deckhearth/.github/workflows/ci.yml
Randall Stillwell c210fd319a convoy: migrate CI to self-hosted axiom runners (briefs 1+2)
Moves 4 of 5 GitHub Actions workflows from `ubuntu-latest` to the new
`stwl-labs` org-level self-hosted pool (CT 111 axiom-runner-1..4) and
rewires the `migrate` job to use CT 102's shared Postgres via per-run
databases.

Changes:
- ci.yml: lint, schema-map-fresh, forbidden-patterns, migrate, test ->
  `[self-hosted, axiom]`. migrate job drops `services.postgres` (saved
  ~30s/run of image pull) and switches to `HOMELAB_CI_POSTGRES_BASE_URL`
  secret + per-run DB (`ci_run_<run_id>_<run_attempt>`) with `always()`
  cleanup so failed migrations don't leak DBs.
- preview-smoke.yml: gate + smoke -> self-hosted. Playwright browser
  cache lives under /opt/appdata/gha-runner/shared-cache/playwright on
  the host bind mount; first PR primes it, subsequent runs reuse.
- visual-diff.yml: gate + visual -> self-hosted (same Playwright cache).
- pr-health-rollup.yml: rollup -> self-hosted.
- agent-context-drift.yml: deliberately LEFT on ubuntu-latest (D4 in
  convoy doc). Weekly cron stays GitHub-hosted so it runs even when
  axiom is down.

Why on this side and not the runner side:
- migrate adds an explicit `sudo apt-get install -y postgresql-client`
  step (~10s, amortized via apt-cache survival). The runner image
  doesn't ship psql; baking it in would require a custom image and
  doesn't earn its keep for one job.

Repo prereqs (set before this PR opens):
- `HOMELAB_CI_POSTGRES_BASE_URL` repo secret set (value pattern:
  `postgres://deckhearth_ci:<pw>@192.168.68.102:5432`)
- `deckhearth_ci` Postgres user created on CT 102 with CREATEDB,
  no superuser
- stwl-labs org Actions settings: "Require approval for all outside
  collaborators" + runner group rejects public repos
- 4 runners online: `axiom-runner-1..4`, status Idle

Follow-ups (per convoy):
- Brief 3: forbidden-pattern gate to catch `runs-on: ubuntu-latest`
  re-introduction outside the agent-context-drift allowlist
- Brief 4: AGENTS.md updates + 1-line revert path (D5)
- Weekly cron on CT 102 to GC any `ci_run_*` DBs older than 7d
  (Risk #4 mitigation)

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-06-05 17:50:27 -05:00

429 lines
18 KiB
YAML

name: CI
# Vercel variant: Vercel builds Preview deployments on every push and gates the
# PR via the Vercel GitHub integration check. Running `npm run build` here too
# would duplicate Vercel's work for ~3-5 minutes per PR with no added signal.
#
# What this CI covers (and Vercel does not):
# - Lint (cheap belt-and-suspenders)
# - Schema-map drift check (docs/SCHEMA_MAP.md updated when scripts/add-*.js changes)
# - Migrations apply cleanly (node-pg-migrate against ephemeral Postgres 16)
# - Unit tests (vitest)
#
# NOTE: tcg-vault is JavaScript (not TypeScript). No `npx tsc --noEmit` step.
# Re-enable a type-check job if migrating to TypeScript.
on:
pull_request:
branches: [main]
# Skip CI entirely for doc-only PRs. The `slash-ci-minutes` convoy
# (2026-06-04) added this filter because every grep / lint / vitest /
# migrate job that fires on a pure-docs PR (e.g. `.convoys/` seeds +
# briefs, AGENTS.md edits, `docs/`) costs GitHub Actions minutes for
# zero signal. The full job set still runs on code-touching PRs.
paths-ignore:
- '.convoys/**'
- '**/*.md'
- 'docs/**'
- 'AGENTS.md'
- '.cursor/**'
- 'README.md'
push:
branches: [main]
paths-ignore:
- '.convoys/**'
- '**/*.md'
- 'docs/**'
- 'AGENTS.md'
- '.cursor/**'
- 'README.md'
concurrency:
group: ci-${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
env:
NODE_VERSION: '20'
jobs:
lint:
name: Lint
runs-on: [self-hosted, axiom]
steps:
- uses: actions/checkout@v4
- uses: actions/setup-node@v4
with:
node-version: ${{ env.NODE_VERSION }}
cache: npm
# Cache node_modules between runs (in addition to setup-node's npm
# cache). `setup-node@v4`'s `cache: npm` caches `~/.npm` but `npm
# ci` still does a fresh install into `node_modules` — that's the
# ~30-45s we want to skip on cache hit. Keyed by package-lock.json
# hash so any dep change invalidates correctly.
- name: Cache node_modules
id: cache-node-modules
uses: actions/cache@v4
with:
path: node_modules
key: node-modules-${{ runner.os }}-node${{ env.NODE_VERSION }}-${{ hashFiles('package-lock.json') }}
- run: npm ci
if: steps.cache-node-modules.outputs.cache-hit != 'true'
- run: npm run lint --if-present
schema-map-fresh:
name: Schema map up to date
runs-on: [self-hosted, axiom]
# Only run when migration scripts or the schema map itself changed.
# If neither changed, nothing to verify.
if: |
contains(github.event.pull_request.changed_files, 'scripts/add-') ||
contains(github.event.pull_request.changed_files, 'scripts/fix-') ||
contains(github.event.pull_request.changed_files, 'scripts/setup-neon-db.js') ||
contains(github.event.pull_request.changed_files, 'migrations/') ||
contains(github.event.pull_request.changed_files, 'docs/SCHEMA_MAP.md')
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 2
- name: Verify schema map updated alongside migration scripts
run: |
MIGRATION_CHANGED=false
MAP_CHANGED=false
if git diff --name-only HEAD~1 | grep -qE '^scripts/(add-|fix-|setup-neon-db\.js)'; then
MIGRATION_CHANGED=true
fi
if git diff --name-only HEAD~1 | grep -qE '^migrations/'; then
MIGRATION_CHANGED=true
fi
if git diff --name-only HEAD~1 | grep -q '^docs/SCHEMA_MAP\.md$'; then
MAP_CHANGED=true
fi
if [ "$MIGRATION_CHANGED" = "true" ] && [ "$MAP_CHANGED" = "false" ]; then
echo "::error::A migration script changed but docs/SCHEMA_MAP.md was not updated."
echo "Update docs/SCHEMA_MAP.md to reflect the schema change, then re-push."
exit 1
fi
echo "OK: schema map and migration scripts are in sync."
# Consolidated grep-based regression gates. Previously 6 separate jobs
# (forbidden-endpoints / -cors-headers / -client-side-llm-keys /
# -modal-shell-without-primitive / -deprecated-color-aliases /
# -stale-strings); the `slash-ci-minutes` convoy (2026-06-04) merged
# them into one job to amortize a single actions/checkout across all
# 6 checks. Each check still emits `::error file=...::msg` so the
# in-file annotations work the same way. The job uses a FAIL flag so
# ALL violations across all 6 checks are reported in a single run
# (matches the previous independent-jobs behavior; nicer than
# stop-at-first-failure).
forbidden-patterns:
name: Forbidden patterns (7 checks)
runs-on: [self-hosted, axiom]
steps:
- uses: actions/checkout@v4
- name: Run all forbidden-pattern checks
run: |
FAIL=0
# ---------- (1) No dev endpoints in pages/api ----------
echo "::group::Check 1/6: No dev endpoints in pages/api"
BAD_PATHS=(
"pages/api/simple.js"
"pages/api/test-auth.js"
"pages/api/test-db.js"
"pages/api/setup-database.js"
)
FOUND=()
for path in "${BAD_PATHS[@]}"; do
if [ -f "$path" ]; then
FOUND+=("$path")
fi
done
while IFS= read -r path; do
FOUND+=("$path")
done < <(find pages/api -maxdepth 4 -type f -name 'test-*.js' 2>/dev/null || true)
if [ ${#FOUND[@]} -gt 0 ]; then
echo "::error::Forbidden dev endpoints present in pages/api/. Delete them or move to scripts/."
for path in "${FOUND[@]}"; do
echo "::error file=${path}::Forbidden dev endpoint."
done
FAIL=1
else
echo "OK: no forbidden dev endpoints under pages/api/."
fi
echo "::endgroup::"
# ---------- (2) No wildcard CORS in pages/api ----------
echo "::group::Check 2/6: No wildcard CORS in pages/api"
MATCHES=$(grep -rEn 'Access-Control-Allow-(Origin|Methods|Headers)' pages/api/ 2>/dev/null || true)
if [ -n "$MATCHES" ]; then
echo "::error::Forbidden CORS headers present under pages/api/. Remove them — same-origin Vercel deployment does not need CORS."
echo "$MATCHES" | while IFS= read -r line; do
file=$(echo "$line" | cut -d: -f1)
lineno=$(echo "$line" | cut -d: -f2)
echo "::error file=${file},line=${lineno}::Forbidden CORS header — delete this line."
done
FAIL=1
else
echo "OK: no Access-Control-Allow-* headers under pages/api/."
fi
echo "::endgroup::"
# ---------- (3) No client-side LLM key leakage ----------
echo "::group::Check 3/6: No client-side LLM key leakage"
if [ -f pages/api/config/gemini.js ]; then
echo "::error file=pages/api/config/gemini.js::Forbidden config endpoint — do not return API keys to browsers."
FAIL=1
fi
CONFIG_MATCHES=$(grep -rEn 'apiKey:' pages/api/config/ 2>/dev/null || true)
if [ -n "$CONFIG_MATCHES" ]; then
echo "::error::Forbidden apiKey response under pages/api/config/."
echo "$CONFIG_MATCHES" | while IFS= read -r line; do
file=$(echo "$line" | cut -d: -f1)
lineno=$(echo "$line" | cut -d: -f2)
echo "::error file=${file},line=${lineno}::Do not return API keys from config endpoints."
done
FAIL=1
fi
AI_OCR_IMPORTS=$(grep -rEn 'from ['\''"].*ai-ocr|import.*ai-ocr' components/ 2>/dev/null || true)
if [ -n "$AI_OCR_IMPORTS" ]; then
echo "::error::Browser code must not import lib/ai-ocr — use POST /api/scan/identify instead."
echo "$AI_OCR_IMPORTS" | while IFS= read -r line; do
file=$(echo "$line" | cut -d: -f1)
lineno=$(echo "$line" | cut -d: -f2)
echo "::error file=${file},line=${lineno}::Remove ai-ocr import; call server-side scan API."
done
FAIL=1
fi
SERVER_ONLY=(
lib/scan-vision.js
)
LLM_PATTERN='generativelanguage\.googleapis\.com|api\.openai\.com|ai-gateway\.vercel\.sh'
LLM_FOUND=()
while IFS= read -r file; do
skip=false
for so in "${SERVER_ONLY[@]}"; do
if [ "$file" = "$so" ]; then
skip=true
break
fi
done
if [ "$skip" = true ]; then
continue
fi
if grep -qE "$LLM_PATTERN" "$file" 2>/dev/null; then
LLM_FOUND+=("$file")
fi
done < <(find components lib pages -name '*.js' ! -path 'pages/api/*' 2>/dev/null || true)
if [ ${#LLM_FOUND[@]} -gt 0 ]; then
echo "::error::Client-side LLM API URLs must not appear outside pages/api/."
for path in "${LLM_FOUND[@]}"; do
echo "::error file=${path}::Move LLM calls server-side or add to server-side-scan-pipeline removal list."
done
FAIL=1
fi
[ ${#LLM_FOUND[@]} -eq 0 ] && [ -z "$CONFIG_MATCHES" ] && [ -z "$AI_OCR_IMPORTS" ] && [ ! -f pages/api/config/gemini.js ] && echo "OK: no client-side LLM key leakage patterns detected."
echo "::endgroup::"
# ---------- (4) No legacy modal shells ----------
echo "::group::Check 4/6: No new modal shells without <Modal> primitive"
MODAL_FOUND=()
while IFS= read -r file; do
MODAL_FOUND+=("$file")
done < <(grep -lE 'fixed inset-0 bg-black bg-opacity-' \
pages components -r --include='*.js' 2>/dev/null \
| sort -u || true)
if [ ${#MODAL_FOUND[@]} -gt 0 ]; then
echo "::error::Legacy modal-shell pattern detected. Use the <Modal> primitive from components/ui/."
for f in "${MODAL_FOUND[@]}"; do
echo "::error file=${f}::Replace 'fixed inset-0 bg-black bg-opacity-' with <Modal open={…} onClose={…} title=…> from components/ui."
done
FAIL=1
else
echo "OK: no legacy modal shells anywhere in pages/ or components/."
fi
echo "::endgroup::"
# ---------- (5) No deprecated color aliases ----------
echo "::group::Check 5/6: No use of deprecated color aliases"
PATTERNS=(
'gradient-text-purple'
'gradient-text-pink'
'gradient-text-blue'
'glow-purple'
'glow-pink'
'glow-blue'
'gradient-bg-purple'
'gradient-bg-blue'
'gradient-bg-pink'
)
ALL_MATCHES=""
for pattern in "${PATTERNS[@]}"; do
P_MATCHES=$(grep -rFn "$pattern" pages components --include='*.js' 2>/dev/null || true)
if [ -n "$P_MATCHES" ]; then
ALL_MATCHES=$(printf '%s\n%s' "$ALL_MATCHES" "$P_MATCHES")
fi
done
if [ -n "$ALL_MATCHES" ]; then
echo "::error::Deprecated color alias detected. Use --accent-ember / --accent-flame / --accent-gold (or the .gradient-bg-ember / .gradient-text-ember utility classes)."
echo "$ALL_MATCHES" | sort -u | while IFS= read -r line; do
[ -z "$line" ] && continue
file=$(echo "$line" | cut -d: -f1)
lineno=$(echo "$line" | cut -d: -f2)
echo "::error file=${file},line=${lineno}::Deprecated color alias — replace with the canonical Deck Hearth warm-palette equivalent."
done
FAIL=1
else
echo "OK: no deprecated color aliases in pages/ or components/."
fi
echo "::endgroup::"
# ---------- (6) No stale ownership/collection copy ----------
echo "::group::Check 6/6: No stale ownership/collection copy"
STRING_PATTERNS=(
'Mark Owned'
'Owned Cards'
'All My Cards'
)
STRING_FOUND=()
for pattern in "${STRING_PATTERNS[@]}"; do
while IFS= read -r line; do
STRING_FOUND+=("$line")
done < <(grep -rFn "$pattern" pages/ components/ \
--include='*.js' \
--exclude-dir=api 2>/dev/null || true)
done
if [ ${#STRING_FOUND[@]} -gt 0 ]; then
echo "::error::Forbidden stale UI string(s) in pages/ or components/. Use lib/collection-vocabulary.js labels — see AGENTS.md § Product vocabulary."
printf '%s\n' "${STRING_FOUND[@]}" | sort -u | while IFS= read -r line; do
file=$(echo "$line" | cut -d: -f1)
lineno=$(echo "$line" | cut -d: -f2)
text=$(echo "$line" | cut -d: -f3-)
echo "::error file=${file},line=${lineno}::${text}"
done
FAIL=1
else
echo "OK: no forbidden stale strings in pages/ or components/."
fi
echo "::endgroup::"
# ---------- (7) No bespoke var(--glass-surface-*) inline styles ----------
echo "::group::Check 7/7: No bespoke var(--glass-surface-*) inline styles"
# unify-glass-panel-surfaces convoy, Brief 7 (PR sequence after #126).
# After Briefs 1-6 land, every panel-shaped surface in the app
# composes via .glass-panel / .glass-panel-strong /
# .page-header-glass / <GlassSurface>. Bespoke inline-style use
# of var(--glass-surface-low|mid|high) under pages/ or
# components/ JSX is the regression vector this gate prevents.
#
# Allowlist (explicit, documented exceptions):
# --- Chrome blocks (D4 of architect plan, ratified) ---
# - components/ui/GlassSurface.js # the primitive itself
# - components/Layout.js # sidebar nav-chip chrome (L850 area)
# - components/ui/TopSearchBar.js # <header> chrome
#
# The pending-migration entries (Button / Input /
# MobileNavigation) were removed by the
# migrate-button-input-mobilenav-to-glass-primitive convoy
# (2026-06-05). Adding a fourth chrome-tier entry is a
# design-system decision — open a new convoy.
GLASS_ALLOWLIST=(
"components/ui/GlassSurface.js"
"components/Layout.js"
"components/ui/TopSearchBar.js"
)
GLASS_FOUND=()
while IFS= read -r file; do
skip=false
for allowed in "${GLASS_ALLOWLIST[@]}"; do
if [ "$file" = "$allowed" ]; then
skip=true
break
fi
done
if [ "$skip" = true ]; then
continue
fi
GLASS_FOUND+=("$file")
done < <(grep -lE "var\\(--glass-surface-(low|mid|high)\\)" \
pages components -r --include='*.js' 2>/dev/null \
| sort -u || true)
if [ ${#GLASS_FOUND[@]} -gt 0 ]; then
echo "::error::Bespoke var(--glass-surface-*) inline styles detected outside the documented allowlist."
echo "Use .glass-panel / .glass-panel-strong / .page-header-glass or compose <GlassSurface> from components/ui/ instead."
for f in "${GLASS_FOUND[@]}"; do
echo "::error file=${f}::Replace inline var(--glass-surface-*) with the appropriate class or primitive."
done
FAIL=1
else
echo "OK: no bespoke glass-surface inline styles outside the allowlist."
fi
echo "::endgroup::"
# ---------- Final exit ----------
if [ "$FAIL" -ne 0 ]; then
echo "::error::One or more forbidden-pattern checks failed. See annotations above."
exit 1
fi
echo "All 7 forbidden-pattern checks passed."
migrate:
name: Migrations apply (node-pg-migrate)
runs-on: [self-hosted, axiom]
# Postgres lives on CT 102 (axiom homelab). Per-run database keeps parallel
# PRs isolated and a `DROP DATABASE` in `always()` keeps the catalog tidy.
# `HOMELAB_CI_POSTGRES_BASE_URL` is a repo secret of the shape
# `postgres://deckhearth_ci:<pw>@192.168.68.102:5432` — no DB name. The
# `deckhearth_ci` Postgres role has CREATEDB but no superuser; a
# compromised runner can't reach other apps' databases.
env:
PGBASE: ${{ secrets.HOMELAB_CI_POSTGRES_BASE_URL }}
DBNAME: ci_run_${{ github.run_id }}_${{ github.run_attempt }}
steps:
- uses: actions/checkout@v4
- uses: actions/setup-node@v4
with:
node-version: ${{ env.NODE_VERSION }}
cache: npm
- name: Install postgresql-client
# myoung34/github-runner image doesn't ship psql. ~10s on first
# apt-update; the underlying apt cache survives between runs because
# the ephemeral runner container is recreated but layers are cached
# by the CT's docker daemon.
run: sudo apt-get update -qq && sudo apt-get install -y -qq postgresql-client
- name: Cache node_modules
id: cache-node-modules
uses: actions/cache@v4
with:
path: node_modules
key: node-modules-${{ runner.os }}-node${{ env.NODE_VERSION }}-${{ hashFiles('package-lock.json') }}
- run: npm ci
if: steps.cache-node-modules.outputs.cache-hit != 'true'
- name: Create per-run database
run: |
psql "$PGBASE/postgres" -c "CREATE DATABASE \"$DBNAME\";"
echo "POSTGRES_URL=$PGBASE/$DBNAME" >> .env.local
- run: npm run migrate up
- name: Drop per-run database
if: always()
run: psql "$PGBASE/postgres" -c "DROP DATABASE IF EXISTS \"$DBNAME\";"
test:
name: Unit tests (vitest)
runs-on: [self-hosted, axiom]
steps:
- uses: actions/checkout@v4
- uses: actions/setup-node@v4
with:
node-version: ${{ env.NODE_VERSION }}
cache: npm
- name: Cache node_modules
id: cache-node-modules
uses: actions/cache@v4
with:
path: node_modules
key: node-modules-${{ runner.os }}-node${{ env.NODE_VERSION }}-${{ hashFiles('package-lock.json') }}
- run: npm ci
if: steps.cache-node-modules.outputs.cache-hit != 'true'
- run: npm run test:run
env:
JWT_SECRET: ci-secret-only-for-tests-do-not-use-in-prod