Repository navigation
test(dashboard): point the credit capture harnesses at the kiro backend - #25348
Conversation
capture-usage-refresh.mjs and capture-credit-pill-states.mjs served a config with agent.acp_backend unset. Since KAS became the default backend that no longer resolves to kiro, so a failed or empty usage read hid the credit pill and closed the Kiro Account modal, and both harnesses timed out before writing their frames. The shared stub now offers KIRO_BACKEND_CONFIG_FIXTURE, and both harnesses serve it. A vitest pins that it resolves to the kiro backend. Closes #25345
UX Review (Opus 5.5) — ✅ PASSUX-level review of UX-Verdict: PASS Nothing to check. The diff only touches screenshot harness scripts and a test, so no user-visible control changes and no screenshot is needed. [UX-REVIEWED] fdcf9ef Description read[DESCRIPTION-READ] 9fdd34d6c31c372cb7fc8372bbd5feecddf69a081e927e55bd992517473c6311 That sha256 names everything this verdict read: the PR description, plus the 3 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: the change only touches test harnesses and fixes the cause directly. It names the kiro backend explicitly, so the harnesses no longer depend on the unset default resolving to kiro. [DESIGN-REVIEWED] fdcf9ef Description read[DESCRIPTION-READ] d7c46f2a52c2a5928b303c885f57742568fba1e37af38918cc15b49aa9c1cec8 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
First Principles Review (Opus 5.5) — ⏭️ skippedRevision |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Tier 1 auto-approve: test (4 files). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, security checklist all-NO, no reviewer-lane findings, readiness signal fresh for this head SHA, declared category matches the diff. Category: test harness only; the credit capture scripts and their fixture test get a config that resolves to the kiro backend, no shipped code touched.
Four utility lanes that sit where a human is waiting to merge move from the shared hosted pool to the `kirocrew-runners` larger-runner group, reached by its `kirocrew-linux-4-core` label inside a `github.repository` conditional so a personal fork keeps a hosted fallback: - `issue-gate.yml`: `issue-gate` - `pr-size-override.yml`: `record` - `sensitive-change-review-comment.yml`: `rerun` - `disposition-deferral-check.yml`: `validate-deferral` Each is seconds of work whose wall clock is almost all queue wait, so the move buys queue position, not capacity. Both branches of every expression are GitHub-hosted Ubuntu VMs carrying no fleet credentials, so each lane's isolation contract is unchanged; none of these events can reach the AWS webhook's actor allowlist, which is why they stay off the CodeBuild fleet. `merged-readiness-audit.yml`'s `audit` job is deliberately NOT included. `readiness_at_merge` in `scripts/audit_bypass_merge.py` returns the newest `PR Readiness` state created at or before `merged_at`, so a run taken hours late records the same value; the lane has no latency harm to remove and would only occupy a capped group. The `docs/ci/ci-and-reviews.md` inventory edit for this wave is deferred to #26425 rather than carried per PR. Part of #25348
…25349) Four utility lanes that sit where a human is waiting to merge move from the shared hosted pool to the `kirocrew-runners` larger-runner group, reached by its `kirocrew-linux-4-core` label inside a `github.repository` conditional so a personal fork keeps a hosted fallback: - `issue-gate.yml`: `issue-gate` - `pr-size-override.yml`: `record` - `sensitive-change-review-comment.yml`: `rerun` - `disposition-deferral-check.yml`: `validate-deferral` Each is seconds of work whose wall clock is almost all queue wait, so the move buys queue position, not capacity. Both branches of every expression are GitHub-hosted Ubuntu VMs carrying no fleet credentials, so each lane's isolation contract is unchanged; none of these events can reach the AWS webhook's actor allowlist, which is why they stay off the CodeBuild fleet. `merged-readiness-audit.yml`'s `audit` job is deliberately NOT included. `readiness_at_merge` in `scripts/audit_bypass_merge.py` returns the newest `PR Readiness` state created at or before `merged_at`, so a run taken hours late records the same value; the lane has no latency harm to remove and would only occupy a capped group. The `docs/ci/ci-and-reviews.md` inventory edit for this wave is deferred to #26425 rather than carried per PR. Part of #25348
Problem / Motivation
Goal: The two Kiro credit screenshot harnesses produce their frames on main again.
On current main,
website/scripts/capture-usage-refresh.mjsexits 1 at frame 02 (the failed read): the Kiro Account modal fades out before it can be checked.website/scripts/capture-credit-pill-states.mjsexits 1 waiting for the same modal in itsfailedscenario.Why it matters
These harnesses are how a PR that touches the credit pill or the Kiro Account modal shows reviewers what it looks like. While they fail, that evidence has to be built by hand, and a broken frame can pass for a product bug, which is how this was found while working on #23655.
Not a goal
KIROCREW_CONFIG_FIXTURE, which every other harness uses.What changed (motivation → approach → change)
Frames time out → the harnesses serve a config with
agent.acp_backendunset, which since #20359 resolves tokas, not kiro, and on a backend that is not kiro a failed or empty read hides the credit pill and closes the modal → give both harnesses a config that names kiro.website/scripts/lib/stub-dashboard-api.mjsKIRO_BACKEND_CONFIG_FIXTURE: the base fixture withacp_backendset tokirocapture-usage-refresh.mjs/api/config/kirocrew; its comments no longer say unset means kirocapture-credit-pill-states.mjs/api/config/kirocrewBackwards compatibility
Compatible: only screenshot harnesses and a test change, and the existing fixture export is untouched.
Removes nothing: the change adds one export and two route answers.
Tests
website/src/test/stubDashboardApiConfigFixture.test.tsxaddsoffers a config that resolves to the Kiro backend for the credit harnesses. It checks that the new fixture resolves to the kiro backend throughisKiroBackend, and that it differs from the base fixture only inagent. On main it fails with aTypeErrorreadingagentofundefined, since the export does not exist. With this change the file passes, 7 of 7.Manual verification
Built the SPA on main at fe317a6 and ran both harnesses, before and after:
capture-usage-refresh.mjscapture-credit-pill-states.mjsRelated Issues
Closes #25345
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)