You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
fix: a failed prompt leaves no workspace verdict, and the workspace telemetry review's follow-ups - #1711
A failed prompt leaves no workspace verdict. A failed attachment cleared the turn's workspace while others still ran, and a ranged file still in its symbol lookup could resolve it again for the next prompt-less loop (summarize, resume). Now every attachment settles first.
An onboarding first-job mark uses the credential of the tool's own turn. The trigger site notes the turn's credential by call id for that call's plugins, and the hook claims under it, so a tenant switch mid-turn or a late return can't drop the other tenant's row. The plugin hook input is unchanged.
Tests: the first-events loop test runs with the workspace feature on, and the telemetry event list has first_prompt_sent and first_job_completed.
One comment no longer names a branch.
The workspace kill switch stops a saved link's first events. With ALTIMATE_DISABLE_WORKSPACE set, a project linked before the switch was set sends no first_prompt_sent or first_job_completed for that workspace, and an onboarding session's own rows name no workspace. Nothing is marked, so the workspace's first rows are sent once the switch is off. Other rows still name the link, as before; the docs say so.
How did you verify your code works?
Each fix's new test fails without it: a missing directory with a slower ranged file, a late return after a tenant switch, a switch mid-turn in a real loop, and the kill switch in a linked turn, an onboarding turn and the onboarding hook. With the feature on, the first-events loop file passes 15/15 and the pinned and turn-boundary files 15/15; each logs one unhandled "All fibers interrupted" error, as on main. test/altimate/telemetry, workspace and plugin pass (1,622 tests). Typecheck and Marker Guard pass.
Cost side effects
None. A failed prompt waits for its other attachments before it reports the error.
Screenshots / recordings
Not a UI change.
Checklist
I have tested my changes locally
I have not included unrelated changes in this PR
I checked for cost side effects (model calls, queries or cloud resources that grow with user data) and described them, or wrote "none"
Fixes a bug where a failed prompt could leave a stale workspace verdict for the next loop to take, and makes the first-job mark use the credential of the turn that ran the tool.
Bug Fixes
Now every attachment settles before a failure is rethrown; a new test in pinned-workspace-loop.test.ts covers the race.
The first-job mark now claims under the turn's credential, noted per call id, so a tenant switch mid-turn or a late return doesn't drop the other tenant's row.
Follow-ups from the workspace telemetry review
With ALTIMATE_DISABLE_WORKSPACE set, no first event names the saved link: a project linked before the switch was set sends no first_prompt_sent or first_job_completed, onboarding rows name no workspace, and nothing is marked, so the workspace's first rows are sent once the switch is off. All other rows still name the link.
The first-events loop test runs with the workspace feature on and clears any pin from the runner's environment.
The telemetry event-type list adds both first-event types.
Written for commit 88756ea. Summary will update on new commits.
Summary by CodeRabbit
Bug Fixes
Workspace attribution for onboarding completion events now follows the workspace resolved for the tool call, avoiding incorrect attribution when a project link differs from the active workspace.
Disabling workspace features now suppresses workspace first-event telemetry and removes workspace attribution from onboarding-session rows. Other turn telemetry continues to include the cached project link.
Prompt processing now waits for all parts to finish before reporting a failure, helping prevent incomplete workspace binding after an error.
Tests
Expanded coverage for workspace attribution, telemetry suppression, and prompt failure scenarios.
…st gaps
- Every attachment part settles before a failed one is rethrown, so a late part can't put back the cleared verdict. Tested.
- The first-events loop test runs with the workspace feature on.
- The telemetry event list has both first-event types.
- A comment no longer names a branch.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VEmKQswEEpkYFmMEngESx8
…link
States what the CLI already does: a cached link still names the turn's rows and still sends the workspace's first events once. No code change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VEmKQswEEpkYFmMEngESx8
No actionable comments were generated in the recent review.
Walkthrough
The change filters workspace first-event telemetry when the workspace kill switch is set and records call-scoped credentials for onboarding completion claims. Prompt input-part processing now waits for all operations to settle before propagating a rejection. Tests and telemetry documentation cover these behaviors.
Changes: Workspace telemetry and prompt processing
First-event functions filter workspace IDs through the kill switch. Tests and documentation cover event suppression, workspace attribution for other turn rows, and event-type completeness.
Prompt call scope and settlement packages/opencode/src/session/prompt.ts, packages/opencode/test/session/pinned-workspace-loop.test.ts, packages/opencode/test/session/workspace-first-events-loop.test.ts
Prompt tool execution records the bound workspace scope for each call. Input-part processing waits for all operations to settle before propagating a rejection. Tests cover call scope and a failed prompt with delayed attachment processing.
Onboarding completion resolves the event workspace through the kill-switch filter and uses the tool call’s recorded scope when claiming the first-job event. Tests cover tenant scope and disabled-workspace behavior.
AltimateAI/altimate-code#1705: Changes the workspace kill-switch behavior and first-event function contracts that this change also updates.
AltimateAI/altimate-code#1513: Adds workspace and turn inputs to resolveTools, which this change extends with workspace scope.
Suggested reviewers:sahrizvi, anandgupta42
Merge Risk: Minimal · up to 88756
The change makes workspace first-event telemetry respect the kill switch, ties the first-job mark to the originating tool call's credential, and waits for attachments to settle before a failed prompt returns. No actionable merge-blocking risk remains after review.
Pre-merge checks: 4
Check name
Status
Explanation
Title check
The title names a real part of the change: a failed prompt now waits for all attachments before reporting an error, so it leaves no stale workspace verdict. It also names the follow-up review fixes, w…
Description check
The description follows the template. It has the issue link, type of change, a detailed 'What does this PR do?' section, verification steps, and a cost side effects section. It is complete, and the ch…
Linked Issues check
The only directly linked issue, #1698, is closed and completed. Under the assessment rules, it supplies background only, so no active linked-issue coding requirements apply. The PR's stated objectives…
Out of Scope Changes check
All changes fall within the workspace telemetry and workspace-binding area that the PR states it addresses. Changes include: (1) Promise.allSettled in packages/opencode/src/session/prompt.ts so a fail…
Full details: Title check
Explanation
The title names a real part of the change: a failed prompt now waits for all attachments before reporting an error, so it leaves no stale workspace verdict. It also names the follow-up review fixes, which the description covers. It is broad because the PR also covers the kill switch, the tenant-scoped first-job mark, and test changes, but none of it is off-topic.
Full details: Description check
Explanation
The description follows the template. It has the issue link, type of change, a detailed 'What does this PR do?' section, verification steps, and a cost side effects section. It is complete, and the checklist is marked. One minor point: the template asks for a closed issue reference and a 'Documentation' box is checked, which matches the docs change.
Full details: Linked Issues check
Explanation
The only directly linked issue, #1698, is closed and completed. Under the assessment rules, it supplies background only, so no active linked-issue coding requirements apply. The PR's stated objectives are not checked against #1698's text as requirements. The PR also says it closes #1698, but this does not create an open requirement.
Full details: Out of Scope Changes check
Explanation
All changes fall within the workspace telemetry and workspace-binding area that the PR states it addresses. Changes include: (1) Promise.allSettled in packages/opencode/src/session/prompt.ts so a failed prompt waits for other attachments before it rethrows, which prevents a stale workspace verdict; (2) per-call credential scope in packages/opencode/src/altimate/workspace/first-events.ts and its use in packages/opencode/src/altimate/plugin/onboarding-telemetry.ts; (3) the ALTIMATE_DISABLE_WORKSPACE filter in firstEventWorkspace, promptSent, and jobCompleted; (4) tests for these paths; (5) adding first_prompt_sent and first_job_completed to the telemetry event-type list; (6) a docs update for the kill-switch behavior; (7) a comment wording change in workspace/state.ts. Each change connects to a stated objective. The evidence is the PR summary and the change summaries. I did not inspect the source diff, so the connection rests on those summaries. No change was found that is clearly unrelated.
Finishing Touches
Generate unit tests (beta)
Commit to this branch
Create a new PR
Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
A rabbit checks the workspace gate, Then tracks each call and scope. First events wait for the right turn, While settled tasks complete. The burrow logs the rows it needs.
Recent review info
Run configuration
Source mode: Auto
Repository: Unavailable
Branch: Unavailable
File: No configuration file from this repository applied
Commit: Unavailable
Layers: Repository UI
Review profile: CHILL
Plan: Advanced
Run ID: 4697acfd-227a-4247-ab6c-f22ae5e62461
Commits
Reviewing files that changed from the base of the PR and between 13a2a01 and 88756ea.
Included review availability: This review used your included allowance. Your plan provides up to 4 included reviews per hour; 2 remain after this review.
…wn turn
The trigger site notes the turn's credential under the call id while that call's plugins run, and the onboarding hook claims under it. A tenant switch mid-turn, or a late return, no longer drops the other tenant's row.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VEmKQswEEpkYFmMEngESx8
…ject's link
A project linked before the switch was set still sent that workspace's first_prompt_sent and first_job_completed. Now neither is sent, onboarding rows name no workspace, and nothing is marked, so the rows come once the switch is off. Other rows still name the link.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VEmKQswEEpkYFmMEngESx8
Review log: claims and disclosed residuals (head 88756ea)
Review against these. A finding is a reproducible trace that violates a numbered claim.
Claims
C1, the verdict: when a prompt fails on an attachment, every part settles before the failure is rethrown (settled, packages/opencode/src/session/prompt.ts:3212, used at :3273 and :3605). So no attachment read still running sets a turn verdict (turnBinding) for the session after prompt has ended it, and the next loop started without a prompt (summarize, resume) resolves its own workspace.
C2, the per-call tenant: an onboarding first_job_completed mark (packages/opencode/src/altimate/plugin/onboarding-telemetry.ts:268-281) is claimed under the credential scope of the turn that ran the tool. The trigger site notes that scope by call id (prompt.ts:2983, packages/opencode/src/altimate/workspace/first-events.ts:95-108), and the claim doesn't read the credential at claim time. A credential switch mid-turn, or a tool returning after its turn was cancelled and replaced, can't mark another tenant's workspace or leave its own unmarked.
C3, the kill switch: with ALTIMATE_DISABLE_WORKSPACE set, no first_prompt_sent or first_job_completed row carries a workspace_id, and no first-event mark is written (first-events.ts:112-140, onboarding-telemetry.ts:268). An onboarding session's own rows are still sent, without workspace_id. With the switch unset, which first-event rows are sent and what they carry is unchanged from main.
Disclosed residuals (not findings)
R1: with the switch set, the turn's other rows (session_start, task_outcome_signal and the rest) still name the project's saved link, read from the local cache. This is the decided behaviour, and the telemetry docs say so.
R2: the switch is read at each first-event decision, so setting it mid-session applies from the next prompt or turn end.
R3: the switch doesn't clear marks written, or rows sent, before it was set.
R4: the loop test files log one unhandled "All fibers interrupted" error at teardown, as on main.
R5 (cubic, 06:54Z): C2 assumes tool call ids are unique among calls running at the same time in one process. A provider that repeats an id across two overlapping calls whose turns ran under different credentials could have one onboarding sample-setup mark take the other's tenant. Provider ids pass through unchanged, and the AI SDK makes them unique.
R6 (Kilo, 07:12Z):LSP.documentSymbol has no timeout, so a hung language server stalls a prompt with a ranged attachment whose range came back empty, as on main. With C1, a failing sibling part also waits for it.
R7 (Kilo): only the native tool path notes a call scope. No other path can complete an onboarding job today: the MCP and task triggers can't reach isJobCompletion as true.
R8 (Kilo): the verdict test orders its parts with a 300 ms timer. On a slow host it can pass without the fix, but it can't fail with it.
@codex review against the numbered claims and the disclosed residuals in the review-log comment on this PR: report only a reproducible trace that violates a numbered claim. Instances of the disclosed residuals are not findings. A round with no claim violation ends review.
The reason will be displayed to describe this comment to others. Learn more.
1 issue found across 9 files
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="packages/opencode/src/altimate/workspace/first-events.ts">
<violation number="1" location="packages/opencode/src/altimate/workspace/first-events.ts:92">
P2: `callScopes` overwrites an earlier scope whenever overlapping tool executions reuse the same `callID`, so a late `tool.execute.after` can claim the workspace under the replacement turn's tenant or fall back to the current tenant. Key the scope by an execution-unique token (or retain per-call entries) rather than relying on provider call IDs being unique across concurrent turns.</violation>
</file>
Reply with feedback, questions, or to request a fix.
The reason will be displayed to describe this comment to others. Learn more.
P2: callScopes overwrites an earlier scope whenever overlapping tool executions reuse the same callID, so a late tool.execute.after can claim the workspace under the replacement turn's tenant or fall back to the current tenant. Key the scope by an execution-unique token (or retain per-call entries) rather than relying on provider call IDs being unique across concurrent turns.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At packages/opencode/src/altimate/workspace/first-events.ts, line 92:
<comment>`callScopes` overwrites an earlier scope whenever overlapping tool executions reuse the same `callID`, so a late `tool.execute.after` can claim the workspace under the replacement turn's tenant or fall back to the current tenant. Key the scope by an execution-unique token (or retain per-call entries) rather than relying on provider call IDs being unique across concurrent turns.</comment>
<file context>
@@ -80,8 +84,38 @@ export async function claim(event: FirstEvent, workspaceId: number, scope?: Scop
+ * call's workspace id belongs to, even when the credential has switched since, or the turn was
+ * cancelled and replaced before the tool returned. Keyed by call id, not session: a session's later
+ * turn has its own calls. */
+const callScopes = new Map<string, Scope>()
+
+/** Notes `scope` for `callID` until the returned handle is disposed. */
</file context>
The reason will be displayed to describe this comment to others. Learn more.
Declined for this PR, recorded as R5 on the review log.
The overwrite needs two tool calls running at the same time in one process with the same call id. The id is the provider's toolCallId. sanitizeToolCallID replaces only a missing or non-string one, and the AI SDK providers return or generate unique ids. Even with a repeated id, the wrong tenant is taken only when the two calls' turns ran under different credentials, and only for an onboarding sample-setup completion. With the same credential, the fallback to the current credential names the same tenant.
Keying by something unique to each execution would mean adding it to the plugin hook input, which this PR keeps unchanged, or carrying it in async context. Either is a bigger change than this case warrants before this release.
The reason will be displayed to describe this comment to others. Learn more.
WARNING: A failed prompt now waits for every attachment, so an attachment that never settles withholds the error indefinitely.
Promise.allSettled removes the fail-fast behaviour: createUserMessage no longer rejects until every part settles. The ranged text/plain attachment part awaits LSP.documentSymbol (packages/opencode/src/session/prompt.ts:3397), which runs client.connection.sendRequest("textDocument/documentSymbol", ...) with no timeout (packages/opencode/src/lsp/lsp.ts:301-304,427-433); the .catch(() => []) there only handles rejection, not a promise that never settles, and ctx.abort for these reads is a fresh never-aborted AbortController (prompt.ts:3435). So with one failing sibling part (e.g. the missing-directory read) and a hung LSP server or unresponsive language server, the failure used to surface immediately and now never surfaces: the session stays busy with no error and no way to know the prompt failed.
Suggestion: keep waiting for the parts that settle, but bound the wait (per-part timeout, or race the settled set against a deadline) so the original error is still reported when a sibling hangs.
The reason will be displayed to describe this comment to others. Learn more.
Declined for this PR, recorded as R6 on the review log.
The hang comes from LSP.documentSymbol having no timeout, and it is on main too. When no sibling part fails, Promise.all on main also waits for that symbol lookup, so a hung language server stalls every prompt with a ranged attachment whose range came back empty. This PR changes only the case where another part also fails: the error now waits for the hung part as well, instead of surfacing first. Bounding the lookup belongs on documentSymbol itself, which fixes both cases, and is its own change.
The reason will be displayed to describe this comment to others. Learn more.
SUGGESTION:callScope(input.callID) is only populated on the native registry path, so this claim silently falls back to the credential at tool-return time on other paths.
WorkspaceFirstEvents.noteCallScope is called only in resolveTools's native tool wrapper (packages/opencode/src/session/prompt.ts:2983). The MCP wrapper also fires tool.execute.after (prompt.ts:3090-3100) and the task tool fires it too (prompt.ts:1420-1430), but neither notes a scope. The MCP loop assigns tools[key] = item after the native loop (prompt.ts:3163), so an MCP server exposing a tool named sample_setup replaces the native one and reaches this hook through the un-noted path. callScope is then undefined, and claim() falls back to currentScope() (packages/opencode/src/altimate/workspace/first-events.ts:53), which reads the credential now — the exact cross-tenant read this PR removes.
This is latent today because sample_setup is native, but the new mechanism's invariant is "the trigger site notes the scope". Noting it at the MCP (and task) trigger sites, or deriving the scope from the turn context rather than a single call site, keeps the invariant true.
The reason will be displayed to describe this comment to others. Learn more.
Declined, recorded as R7 on the review log: not reachable for this mark.
Only sample_setup completes a job (isJobCompletion, onboarding-telemetry.ts:183), and only the native tool reaches it by that name. The task trigger isn't a job (jobForTool maps no task). An MCP tool's key is <server>_<tool> (mcp/index.ts:1251), so it would take a server named sample with a tool named setup. Even then the MCP trigger passes the raw call result, which carries no metadata.success, so isJobCompletion is false and nothing is claimed. If a job-completing tool is ever added on another path, its trigger site should note the scope as the native one does.
The reason will be displayed to describe this comment to others. Learn more.
SUGGESTION: The 300 ms timer is the only ordering guarantee, so on a slow host this regression test can pass without the production fix.
The test exists to prove a part that finishes after the failure cannot leave a turn verdict behind. If the missing-directory ReadTool takes longer than the 300 ms symbol lookup on a cold CI runner, the ranged part resolves its verdict first and the directory failure lands afterwards, so endTurnBinding is not followed by any late read and both the fixed and unfixed code reach the same RELINKED assertion.
packages/opencode/test/AGENTS.md asks for a published readiness signal instead of a fixed sleep. For example, have the failing part set a deferred/flag that symbolLookup awaits, so the late part provably settles after the failure rather than relying on relative timing.
The reason will be displayed to describe this comment to others. Learn more.
Declined for this PR, recorded as R8 on the review log.
Agreed that the timer is the ordering guarantee. On a slow host the test can only pass without the fix; it can't fail with it, so it never blocks CI. It failed without the fix locally when it was written. Replacing the timer with a readiness signal is a test-only change worth making, and it can follow without touching the release.
Status: 3 Issues Found | Recommendation: Address before merge
Overview
Severity
Count
CRITICAL
0
WARNING
1
SUGGESTION
2
Issue Details (click to expand)
WARNING
File
Line
Issue
packages/opencode/src/session/prompt.ts
3273
Switching to Promise.allSettled means a failed prompt now waits for every attachment; the ranged text/plain part awaits LSP.documentSymbol, which has no timeout and only catches rejections, so a hung language server withholds the prompt's error indefinitely (session stays busy). Bound the wait while still letting parts settle.
noteCallScope is only called on the native registry tool path; the MCP and tasktool.execute.after triggers do not note one, and MCP tools replace native ones in the tools map, so a job-mapped MCP tool falls back to currentScope() at claim time — the cross-tenant read this PR removes.
The 300 ms timer is the test's only ordering guarantee; on a slow CI host the ranged part can finish before the failing directory read, letting the regression test pass without the production fix. Use a readiness signal instead of a fixed sleep.
Note: the callScopes call-id collision raised by an existing inline comment (first-events.ts:92) was reviewed and not re-raised, since that thread is already active.
The reason will be displayed to describe this comment to others. Learn more.
Consensus review (Claude lead + Codex, both rounds completed): APPROVE
All three claims hold for ordinary inputs, and three of the four earlier minors are fixed. M1 (the onboarding first-job credential) is partly fixed: it is correct for distinct call ids, and minors 2 and 4 cover what is left. Every fix has a test that fails when the fix is reverted (10 of 10 mutations caught). test/session, test/altimate/{telemetry,workspace,plugin}, test/plugin and test/telemetry pass at head and at base, typecheck passes, and every prompt.ts edit sits inside altimate_change markers. The "All fibers interrupted" unhandled error also appears on base.
No critical or major finding is open. The remaining minors are inline, and none blocks:
A failed prompt never reports its error when a sibling's symbol lookup hangs (prompt.ts:3273): an edge input on a hang that already exists on main.
A repeated raw call id can swap or drop the noted tenant (first-events.ts:97): an edge input, safe in a follow-up.
The error reported is the first failure in input order, not in time (prompt.ts:3213): cosmetic.
A hook that never settles keeps its call-scope entry (prompt.ts:2983): an edge input.
The verdict test's slow-lookup seam is never called (pinned-workspace-loop.test.ts:372): test-only.
Codex first rated minor 1 as major, then revised it to minor after a probe showed main hangs on the same input. Nits are in a separate comment.
The reason will be displayed to describe this comment to others. Learn more.
Minor (Codex; reproduced by the lead). A failed prompt now waits for every sibling. If a ranged ?start=N&end=N file's LSP.documentSymbol never answers (no timeout, R6), the error is never reported.
Probe: a missing directory plus missing.txt?start=3&end=3, with the legacy LSP.documentSymbol spied to return a promise that never settles. Head was still pending after 8 s; base rejected in 537 ms.
Not a blocker: on main the same hung lookup already stalls every prompt with that attachment, and no loop starts. Safe in a follow-up: bound documentSymbol, or report the first rejection and clear the verdict once every part has settled.
The reason will be displayed to describe this comment to others. Learn more.
Minor (Codex).find picks the first rejected part in input order. Promise.all reported the first rejection in time. Example: a missing directory first and a malformed URL second. Main shows the URL error; head waits and shows the directory error. This only changes which diagnostic is shown, so it's cosmetic, but the PR's "error is still the first failure" doesn't hold literally.
The reason will be displayed to describe this comment to others. Learn more.
Minor (Codex). The key is the raw options.toolCallId (prompt.ts:2870), and processor.ts:158 says a provider may repeat one malformed raw id for concurrent calls. So R5's "the AI SDK makes them unique" doesn't hold for this repo.
When two overlapping calls share an id under different turn credentials, the second set overwrites the first. If the second call disposes first, the first call's lookup returns undefined and claim falls back to the claim-time credential.
This is an edge input that costs one first-event attribution. Safe in a follow-up, for example by keying on the processor's execution identity.
The reason will be displayed to describe this comment to others. Learn more.
Minor (Codex). Disposal happens only when this block exits. A tool.execute.after hook that never settles keeps its entry, because cancelling the session doesn't unwind a pending hook. That leaves one entry per hung hook, with no hard cap. It's an edge input: a hung hook already hangs this call on main. Success and thrown-error exits dispose correctly.
The reason will be displayed to describe this comment to others. Learn more.
Minor, test-only (lead; Codex agreed). This seam is never called. The stub replaces documentSymbol on LSP.Service (@/lsp/lsp, line 83), but prompt.ts:74 imports LSP from ../lsp, the legacy namespace. An instrumented symbolLookup never printed.
The test still fails when allSettled is reverted to Promise.all (3 of 3 runs), but through natural ordering, not this timer. So the comment above and R8 are inaccurate, and nothing tests a slow or hung lookup. Spying on the namespace (spyOn(LSP, "documentSymbol") from src/lsp) would make the ordering explicit.
Consensus review nits (Claude + Codex), not blocking
No test for call-scope cleanup. Making noteCallScope's dispose a no-op leaves every test green. The onboarding "late return" test also calls resetForTest() before its using handles leave scope, which would hide a missing delete.
The ALL_EVENT_TYPES check is a count. Reverting the new names together with the count stays green, and nothing checks that the list covers the whole Telemetry.Event union.
Marker guard. It reports "ok" because none of the changed paths exist in upstream/dev. Checked by hand instead: every prompt.ts edit is inside altimate_change blocks, and telemetry.test.ts is fork-only, so it needs none.
This branch has not been deployed
No deployments
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
![kilo-code-bot[bot]](https://avatars.githubusercontent.com/in/2193792?s=60&v=4){kind=small}
sahrizvi
left a comment
Issue for this PR
Closes #1698
Type of change
What does this PR do?
first_prompt_sentandfirst_job_completed.ALTIMATE_DISABLE_WORKSPACEset, a project linked before the switch was set sends nofirst_prompt_sentorfirst_job_completedfor that workspace, and an onboarding session's own rows name no workspace. Nothing is marked, so the workspace's first rows are sent once the switch is off. Other rows still name the link, as before; the docs say so.How did you verify your code works?
Each fix's new test fails without it: a missing directory with a slower ranged file, a late return after a tenant switch, a switch mid-turn in a real loop, and the kill switch in a linked turn, an onboarding turn and the onboarding hook. With the feature on, the first-events loop file passes 15/15 and the pinned and turn-boundary files 15/15; each logs one unhandled "All fibers interrupted" error, as on
main.test/altimate/telemetry,workspaceandpluginpass (1,622 tests). Typecheck and Marker Guard pass.Cost side effects
None. A failed prompt waits for its other attachments before it reports the error.
Screenshots / recordings
Not a UI change.
Checklist
🤖 Generated with Claude Code
https://claude.ai/code/session_01VEmKQswEEpkYFmMEngESx8
Summary by cubic
Fixes a bug where a failed prompt could leave a stale workspace verdict for the next loop to take, and makes the first-job mark use the credential of the turn that ran the tool.
Bug Fixes
pinned-workspace-loop.test.tscovers the race.Follow-ups from the workspace telemetry review
ALTIMATE_DISABLE_WORKSPACEset, no first event names the saved link: a project linked before the switch was set sends nofirst_prompt_sentorfirst_job_completed, onboarding rows name no workspace, and nothing is marked, so the workspace's first rows are sent once the switch is off. All other rows still name the link.Written for commit 88756ea. Summary will update on new commits.
Summary by CodeRabbit
Bug Fixes
Tests