refactor(app): migrate App.vue to Pinia (47 commits — App 2211→72 code lines, + store tests/debt) #1511

Workflow file for this run

	name: CI

	on:
	push:
	branches: [main]
	pull_request:
	branches: [main]

	# CI only reads code — no releases or package pushes.
	permissions:
	contents: read

	jobs:
	# ── Job 1: All linters ──────────────────────────────────────────────────
	lint:
	runs-on: ubuntu-latest
	# Extra scopes are needed by EnricoMi/publish-unit-test-result-action so
	# it can post the sticky comment + checks summary on pull requests. The
	# workflow-level default is contents:read; per-job permissions replace
	# rather than merge, so contents:read is re-listed below.
	permissions:
	contents: read
	checks: write
	pull-requests: write
	issues: read
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6

	# mise provisions this job's slice of the toolchain from mise.toml —
	# go, node, task, golangci-lint (built from source with this Go so it
	# accepts the project's go directive — a prebuilt binary on an older Go
	# would reject it), shfmt, shellcheck, ruff, jq, go-junit-report — and
	# loads mise.toml [env] (SPECTRAL_VERSION, …, GOTOOLCHAIN=local) into
	# $GITHUB_ENV regardless of install_args. Replaces setup-go + the
	# tool-versions.env grep + every per-tool `go install`/pipx step.
	- uses: jdx/mise-action@dba19683ed58901619b14f395a24841710cb4925 # v4.1.0
	with:
	install_args: >-
	go node task ruff shfmt shellcheck jq
	go:github.com/golangci/golangci-lint/v2/cmd/golangci-lint
	go:github.com/jstemmer/go-junit-report/v2

	# Frontend — set up Node and build frontend/dist before Go linting so the
	# //go:embed all:frontend/dist directive in assets.go passes type-checking,
	# AND so the bundle-size budget step below can measure the real artifacts.
	- name: Prepare frontend/dist (real Vite build)
	uses: ./.github/actions/prepare-frontend-dist
	with:
	real-assets: "true"

	# Bundle-size budget. Catches accidental regressions (e.g. someone
	# imports a 200KB library by mistake). Two budgets:
	# - Initial chunk (index-*.{js,css}): what every page-load pays.
	# The four view components are lazy-loaded via
	# defineAsyncComponent so each becomes a separate chunk that
	# isn't counted here — only the masthead/modal/router-shell
	# code lands in index-*.
	# - Total assets/: catches the case where lazy-splitting masks a
	# real regression by smearing it across chunks.
	# Bump deliberately when a real feature needs the room.
	# Bundle-size budget. Logic lives in scripts/ci/check-bundle-size.sh
	# so the lefthook pre-push hook can reuse the same assertion —
	# contributors fail locally before push instead of finding out
	# from a red CI badge. Per-chunk + total caps live in the script
	# as defaults; override via env var if a one-off needs room.
	- name: Enforce bundle-size budget
	run: bash scripts/ci/check-bundle-size.sh

	- name: Lint Go (default build tags)
	run: golangci-lint run ./...

	- name: Lint Go (serveronly build tag)
	run: golangci-lint run --build-tags serveronly ./...

	# Frontend linters — deps already installed above
	- name: Lint JavaScript/Vue (ESLint)
	run: cd frontend && npm run lint:js

	- name: Lint CSS (Stylelint)
	run: cd frontend && npm run lint:css

	- name: Lint HTML (HTMLHint)
	run: cd frontend && npm run lint:html

	# Python — ruff lint + format-check. ruff comes from mise (pinned in
	# mise.toml [tools]); no pipx install needed.
	- name: Lint Python (ruff)
	run: task lint-py

	# Shell scripts in scripts/ — shellcheck for correctness, shfmt for
	# style. Both come from mise (pinned in mise.toml), so no install step.
	- name: Lint shell scripts (shellcheck + shfmt)
	run: task lint-shell

	# Dockerfile
	- name: Lint Dockerfile (Hadolint)
	uses: hadolint/hadolint-action@2332a7b74a6de0dda2e2221d575162eba76ba5e5 # v3.3.0
	with:
	dockerfile: Dockerfile.build

	# OpenAPI spec — Spectral with the spectral:oas ruleset + project
	# overrides in .spectral.yaml. SPECTRAL_VERSION comes from mise.toml
	# [env] (loaded into $GITHUB_ENV by mise-action).
	- name: Lint OpenAPI (Spectral)
	run: task lint-openapi

	# Spell-check code, docs, and identifiers. The action ships the
	# typos binary so the runner doesn't need a separate install
	# step; config lives in _typos.toml at the repo root.
	- name: Spell-check (typos)
	uses: crate-ci/typos@37bb98842b0d8c4ffebdb75301a13db0267cef89 # v1.47.2

	# Markdown lint — mirrors `task lint-md`. Config in
	# .markdownlint-cli2.yaml at the repo root; npx pulls the
	# version on demand so no global install is required.
	- name: Lint Markdown (markdownlint-cli2)
	run: npx --yes markdownlint-cli2

	# GitHub Actions workflow lint — catches deprecated inputs,
	# syntax errors, and shellcheck issues in `run:` blocks. The
	# action ships the actionlint binary so no install step.
	- name: Lint workflows (actionlint)
	uses: raven-actions/actionlint@205b530c5d9fa8f44ae9ed59f341a0db994aa6f8 # v2

	# Enforce the project's SHA-pinning policy for third-party
	# GitHub Actions. A tag-pinned action could be silently
	# re-pointed to a malicious commit; see CONTRIBUTING.md
	# → "Pinning GitHub Actions".
	- name: Enforce SHA-pinned actions
	run: bash scripts/ci/check-action-pins.sh

	# Go unit tests (merge logic, inference helpers, screenshotType
	# classification). Parser golden-file tests are skipped by -short.
	# Tee through go-junit-report so the XML is always produced — the
	# final pass/fail decision is delegated to the EnricoMi step below
	# via `fail_on: test failures` so a fork PR without comment perms
	# still surfaces failures via the workflow conclusion.
	# Test-skip inventory gate. Diffs every `t.Skip*` line under
	# pkg/ against `scripts/ci/test-skips-allow.txt`. A new skip
	# without an allow-list entry fails CI — flake-suppression
	# skips are forbidden; the allow-list is for documented
	# environment gates (OS-conditional probe tests, `-short`-mode
	# tesseract integration). Mirrors the lefthook pre-push
	# `test-skips` hook so the gate fires both locally and in CI.
	- name: Inventory test skips
	run: bash scripts/ci/check-test-skips.sh

	- name: Run Go unit tests
	id: gotest
	# continue-on-error so BOTH test steps run and the JUnit XML always
	# uploads + EnricoMi always posts the "Unit test results" check. The
	# real pass/fail is enforced by the "Fail the job on any unit-test
	# failure" gate at the end of this job (reads `.outcome`) — so a
	# failure, including a -race-only one, reds the workflow run, not just
	# the check-run.
	continue-on-error: true
	run: \|
	set -o pipefail
	go test -race -short -v ./... 2>&1 \
	\| tee go-test.log \
	\| go-junit-report -set-exit-code > go-junit.xml

	# Frontend unit tests (Vitest) — pure-helper coverage for the
	# extracted match-helpers.ts module. `--reporter=default` keeps the
	# log readable in the Actions run; `--reporter=junit` adds the XML
	# consumed by the EnricoMi step. The `--outputFile.junit=` form
	# scopes the path to the junit reporter only, so the default
	# reporter still writes to stdout.
	- name: Run frontend unit tests
	id: fetest
	continue-on-error: true
	run: \|
	cd frontend
	npx vitest run \
	--reporter=default \
	--reporter=junit \
	--outputFile.junit=../frontend-junit.xml

	# Upload the JUnit XML so the coverage-comment job can fold its
	# numbers + per-failure details into the single combined PR
	# comment. `if: always()` so the artifact still uploads when a
	# test step exited non-zero (which is the normal path on a
	# failing test — `continue-on-error: true` makes the step succeed,
	# but EnricoMi below is what flips the workflow to red).
	- name: Upload unit-test JUnit XML
	if: always()
	uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
	with:
	name: unit-test-results
	path: \|
	go-junit.xml
	frontend-junit.xml
	if-no-files-found: warn
	retention-days: 7

	# EnricoMi posts the "Unit test results" check (red via `fail_on:
	# test failures` when the JUnit XML carries a failure) — that's the
	# per-PR/commit status + the detailed breakdown. It does NOT fail the
	# workflow run itself, so the explicit `.outcome` gate at the end of
	# this job is what reds CI. Its own commenter is OFF — the
	# coverage-comment job renders one combined sticky comment.
	- name: Publish unit-test results check
	if: always()
	uses: EnricoMi/publish-unit-test-result-action@c950f6fb443cb5af20a377fd0dfaa78838901040 # v2
	with:
	check_name: Unit test results
	comment_mode: off
	files: \|
	go-junit.xml
	frontend-junit.xml
	fail_on: test failures

	# TypeScript type-check — confirms api.ts agrees with the
	# OpenAPI-generated types in api.gen.d.ts.
	- name: TypeScript type-check
	run: cd frontend && npm run typecheck

	# Verify the committed types match the current OpenAPI spec.
	# Catches "you forgot to run `task gen-types` after editing
	# api/openapi.yaml" before it lands on main.
	- name: Verify generated types are in sync with OpenAPI spec
	run: \|
	task gen-types
	if ! git diff --quiet frontend/src/api.gen.d.ts; then
	echo "::error::frontend/src/api.gen.d.ts is out of sync with api/openapi.yaml — run 'task gen-types' and commit"
	git diff frontend/src/api.gen.d.ts \| head -50
	exit 1
	fi

	# Hard-fail CI on any unit-test failure. The two test steps are
	# continue-on-error (so JUnit always uploads + EnricoMi always posts its
	# check), which means their failure alone only reds the "Unit test
	# results" check — the workflow run stays green. Gate on their real
	# `.outcome` here so a failing test, including one that only fails under
	# `-race`, turns this job (and the CI run) red directly. Last step in the
	# job so every other check still runs first.
	- name: Fail the job on any unit-test failure
	if: always() && (steps.gotest.outcome == 'failure' \|\| steps.fetest.outcome == 'failure')
	env:
	GO_OUTCOME: ${{ steps.gotest.outcome }}
	FE_OUTCOME: ${{ steps.fetest.outcome }}
	run: \|
	echo "::error::Unit tests failed (Go=${GO_OUTCOME}, frontend=${FE_OUTCOME}). See the 'Unit test results' check and the test-run step logs."
	exit 1

	# ── Job 2: Linux/amd64 Wails app ────────────────────────────────────────
	build-linux:
	runs-on: ubuntu-latest
	env:
	DOCKER_BUILDKIT: "1"
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
	- uses: docker/setup-buildx-action@d7f5e7f509e45cec5c76c4d5afdd7de93d0b3df5 # v4
	# Docker does the heavy build; the host only needs task + jq (the
	# build-* tasks compute BUILD_VERSION via jq), so install just those.
	- uses: jdx/mise-action@dba19683ed58901619b14f395a24841710cb4925 # v4.1.0
	with:
	install_args: task jq
	- name: Build Linux/amd64 Wails app
	run: task build-linux

	# ── Job 3: Windows/amd64 Wails app ──────────────────────────────────────
	build-windows:
	runs-on: ubuntu-latest
	env:
	DOCKER_BUILDKIT: "1"
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
	- uses: docker/setup-buildx-action@d7f5e7f509e45cec5c76c4d5afdd7de93d0b3df5 # v4
	- uses: jdx/mise-action@dba19683ed58901619b14f395a24841710cb4925 # v4.1.0
	with:
	install_args: task jq
	- name: Build Windows/amd64 Wails app
	run: task build-windows

	# ── Job 4: Server binaries + container image ─────────────────────────────
	build-server:
	runs-on: ubuntu-latest
	env:
	DOCKER_BUILDKIT: "1"
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
	- uses: docker/setup-buildx-action@d7f5e7f509e45cec5c76c4d5afdd7de93d0b3df5 # v4
	- uses: jdx/mise-action@dba19683ed58901619b14f395a24841710cb4925 # v4.1.0
	with:
	install_args: task jq
	- name: Build all server binaries (Linux, Windows, macOS arm64+amd64)
	run: task build-server-all
	- name: Build Linux server container image (with Tesseract)
	run: task build-server-container

	# ── Job 5: macOS Wails build (requires Apple SDK / Xcode CLT) ────────────
	build-mac:
	runs-on: macos-latest
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6

	# Go + Node + Wails CLI (+ task, jq) from mise.toml's pins.
	- uses: ./.github/actions/wails-build-env

	- name: Build macOS universal Wails app
	run: task build-mac

	# ── Job 6: Vulnerability scan ────────────────────────────────────────────
	# Mirrors `task trivy` but runs in CI and uploads results to the GitHub
	# Security tab (Alerts → Code scanning) as well as failing the build.
	trivy:
	runs-on: ubuntu-latest
	permissions:
	contents: read
	security-events: write
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6

	- name: Run Trivy vulnerability scan
	uses: aquasecurity/trivy-action@ed142fd0673e97e23eac54620cfb913e5ce36c25 # v0.36.0
	with:
	scan-type: fs
	scan-ref: .
	scanners: vuln
	severity: HIGH,CRITICAL
	exit-code: '1'
	format: sarif
	output: trivy.sarif

	- name: Upload Trivy results to GitHub Security tab
	if: always()
	uses: github/codeql-action/upload-sarif@8aad20d150bbac5944a9f9d289da16a4b0d87c1e # v4
	with:
	sarif_file: trivy.sarif

	# ── Job 7: Go-specific vulnerability scan ────────────────────────────────
	# govulncheck is call-graph-aware: it only flags CVEs in code paths Recall
	# actually invokes. Complements Trivy (which scans all module versions
	# regardless of reachability) and typically produces fewer false positives.
	govulncheck:
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
	# go + govulncheck from mise.toml (govulncheck built with this Go).
	- uses: jdx/mise-action@dba19683ed58901619b14f395a24841710cb4925 # v4.1.0
	with:
	install_args: go go:golang.org/x/vuln/cmd/govulncheck
	- name: Scan Go modules (default build tag)
	run: govulncheck ./...
	- name: Scan Go modules (serveronly build tag)
	run: govulncheck -tags serveronly ./...

	# gosec is a Go-idiom SAST: catches issues CodeQL and govulncheck don't
	# (Slowloris timeouts, file-mode bugs, weak crypto, taint-tracked path/
	# command/log injection). Mirrors `task lint-gosec`. Both build tags are
	# swept so the Wails and serveronly code paths both get covered.
	gosec:
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
	# go + gosec from mise.toml. mise's go: backend compiles gosec with the
	# pinned Go 1.26.4 — the same reason the old job avoided securego/gosec's
	# docker action (it bundles a stale Go that can't load a go 1.26 module
	# under GOTOOLCHAIN=local). Version pinned in mise.toml (v2.27.1).
	- uses: jdx/mise-action@dba19683ed58901619b14f395a24841710cb4925 # v4.1.0
	with:
	install_args: go go:github.com/securego/gosec/v2/cmd/gosec
	# gosec only needs Go source to compile, not real frontend
	# assets — stub the embed directory rather than running a real
	# Vite build (saves ~30s per CI run).
	- name: Prepare frontend/dist (stub for //go:embed)
	uses: ./.github/actions/prepare-frontend-dist
	with:
	real-assets: "false"
	- name: Run gosec (default build tag)
	run: gosec -exclude-dir=frontend ./...
	- name: Run gosec (serveronly build tag)
	run: gosec -exclude-dir=frontend -tags=serveronly ./...

	# JS/TS SAST in CI is handled by CodeQL's javascript-typescript
	# language matrix in .github/workflows/codeql.yml. Semgrep runs
	# locally only — via `task lint-semgrep` and the pre-push lefthook
	# hook — so contributors get fast feedback before push without
	# duplicating coverage with CodeQL.

	# ── Job 8: Cyclomatic complexity (REPORT ONLY) ───────────────────────────
	# gocyclo (Go) + ESLint's `complexity` rule (frontend) at threshold 10.
	# `continue-on-error: true` keeps the job NON-BLOCKING — a high number
	# never fails the build, it just shows up in the job log. We watch the
	# trend across PRs to know when to refactor; tightening into a hard
	# gate is its own deliberate decision, not a side-effect of this job.
	# Threshold + tooling pins both live in scripts/ci/check-complexity.sh +
	# mise.toml (gocyclo).
	complexity:
	runs-on: ubuntu-latest
	continue-on-error: true
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
	# go (gocyclo) + node (eslint pass) from mise.toml.
	- uses: jdx/mise-action@dba19683ed58901619b14f395a24841710cb4925 # v4.1.0
	with:
	install_args: go node go:github.com/fzipp/gocyclo/cmd/gocyclo
	# gocyclo doesn't need real frontend assets — Go source only.
	- name: Prepare frontend/dist (stub for //go:embed)
	uses: ./.github/actions/prepare-frontend-dist
	with:
	real-assets: "false"
	- name: Install frontend deps (for the eslint pass)
	run: npm --prefix frontend ci
	- name: Run complexity sweep (Go + frontend, threshold 10)
	run: bash scripts/ci/check-complexity.sh
	# Diff the current Go scores against the checked-in top-20
	# baseline. Same threshold, same exclusions; emits a summary
	# line + per-jump/newcomer record. continue-on-error keeps the
	# job non-blocking; a future PR-comment step can hook into
	# this output once the baseline shape stabilises.
	- name: Diff against baseline (report-only)
	run: bash scripts/ci/check-complexity-baseline.sh

	# ── Job 9: Dead code analysis ────────────────────────────────────────────
	# deadcode does whole-program call-graph analysis for Go (serveronly variant
	# only — the Wails variant registers App methods via reflection so deadcode
	# would report false positives; golangci-lint `unused` covers that variant).
	# knip scans the frontend for unused TypeScript exports and stale deps.
	dead-code:
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6

	# go (+ deadcode) and node (+ knip) from mise.toml.
	- uses: jdx/mise-action@dba19683ed58901619b14f395a24841710cb4925 # v4.1.0
	with:
	install_args: go node go:golang.org/x/tools/cmd/deadcode

	- name: Install frontend dependencies
	run: cd frontend && npm ci

	- name: Build frontend (required for Go embed typecheck)
	run: cd frontend && npm run build

	- name: Dead Go code (serveronly build tag)
	# Delegates to scripts/ci/deadcode-check.sh; the allow-list of
	# intentional unreachables lives in scripts/ci/deadcode-allow.txt
	# and is shared with Make + lefthook so all three call sites
	# stay in sync.
	run: bash scripts/ci/deadcode-check.sh

	- name: Dead TypeScript code (knip)
	run: cd frontend && npm run dead:ts

	# ── Job 10: Frontend coverage report ─────────────────────────────────────
	# Informational only — no thresholds enforced. The HTML report and lcov
	# file are uploaded as a workflow artifact so they can be downloaded from
	# the Actions run page for any push or PR.
	coverage-frontend:
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6

	- uses: jdx/mise-action@dba19683ed58901619b14f395a24841710cb4925 # v4.1.0
	with:
	install_args: node

	- name: Install frontend dependencies
	run: cd frontend && npm ci

	- name: Generate coverage report
	run: cd frontend && npm run test:coverage

	- name: Upload coverage report
	uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
	with:
	name: frontend-coverage
	path: frontend/coverage/
	retention-days: 30

	# ── Job 11: Go coverage report ───────────────────────────────────────────
	# Informational only — no thresholds enforced. The HTML report and func
	# summary are uploaded as a workflow artifact so they can be downloaded
	# from the Actions run page for any push or PR.
	coverage-go:
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6

	# go + gocover-cobertura (Cobertura XML converter) from mise.toml.
	- uses: jdx/mise-action@dba19683ed58901619b14f395a24841710cb4925 # v4.1.0
	with:
	install_args: go task go:github.com/boumenot/gocover-cobertura

	# Frontend build is required because assets.go uses
	# //go:embed all:frontend/dist — compiling the root `recall`
	# package (via `go test ./...`) fails if dist/ is missing.
	- name: Prepare frontend/dist (real Vite build)
	uses: ./.github/actions/prepare-frontend-dist
	with:
	real-assets: "true"

	- name: Generate Go coverage report
	run: task cover-go

	# gocover-cobertura converts Go's native coverprofile into the
	# Cobertura XML format that irongut/CodeCoverageSummary parses.
	# Bundled into the artifact so the coverage-comment job downloads
	# both formats in a single fetch.
	- name: Convert coverage profile to Cobertura XML
	run: gocover-cobertura < coverage/go/coverage.out > coverage/go/cobertura.xml

	- name: Upload coverage report
	uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
	with:
	name: go-coverage
	path: coverage/go/
	retention-days: 30

	# ── Job 12: OpenAPI ↔ implementation drift check ─────────────────────────
	# Spectral checks the spec is well-formed; schemathesis checks the spec
	# matches the running server. Fuzzes spec-derived requests against a
	# freshly-built serveronly binary and verifies every response conforms to
	# the declared schema. Catches the "I added a route but forgot to update
	# api/openapi.yaml" class of bug that Spectral can't see.
	schemathesis:
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6

	# go (build serveronly) + schemathesis (pipx) from mise.toml; also loads
	# TESSERACT_VERSION + SCHEMATHESIS_VERSION from [env] into $GITHUB_ENV.
	- uses: jdx/mise-action@dba19683ed58901619b14f395a24841710cb4925 # v4.1.0
	with:
	install_args: go task pipx:schemathesis

	- name: Prepare frontend/dist (real Vite build)
	uses: ./.github/actions/prepare-frontend-dist
	with:
	real-assets: "true"

	- name: Install Tesseract (no-op for API surface but Startup checks for it)
	# Assert installed major.minor matches TESSERACT_VERSION from mise.toml
	# [env] (loaded into $GITHUB_ENV by mise-action) so OCR drift can't
	# silently land between what dev baselined the goldens against and what
	# CI runs. Loud-fail when Ubuntu bumps the package version — that's the
	# signal to re-baseline goldens + bump the pin.
	run: \|
	sudo apt-get update -y
	sudo apt-get install -y --no-install-recommends tesseract-ocr
	installed=$(tesseract --version 2>&1 \| head -1 \| grep -oE '[0-9]+\.[0-9]+\.[0-9]+' \| head -1)
	expected_mm=$(printf '%s' "$TESSERACT_VERSION" \| cut -d. -f1-2)
	installed_mm=$(printf '%s' "$installed" \| cut -d. -f1-2)
	if [ "$installed_mm" != "$expected_mm" ]; then
	echo "::error::Tesseract major.minor mismatch: installed=$installed_mm expected=$expected_mm (TESSERACT_VERSION=$TESSERACT_VERSION in mise.toml). Bump the pin + re-baseline testdata/*.golden.json."
	exit 1
	fi
	echo "Tesseract $installed matches pinned major.minor $expected_mm"

	- name: Build + boot + fuzz + teardown via shared script
	# scripts/ci/check-api-drift.sh handles the build-serveronly →
	# boot-on-isolated-tempdir → schemathesis-run → teardown
	# sequence. Same script lefthook's pre-push schemathesis hook
	# invokes so the local + CI signals stay 1:1. The script
	# short-circuits with an actionable error if schemathesis or
	# frontend/dist are missing.
	run: SKIP_FRONTEND_BUILD=1 bash scripts/ci/check-api-drift.sh

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

refactor(app): migrate App.vue to Pinia (47 commits — App 2211→72 code lines, + store tests/debt) #1511

Workflow file

refactor(app): migrate App.vue to Pinia (47 commits — App 2211→72 code lines, + store tests/debt) #1511

Uh oh!

Workflow file for this run