stale review October 2, 2026 20:18

Dismissing prior approval to re-evaluate db03ae5

@coderabbitai

coderabbitai Bot commented Oct 2, 2026 •

Copy link
Copy Markdown
Contributor

Review in Change Stack →

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

Important

Review skipped

Review was skipped as selected files did not have any reviewable changes.

⚙️ Run configuration
  • Configuration used: Repository: pingdotgg/t3code/.coderabbit.yaml
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: b5fd4351-cbb1-4b55-aa06-89b55f36bf31
📥 Commits

Reviewing files that changed from the base of the PR and between db03ae5 and e0c3f55.

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

  • 🔍 Trigger review

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info
⚙️ Run configuration

Configuration used: Repository: pingdotgg/t3code/.coderabbit.yaml

Review profile: CHILL

Plan: Advanced

Run ID: a45ce577-d46e-484d-97bb-17c0b58500f8

📥 Commits

Reviewing files that changed from the base of the PR and between cc1e634 and db03ae5.

📒 Files selected for processing (1)
  • apps/server/src/orchestration-v2/Adapters/AcpAdapterV2.test.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 6 remain after this review.


📝 Walkthrough

Walkthrough

Eleven background-work interruption tests now synchronize through afterPromptSettledWithBackgroundWork. Each test signals and awaits a Deferred before interrupting the turn, replacing protocol-event polling and yields.

Changes

ACP interruption test synchronization

Layer / File(s) Summary
Synchronize interruptions on the prompt-settled hook
apps/server/src/orchestration-v2/Adapters/AcpAdapterV2.test.ts
Eleven tests signal and await a Deferred through afterPromptSettledWithBackgroundWork before interrupting. This replaces polling for protocol events and yielding. The tested scenarios remain unchanged.

Priority: ⬇️ Low

Estimated code review effort: 2 (Simple) | ~8 minutes

Change: Other

Suggested reviewers: juliusmarminge

Merge Risk: ⚪ Minimal · up to db03a

The interruption tests now wait for prompt settlement without an identified blocking path. No actionable merge risk remains beyond normal test checks.

Architecture Summary

Architecture risk: 🔵 Low · up to db03a

The change affects 1 system.

Changed systems: apps/server

Architecture concerns
No architecture-level concerns identified.

Review details

Systems and components

  • observed — apps/server (service) was modified; 1 changed file maps to changed impact.

Before / after behavior

  • observed — Modified behavior in apps/server/src/orchestration-v2/Adapters/AcpAdapterV2.test.ts: The test now signals and awaits a Deferred through afterPromptSettledWithBackgroundWork, replacing protocol-event polling and two yields as its synchronization point.
  • observed — Modified behavior in apps/server/src/orchestration-v2/Adapters/AcpAdapterV2.test.ts: The test now waits for the prompt-settled hook instead of polling for an incoming stopReason and yielding twice before interrupting.
  • observed — Modified behavior in apps/server/src/orchestration-v2/Adapters/AcpAdapterV2.test.ts: The carryover test adds a Deferred signaled by afterPromptSettledWithBackgroundWork to coordinate its interrupt.
  • observed — Modified behavior in apps/server/src/orchestration-v2/Adapters/AcpAdapterV2.test.ts: The test awaits the prompt-settled hook instead of polling for stopReason and yielding twice.
🚥 Pre-merge checks | ✅ 5
✅ Passed checks (5 passed)
Check name Status Explanation
Docstring Coverage ✅ Passed No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0…
Linked Issues check ✅ Passed Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check ✅ Passed Check skipped because no linked issues were found for this pull request.
Title check ✅ Passed The title clearly and concisely describes the main change: updating ACP adapter tests to avoid prompt-settlement races.
Description check ✅ Passed The description includes all required sections. It explains the intermittent failure, the test-only fix, scope justification, reproduction steps, verification results, limitations, and agent details.
✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create a new PR
  • Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Autopilot is currently an internal CodeRabbit preview.


Comment @coderabbitai help to get the list of available commands.

…ssing

Ten AcpAdapterV2 tests waited for the raw `stopReason` line on the wire,
yielded twice, and assumed the adapter had marked the prompt settled. The
adapter only does that several async hops later, in the prompt completion
callback. When Stop won that race the interrupt took the cancel-and-wait
path, and the test's `TestClock.adjust("10 seconds")` fired the interrupt
timeout immediately.

Wait on the existing `afterPromptSettledWithBackgroundWork` test hook
instead, as three other tests in the file already do.
@tris203
tris203 force-pushed the fix/acp-adapter-test-settle-race branch from db03ae5 to e0c3f55 Compare October 2, 2026 23:19
t3dotgg added a commit that referenced this pull request Oct 3, 2026
)

Takes over #14876.

Co-authored-by: tris203 <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>
@t3dotgg

t3dotgg commented Oct 3, 2026

Copy link
Copy Markdown
Member

Note

🤖 Claude Opus 5.5 responding on behalf of Theo

Landed in #15330 (a21702d), so I'm closing this one. Your commit went in unchanged, and the squash commit lists you as co-author.

The only addition is a follow-up commit. Your hook wait means interruptTurn always finds the prompt settled, so the forked interrupt, the TestClock.adjust("10 seconds"), and the unused protocol log queues could go. Thanks for tracking down this flake, @tris203. It was also failing on main.

@t3dotgg t3dotgg closed this Oct 3, 2026
@tris203
tris203 deleted the fix/acp-adapter-test-settle-race branch October 3, 2026 21:35
aorwall pushed a commit to aorwall/t3code that referenced this pull request Oct 7, 2026
* chore: docs, dev scripts and CI catch up with orchestration V2 (pingdotgg#15041)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): Claude V2 turns start on Windows with the default binary path (pingdotgg#15021)

Co-authored-by: maria-rcks <[email protected]>

* fix(web): diff panel opens on all branch changes, not just uncommitted (pingdotgg#15005)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): threads stay working while Claude starts a wake turn (pingdotgg#15055)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(web): mod+alt+enter on an existing thread sends and opens a new thread (pingdotgg#15050)

Co-authored-by: maria-rcks <[email protected]>

* fix(chat): sending on an older thread no longer jumps to the top (pingdotgg#15059)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* chore: add bmdavis419 to triage exemptions (pingdotgg#15062)

* fix(server): runs no longer get stuck (pingdotgg#15048)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(usage): Codex Fast and Ultrafast now cost what they bill (pingdotgg#15101)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(clients): a dev server left running no longer says the thread is waiting (pingdotgg#15114)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): a thread that left a shell running shows its unseen completion (pingdotgg#14910)

Co-authored-by: Theo Browne <[email protected]>

* fix(web): mod+enter starts a new thread in the background again (pingdotgg#15060)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(usage): show cost by token type, speed, and model detail (pingdotgg#15108)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(server): agents can watch a PR and get woken when checks, reviews, or conflicts need them (pingdotgg#15057)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): keep delegated review rounds on the task API (pingdotgg#15115)

* fix(shared): classify workspace previews by literal filenames (pingdotgg#10311)

Co-authored-by: yashranaway <[email protected]>
Co-authored-by: shivam <[email protected]>

* fix(mobile): iOS threads no longer jump to the top (pingdotgg#14808)

* fix(web): reduce the gap above the draft composer (pingdotgg#15196)

* fix(mobile): a dev server left running no longer shows the waiting bolt (pingdotgg#15194)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): a Claude command you stop shows as interrupted (pingdotgg#14896)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): editors appear once a slow discovery scan finishes (pingdotgg#13917)

* fix(server): Claude threads no longer stay stuck in plan mode Claude entered itself (pingdotgg#15224)

* fix(mobile): show complete subagent details (pingdotgg#15189)

* fix(mobile): an expired Live Activity no longer leaves a second card (pingdotgg#15254)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* test(web): remove redundant thread sort fallback tests (pingdotgg#15095)

Co-authored-by: t3-code[bot] <269035359+t3-code[bot]@users.noreply.github.com>

* fix(web): thinking row after a failed tool expands the run's tool calls (pingdotgg#15056)

* perf(web): DOM changes no longer restyle the whole page (pingdotgg#15265)

* perf(usage): cut warm usage scans from seconds to milliseconds on large histories (pingdotgg#15149)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* perf(web): virtualize command palette results (pingdotgg#15266)

* chore(lint): flag :has() variants that restyle the whole page (pingdotgg#15274)

* fix(web): workspace card docks beside chat when the window is narrow (pingdotgg#14992)

Chat stays centered while the workspace card fits beside it with 32px to spare. When it does not fit, chat moves left only as far as needed, narrows only after it reaches the left padding, and the card becomes a popover below a 640px chat. The card is lighter: 280px wide, 32px rows, no section labels, no "Project folder" hint.

Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>

* feat(web): render mermaid code blocks as diagrams (pingdotgg#15067)

Co-authored-by: maria-rcks <[email protected]>

* feat(web): Nightly tells you to get the beta mobile app (pingdotgg#15070)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* test(server): ACP adapter tests no longer race the prompt settle (pingdotgg#15330)

Takes over pingdotgg#14876.

Co-authored-by: tris203 <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(usage): fold preview model IDs into the model they belong to (pingdotgg#15333)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* refactor(server): check RPC scopes in group middleware (pingdotgg#15324)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(mobile): beta Working section hides busy threads until they need you (pingdotgg#15346)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(settings): symlinked settings files stay linked when saved (pingdotgg#15009)

Signed-off-by: Yordis Prieto <[email protected]>

* fix(server): Stop ends a dev server left running before a provider switch (pingdotgg#15355)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): merged threads settle even after the agent wakes on its own (pingdotgg#15388)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): no-project drafts can switch machines (pingdotgg#15356)

* fix(web): highlight tool inputs and remove nested work log indentation (pingdotgg#15384)

* fix(server): restarts keep delegated tasks, queued threads, and stops intact (pingdotgg#15323)

* fix(web): sending past the resume banner compacts first (pingdotgg#15290)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(codex): resume archived native sessions (pingdotgg#15389)

* feat(web): morph composer and panel action icons (pingdotgg#14924)

Co-authored-by: maria-rcks <[email protected]>

* fix(web): subagents sent a follow-up show as running in Lineage (pingdotgg#15334)

Co-authored-by: scratchyone <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): clear stale chat action shortcuts (pingdotgg#15394)

* fix(orchestration-v2): restore earlier app agent transcript pages (pingdotgg#14104)

* fix(web): remove the square thread info panel shadow (pingdotgg#15069)

* fix(mobile): Android usage widget no longer sticks on "Loading widget" in release builds (pingdotgg#15142)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): size the model picker to its content (pingdotgg#15152)

Co-authored-by: saphid <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* test(server): replay checks a Claude subagent's thread takes its reported model (pingdotgg#15022)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): subagent finish notifications look like subagent cards (pingdotgg#15281)

* fix(web): thread status dot has an accessible name (pingdotgg#14587)

* fix(web): legacy sidebar options button has a label (pingdotgg#14602)

* fix(web): imported themes keep switches and focus rings visible (pingdotgg#14498)

* fix(web): links to issues no longer strand the pull request viewer (pingdotgg#14242)

* fix(web): repo/task breadcrumb no longer bounces when the sidebar collapses (pingdotgg#15046)

* fix(web): Pull request panel entry works for linked PRs (pingdotgg#15061)

* fix(web): add context menu to draft threads in the sidebar (pingdotgg#10637)

* fix(web): keep sidebar branding and build pills from clipping at varying font sizes and zoom levels (pingdotgg#12141)

* fix(usage): model shares and order follow the selected metric (pingdotgg#11391)

* feat(web): sweep sidebar buttons to settle, un-settle, and wake threads (pingdotgg#14768)

Co-authored-by: maria-rcks <[email protected]>

* feat: retry a failed workspace preparation (pingdotgg#15326)

* fix(server): registry test stubs no longer outlive the test run (pingdotgg#15457)

Signed-off-by: Yordis Prieto <[email protected]>

* test(server): the registry's fake Claude CLI is a fixture file, not a generated string (pingdotgg#15463)

Signed-off-by: Yordis Prieto <[email protected]>

* refactor(clients): share opening a machine's No project folder (pingdotgg#14759)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>
Co-authored-by: Julius Marminge <[email protected]>

* test(server): the git-ssh wrapper's fake SSH script is a fixture file, not a generated string (pingdotgg#15480)

Signed-off-by: Yordis Prieto <[email protected]>

* test(server): the ACP registry's fake npm is a fixture file, not a generated string (pingdotgg#15483)

Signed-off-by: Yordis Prieto <[email protected]>

* test(server): the ACP registry's fake uv is a fixture file, not a generated string (pingdotgg#15484)

Signed-off-by: Yordis Prieto <[email protected]>

* feat(clients): step a new thread to the next machine from the keyboard (pingdotgg#15391)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): promoting a draft thread no longer logs a React key warning (pingdotgg#15458)

Signed-off-by: Yordis Prieto <[email protected]>

* test(server): the text generation's fake Claude CLI is a fixture file, not a generated string (pingdotgg#15479)

Signed-off-by: Yordis Prieto <[email protected]>

* fix(mobile): keep dictation running across navigation behind an edge pill (pingdotgg#15502)

Co-authored-by: Bil0000 <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(client-runtime): relay disconnects no longer show as thread errors (pingdotgg#15470)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): subagent cards name the provider account (pingdotgg#15493)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): read paginated review replies when watching PRs (pingdotgg#15427)

Co-authored-by: maria-rcks <[email protected]>

* fix(server): offer one-click provider updates for every install (pingdotgg#15416)

* fix(mobile): keep the dictation timer from shifting width (pingdotgg#15504)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(relay): T3 Connect links no longer fail on colliding prepared statements (pingdotgg#15411)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): sqlite transactions wait for the write lock instead of failing (pingdotgg#15488)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): unpin button shows the pin-off icon on hover (pingdotgg#15425)

* fix(mobile): make queued message removal tappable (pingdotgg#15417)

* fix(web): keep workspace panels below dialogs (pingdotgg#15454)

* fix(clients): Working section keeps its order while agents finish and wake (pingdotgg#15418)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(mobile): full-screen simulator viewer with on-demand controls (pingdotgg#15551)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(client-runtime): closing a busy stream no longer drops the connection (pingdotgg#15563)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(web): add shift-held pull request quick actions (pingdotgg#15549)

Co-authored-by: maria-rcks <[email protected]>

* fix: expanded tool calls show their output, empty ones don't expand (pingdotgg#15505)

Co-authored-by: maria-rcks <[email protected]>

* fix(web): show device diagnostics before hub readiness (pingdotgg#15435)

* fix(web): open thread picker for unsent drafts (pingdotgg#15436)

* fix(desktop): print version before initializing the app (pingdotgg#15440)

* fix(server): recover claude skill scalar frontmatter (pingdotgg#15452)

* fix(server): keep settled threads asleep after restarts (pingdotgg#15604)

* fix(server): avoid inferring forgejo conflicts from mergeability (pingdotgg#15441)

* fix(web): dismiss hovered timeline tooltips on scroll (pingdotgg#15455)

* fix(server): discover Claude commands in each workspace (pingdotgg#15462)

* fix(web): restore project action preview opening (pingdotgg#15490)

* fix(desktop): keep titlebar controls inset when zoomed (pingdotgg#15496)

* fix(source-control): use the Azure DevOps mark (pingdotgg#15512)

* fix(web): open provider update details from both icons (pingdotgg#15501)

* fix(web): reveal sidebar actions for secondary hovering pointers (pingdotgg#15536)

* fix(markdown): preserve descriptive file-link labels (pingdotgg#15509)

* feat(clients): tool calls show the call above a muted result, without cards (pingdotgg#15506)

Co-authored-by: maria-rcks <[email protected]>

* fix(chat): repair unclosed local file links in assistant responses (pingdotgg#15520)

* fix(server): match manual update commands to installed cli (pingdotgg#15539)

* fix(server): preserve staging during commit message generation (pingdotgg#15532)

* fix(mobile): Keep the last line of iOS markdown replies visible (pingdotgg#15737)

* feat(release): include nightly changelogs in Discord announcements (pingdotgg#15754)

* revert(web): remove automatic compaction before resume (pingdotgg#15771)

* fix(server): Claude threads no longer get stuck after background commands (pingdotgg#15770)

* fix(cli): reject accidental server launches (pingdotgg#15795)

* feat(clients): reach one environment over several routes (pingdotgg#15467)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(clients): learn an environment's LAN and tailnet addresses (pingdotgg#15468)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): share MCP tool presentation across providers (pingdotgg#15475)

Co-authored-by: Bil0000 <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* revert(chat): remove automatic file-link repair (pingdotgg#15824)

Co-authored-by: maria-rcks <[email protected]>

* perf(web): validate monospace fonts when selected (pingdotgg#15642)

* fix(server): expand home-relative media paths (pingdotgg#15618)

* fix(server): recover Linux runtime directory for device hub (pingdotgg#12402)

* fix(web): center icons in thread details icon buttons (pingdotgg#15669)

Co-authored-by: Claude Opus 5.5 <[email protected]>

* fix(mobile): back from an agent's thread returns to its parent (pingdotgg#15068)

* fix(dev): worktree setup never deletes a real env file (pingdotgg#15845)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): drop the duplicate Option import that breaks main CI (pingdotgg#15847)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(dev): write bootstrap warnings directly to stderr (pingdotgg#15865)

* fix(mobile): a message that fails to send now says why in the thread (pingdotgg#15807)

Co-authored-by: T3 Code Test <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): queue background notifications during active tools (pingdotgg#15892)

* refactor(server): share one keyed lock that releases idle keys (pingdotgg#15577)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(server): T3 MCP tools take explicit thread and project targets (pingdotgg#15219)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(desktop): V2 imports stashed prompts and drafts from the V1 profile (pingdotgg#15072)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(web): shell commands in the timeline are syntax highlighted (pingdotgg#15037)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>
Co-authored-by: Yash Singh <[email protected]>

* fix(mobile): upgrade Uniwind and remove local patch (pingdotgg#14597)

* fix(server): Stop ends a Codex command after its thread was settled (pingdotgg#15546)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): subagents no longer inherit parent pull-request links (pingdotgg#14918)

Co-authored-by: Simone <[email protected]>

* fix(prs): queue fast actions and close batches by dragging (pingdotgg#15851)

Co-authored-by: maria-rcks <[email protected]>

* perf(prs): share concurrent github routing metadata probes (pingdotgg#15853)

Co-authored-by: maria-rcks <[email protected]>

* fix(mobile): back from a finished subagent in the feed returns to its parent (pingdotgg#15844)

* fix(desktop): bound preview inspector retention and record renderer identity (pingdotgg#16032)

* fix(web): show fast mode beside reasoning as text (pingdotgg#16069)

Co-authored-by: maria-rcks <[email protected]>

* fix(mobile): make the routes list match the other settings rows (pingdotgg#15958)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(threads): stop pull request watches when settling (pingdotgg#16095)

* feat(contracts): clients tolerate union members they don't know yet (pingdotgg#15951)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* refactor(contracts): project icons decode forward-compatibly instead of encoding a fallback (pingdotgg#16118)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* Removed an unused helper from the Android push payload builder (pingdotgg#16116)

* perf(mobile): reduce shell cache encoding work (pingdotgg#15096)

* perf(mobile): defer audio recorder creation until dictation (pingdotgg#15248)

* feat(server): bump Antigravity ACP agent to 1.3.0 (pingdotgg#15746)

* feat(acp): support local provider commands (pingdotgg#16021)

* fix(server): honor submodule settings when creating worktrees (pingdotgg#15594)

* chore(deps): upgrade Effect to stable 4.0.1 (pingdotgg#16138)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): worktree threads survive a local branch named t3code (pingdotgg#16167)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* chore(refs): sync Effect and Alchemy references to 4.0.1 and beta.80 (pingdotgg#16170)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): match subagent timestamp fonts to chat (pingdotgg#16151)

* fix(web): wrap full status text in composer hover details (pingdotgg#16158)

* ci: run the transfer report job on Blacksmith (pingdotgg#16178)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): Stop also stops delegated tasks and pull request watches (pingdotgg#16002)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat: native /goal for Codex and Claude, with goal status in the UI (pingdotgg#15592)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): use current SQL import in thread stop tests

* fix(server): name the cause of a failed git command (pingdotgg#8645)

Co-authored-by: Julius Marminge <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): PR watch wakes the agent when a bot edits its review comment (pingdotgg#15415)

* feat(source-control): omit agent credits from PR merge messages (pingdotgg#16192)

* fix(clients): dropped connections say why in the client trace (pingdotgg#16200)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): PR watch reports a required check that first appears already passed (pingdotgg#15804)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix: a failed DPoP key load and a fresh maintenance read are no longer cached (pingdotgg#15500)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix: tool screenshots show as images, not base64 text (pingdotgg#16199)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* docs(mcp): thread tools reach threads in any project (pingdotgg#15947)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(threads): threads watching a PR stay in Working instead of bouncing to the inbox (pingdotgg#16204)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* chore(deps): bump cursor sdk and astro to clear vulnerable transitives (pingdotgg#16214)

Co-authored-by: Claude Fable 5.1 <[email protected]>

* fix(server): PR sync waits out a GitHub rate limit pause instead of failing every PR (pingdotgg#16203)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(server): scheduled tasks can run on a webhook (pingdotgg#15085)

* feat(relay): forward webhook requests to the environment's tunnel (pingdotgg#15086)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(mobile): create and copy webhook automations (pingdotgg#15087)

* feat(web): create webhook automations and inspect their deliveries (pingdotgg#15088)

* feat(relay,server,web,mobile): opt-in to hold webhooks while offline (pingdotgg#15487)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): PR watches stop burning GitHub's rate limit and giving up (pingdotgg#16208)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): delegation sees a fixed provider without the app open (pingdotgg#16219)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat: new branches use the shorter t3/ prefix (pingdotgg#16220)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* perf: cheaper shell refreshes, one copy of Codex streaming text, no MCP wait polling (pingdotgg#15033)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat: agents can show HTML pages inline in threads (pingdotgg#15968)

Co-authored-by: Ben Davis <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>
Co-authored-by: Julius Marminge <[email protected]>

* chore(relay): match Alchemy to the PS-80 Postgres cluster (pingdotgg#16228)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat: agents ask the user for a secret through a private card (pingdotgg#15907)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(web): see and stop pull request watches in the thread details card (pingdotgg#16235)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): ACP mode states with null descriptions are no longer dropped (pingdotgg#16218)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): finished outbox rows and old PR cache files are pruned (pingdotgg#16247)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(relay): releasing a tunnel that still has a connector no longer 500s (pingdotgg#16250)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* refactor(server,relay): webhook capabilities live in services, not handlers (pingdotgg#16232)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): a T3 Connect preferences save finishes even if the client disconnects (pingdotgg#16266)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* refactor(server): import service modules as namespaces, not aliased layers (pingdotgg#16267)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(server): log how long PR watches stay quiet before they end (pingdotgg#16262)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(server,web): choose where new worktrees are created (pingdotgg#16231)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* perf(server): idle status polls and PR sweeps start fewer git processes (pingdotgg#16272)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* perf(server): PR watches spend ~90% fewer GitHub points by checking a 1-point fingerprint first (pingdotgg#16270)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(pull-requests): PR detail reads no longer drain the GitHub quota (pingdotgg#16280)

Takes over pingdotgg#13841. A PR query refreshes on the server's refresh signal only while something reads it, and the server shares detail, activity, and preview for 60 seconds, or 10 minutes once merged.

Co-authored-by: Wout Stiens <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* refactor: layer variables are named layer or layerXyz (pingdotgg#16282)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* refactor(server): T3 Connect link capabilities live in a CloudLink service (pingdotgg#16265)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): sidebar drag and drop no longer snaps back (pingdotgg#16291)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): inline HTML renders no longer trap the thread's scroll (pingdotgg#16283)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* refactor(server): one module per service instead of Services/ and Layers/ folders (pingdotgg#16295)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* chore(review): configure CodeRabbit in TypeScript (pingdotgg#16281)

* docs: put the Effect and web UI review rules in the docs (pingdotgg#16286)

* chore(lint): require a reason on every lint and type-checker suppression (pingdotgg#16294)

* refactor(relay): import HookInboxObject once, as a namespace (pingdotgg#16307)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): a rejected desktop-local credential is not retried every poll (pingdotgg#16273)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(desktop): the renderer's bootstrap token rotates every 12 hours (pingdotgg#16275)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): recover from a closed IndexedDB connection (pingdotgg#16311)

Co-authored-by: Lakshmi Tanmay <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(relay): stop forcing manual relay deploys by default (pingdotgg#13563)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(relay): measure the managed tunnel backlog (pingdotgg#13564)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(relay): clean up tunnels of hosts that never registered recovery (pingdotgg#13565)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* perf(relay): delete expired tunnels four at a time within a time budget (pingdotgg#13566)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* feat(connect): tell users when an idle tunnel was removed (pingdotgg#13567)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* docs(relay): add the legacy tunnel cleanup rollout runbook (pingdotgg#13568)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* chore(review): point CodeRabbit at the web UI conventions (pingdotgg#16324)

* chore(review): turn off CodeRabbit's docstring coverage check (pingdotgg#16328)

* refactor(server): CloudLink keeps only the link lifecycle; pure checks live beside it (pingdotgg#16340)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* refactor(server): CloudLink fails with its own errors; the connect routes map them to HTTP (pingdotgg#16341)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* refactor(server): replay guards stay in CloudLink (pingdotgg#16349)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): forks no longer merge into their upstream repo's project group (pingdotgg#16353)

Fixes pingdotgg#4880. Originally pingdotgg#14639 by @Project516.

Co-authored-by: Project516 <[email protected]>

* fix(server): stop the startup project sync from delaying the app window (pingdotgg#14912)

* fix(web): avoid blocking image preparation conversions (pingdotgg#13342)

* fix(server): return partial workspace index on timeout (pingdotgg#11500)

* fix(server): probe project favicon candidates concurrently (pingdotgg#12543)

* fix(observability): a failing trace disk no longer stalls the server (pingdotgg#13758)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): status polling no longer locks the git index (pingdotgg#14718)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* perf(shared): scan PATH once per command before spawning, not on every spawn (pingdotgg#12600)

* fix(server): main's startup auto-pull test compiles again (pingdotgg#16357)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): project favicons stop being rescanned every minute (pingdotgg#16206)

Favicons in ProjectEnrichmentService now keep for 15 minutes. Repository identity keeps its 1-minute TTL, so remote changes still show within a minute.

Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): Claude limits load again for users with large transcript histories (pingdotgg#16358)

The Claude capabilities probe now asks for usage with skipBehaviors, so it no longer scans every local transcript and misses its 4 s deadline. Takes over pingdotgg#14456.

Co-authored-by: Ashkaan <[email protected]>

* Add esthor to the list of GitHub users

* fix(server): caches and ids are written atomically (pingdotgg#16242)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): one-shot initializers no longer race (pingdotgg#16260)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): the PR cache sweep only removes real entry files (pingdotgg#16285)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* chore: keep one copy each of undici 8 and ws 8 (pingdotgg#16211)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(shared): DrainableWorker keeps running after a failed item (pingdotgg#16223)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): metrics count interrupted work on the monotonic clock (pingdotgg#16207)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* refactor(web): import connection storage as a namespace in its test (pingdotgg#16315)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(contracts): trimmed IDs round-trip (pingdotgg#16300)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(server): main's settings, keybindings and session tests compile again (pingdotgg#16363)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* chore(lint): catch known tags with Effect.catchTags (pingdotgg#16361)

* fix(observability): T3 Connect tracing stops at the relay boundary (pingdotgg#16314)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(relay): error and deadline responses carry CORS headers (pingdotgg#16253)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* fix(web): bring back the live shimmer on work log rows (pingdotgg#16372)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* refactor: Effect code gets UUIDs and SHA-256 from Effect's Crypto (pingdotgg#16377)

* fix(relay): export traces through one tracer, one request span each (pingdotgg#16382)

Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>

* chore(contracts): take the Moatless V2 backend into the upstream merge

Regenerate the threads.getShell fixture as a V2 row, decode it as the RPC
layer does, drop four UnsupportedMethodError entries the V2 backend now
serves, and reconcile docs/fork/gaps.md with soaplabs/moatless#1068.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

* test(contracts): decode a Moatless V2 thread projection fixture

The fixture comes from the moatless feat/t3code-v2-timeline-and-sessions
branch and holds every timeline item kind it translates tool calls into,
its plans, and the provider rows that let a client steer a running turn.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

* ci: run the contracts tests in the Typecheck workflow

Its Moatless fixtures are the one check that a backend response decodes
against the schemas the client reads, and the package is small enough for
the 4 CPU / 8 GiB runner.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

* ci: run the contracts tests before the typecheck

The runner loses contact during pnpm typecheck, which skipped the
fixture decodes queued after it.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

* test(contracts): decode the V2 projection's node rows

Regenerated from soaplabs/moatless#1071 at ed50318e, which backs every
rootNodeId and nodeId with a node row.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

* docs(fork): narrow the V2 gap to what moatless#1071 still refuses

Co-Authored-By: Claude Opus 5.5 <[email protected]>

---------

Signed-off-by: Yordis Prieto <[email protected]>
Co-authored-by: Theo Browne <[email protected]>
Co-authored-by: Claude Opus 5.5 (1M context) <[email protected]>
Co-authored-by: maria <[email protected]>
Co-authored-by: maria-rcks <[email protected]>
Co-authored-by: Ben Davis <[email protected]>
Co-authored-by: Igor Makowski <[email protected]>
Co-authored-by: Aditya Garud <[email protected]>
Co-authored-by: yashranaway <[email protected]>
Co-authored-by: shivam <[email protected]>
Co-authored-by: Noé <[email protected]>
Co-authored-by: Bilal Bakr <[email protected]>
Co-authored-by: Jake Leventhal <[email protected]>
Co-authored-by: Dara Adedeji <[email protected]>
Co-authored-by: Bob Fowler <[email protected]>
Co-authored-by: Julius Marminge <[email protected]>
Co-authored-by: t3-code[bot] <269035359+t3-code[bot]@users.noreply.github.com>
Co-authored-by: oliver <[email protected]>
Co-authored-by: tris203 <[email protected]>
Co-authored-by: Yordis Prieto <[email protected]>
Co-authored-by: Yash Singh <[email protected]>
Co-authored-by: scratchyone <[email protected]>
Co-authored-by: scratchyone <[email protected]>
Co-authored-by: Alex <[email protected]>
Co-authored-by: Alex Southwell <[email protected]>
Co-authored-by: saphid <[email protected]>
Co-authored-by: Ryan Ilano <[email protected]>
Co-authored-by: Argo <[email protected]>
Co-authored-by: eimexdev <[email protected]>
Co-authored-by: Mike Olson <[email protected]>
Co-authored-by: Exotic <[email protected]>
Co-authored-by: Rakshith Bhat <[email protected]>
Co-authored-by: AKolenda <[email protected]>
Co-authored-by: T3 Code Test <[email protected]>
Co-authored-by: Hubert Bieszczad <[email protected]>
Co-authored-by: Simone <[email protected]>
Co-authored-by: Simone <[email protected]>
Co-authored-by: Kriday Dave <[email protected]>
Co-authored-by: Dipangshu Roy <[email protected]>
Co-authored-by: Rahul Mishra <[email protected]>
Co-authored-by: Wout Stiens <[email protected]>
Co-authored-by: Utkarsh Patil <[email protected]>
Co-authored-by: Julius Marminge <[email protected]>
Co-authored-by: Muhammad Waleed <[email protected]>
Co-authored-by: Guillermo Casanova <[email protected]>
Co-authored-by: Scott Norteman <[email protected]>
Co-authored-by: Erik Thorelli <[email protected]>
Co-authored-by: Lakshmi Tanmay <[email protected]>
Co-authored-by: Project516 <[email protected]>
Co-authored-by: Michel Liao <[email protected]>
Co-authored-by: Ishaan Kothari <[email protected]>
Co-authored-by: ahalekelly <[email protected]>
Co-authored-by: SkiTee3000 <[email protected]>
Co-authored-by: Ashkaan <[email protected]>
Co-authored-by: soap-agentops[bot] <310870250+soap-agentops[bot]@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

macroscope-review Opt PRs made by unvouched contributors in for Macroscope review. Vouched contributors auto-reviews size:L 100-499 changed lines (additions + deletions). vouch:trusted Author is trusted by repo permissions or the VOUCHED list.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants