Every `just deploy` now takes a pg_dumpall to
/opt/colectivo/backups/predeploy-<ts>-<sha>.sql.gz BEFORE rebuilding the
app or applying migrations, aborts the deploy if the dump is < 1 KiB,
and (on success) appends `<iso-ts> \t <sha> \t <backup-file>` to
/opt/colectivo/.deploys.log on ambrosio. Keeps the newest 10 backups
(prunes older predeploy-* files).
New `infra/scripts/rollback-erosi.sh` reads .deploys.log and pairs a
code rollback with a DB restore atomically:
just rollback-list # show recent deploys
just rollback # roll back to N-1
just rollback-to <sha> # roll back to a specific deploy
just rollback-code # roll back code only, keep current DB
Rollback safety:
- 5-second abort window.
- Verifies the target SHA exists locally + the backup is still on prod
(warns if it's been pruned past the 10-deploy window).
- Uses a temporary git worktree so the user's working tree isn't
disturbed.
- Stops app/auth/rest/realtime/storage before the gunzip|psql restore.
- Rebuilds the app image at the rolled-back SHA with the same GIT_SHA
build-arg path the deploy uses (so __APP_COMMIT__ in the bundle
matches the running code).
- Does NOT write a new .deploys.log entry — rollback is intentionally
not a deploy event; the next `just deploy` is from current HEAD.
- Storage volume (/var/lib/storage user uploads) is NOT rolled back.
Justfile `deploy` recipe repointed from the stale `deploy.sh` (which
referenced GHCR pull) to `deploy-erosi.sh` (the active prod path).
New project-scoped skill: `.claude/skills/deploy/SKILL.md` documents the
flow + safety boundaries for Claude-assisted invocations. `.gitignore`
keeps `.claude/settings.local.json` ignored but tracks `.claude/skills/`.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
12 new Playwright tests closing the UI coverage gap in the collective
lifecycle (onboarding → invitation → admin manage). Writing them surfaced
three product-code bugs that have silently been present since Fase 1;
fixed as part of this phase.
Tests:
- tests/e2e/onboarding.test.ts O-01..O-03 (Eva auto-redirect, happy
path create, empty-name submit-disabled)
- tests/e2e/invitation.test.ts I-01..I-05 (token generation, accept
logged-out + logged-in, expired, used)
- tests/e2e/manage-collective.test.ts MC-02..MC-05 (promote, demote, remove
member, generate pending invite)
- tests/fixtures/db.ts resetEva, seedExpiredInvitation,
seedUsedInvitation, restoreSeedMembership,
countMembership — all via raw SQL so they
don't depend on SUPABASE_SERVICE_ROLE_KEY
Bugs found & fixed:
- supabase/migrations/012_create_collective_rpc.sql: atomic
`create_collective(name, emoji)` SECURITY DEFINER RPC. Fase 1 onboarding
used `.insert(...).select().single()` which triggers the SELECT policy on
the fresh row — creator isn't a member yet, so the INSERT was rejected with
"row-level security" even though the INSERT policy passed. onboarding now
calls the RPC which inserts both the collective AND the creator-as-admin
membership in one transaction, bypassing RLS. F-08 in rls-isolation.test.ts
has been silently masking this bug (only the spoof branch was asserted).
- routes/+layout.svelte: post-login redirect now reads sessionStorage
`pendingInvitationToken` and routes back to /invitation/<token> instead of
defaulting to /onboarding or /lists. The invitation page already stashed
the token before kicking off Keycloak, but nothing read it back.
- routes/(app)/collective/manage/+page.svelte: loadMembers now subscribes
to currentCollective so it re-runs when the store resolves after cold
navigation. onMount alone races with the auth listener's first emit and
the member list stayed empty on direct URL hits.
- routes/invitation/[token]/+page.svelte: awaits authLoading before deciding
whether to redirect to login. Previously raced with the auth listener and
triggered a redundant Keycloak round-trip for freshly-authenticated users.
Stable selectors added (Playwright):
- Onboarding: onboarding-create-tab, onboarding-join-tab,
collective-name-input, collective-submit
- Manage: member-row-<uid>, role-select-<uid>, remove-member-<uid>,
generate-invite, invite-link
- Invitation: invitation-accept, invitation-error (+ data-error-key attr)
Scope dropped from the plan — UI does not exist, would be feature work:
- O-04: "+ new collective" affordance in the sidebar.
- MC-01: rename-collective input in /collective/manage.
Both flagged as follow-ups in plan/fase-7-collective-flow-tests.md.
Test totals: 34 pgTAP + 140 Vitest integration + 15 Vitest unit + 58
Playwright + 1 gated rate-limit. 3 skipped (2 Realtime presence, 1 gated).
The dev stack previously mixed two hostnames (`localhost` in some places,
`keycloak` in others). That only worked because the laptop's /etc/hosts
maps `keycloak` to 127.0.0.1. Phones, tablets, or a second laptop had no
such mapping, so the OAuth flow broke the moment you tried to log in from
any non-laptop device.
Switched the SPA, Kong, GoTrue, and Keycloak client to all agree on a
single external host — the laptop's LAN IP (192.168.1.167) — so the full
Keycloak → GoTrue → Supabase → SvelteKit round-trip works from any device
on the same Wi-Fi.
Changes
.env (root, ungitignored): PUBLIC_SUPABASE_URL / PUBLIC_KEYCLOAK_URL /
PUBLIC_APP_URL now use the LAN IP (updated locally; tracked via
apps/web/.env.development below).
apps/web/.env.development: mirror the LAN IP so `$env/static/public`
resolves the same on laptop and phone.
apps/web/vite.config.ts: `server.host: true` — bind Vite to 0.0.0.0.
apps/web/playwright.config.ts: `baseURL: PUBLIC_APP_URL ?? localhost`
so E2E uses the same origin the stack is configured with.
apps/web/tests/fixtures/login.ts:
- wait for the collective button in the sidebar to render before
returning (LAN-IP latency surfaced a pre-existing race in
`handleCreate` where Enter fired before `$currentCollective`
hydrated, silently no-op'ing).
- derive expected app origin from PUBLIC_APP_URL.
apps/web/tests/e2e/items.test.ts: scope D-04 delete assertion to the
`[role="listitem"]` locator — with undoQueue enabled the toast text
"Deleted <name>" also matches `getByText(itemName)` and shadowed the
original check.
infra/docker-compose.dev.yml:
- GOTRUE_EXTERNAL_KEYCLOAK_URL and REDIRECT_URI now read from
PUBLIC_KEYCLOAK_URL / PUBLIC_SUPABASE_URL so they follow the same
host as the rest of the stack.
- NEW: GOTRUE_URI_ALLOW_LIST with `/auth/callback` on both the LAN
IP and localhost. Without an explicit allow-list GoTrue rejected
`redirect_to=.../auth/callback` and fell back to SITE_URL root,
stranding `?code=` at `/` and re-triggering signIn → infinite loop.
keycloak/realm-export.json: `colectivo-web` client gains LAN-IP
redirectUris and webOrigins (alongside the existing localhost
entries). Persisted + live-applied via admin API.
.gitignore: add .claude/ (per-project scheduler runtime state).
Verification
just test-all → 224 passed, 4 skipped:
34 pgTAP
140 Vitest integration + 2 skipped (Realtime presence, upstream bug)
11 Vitest unit
39 Playwright + 2 skipped (mobile-swipe touch, WebKit-only)
Phone sanity check: open http://192.168.1.167:5173 on a LAN device,
log in as a seed user, full RLS-respecting session.
Caveats
LAN IP (192.168.1.167) is DHCP-assigned — if it rotates, rerun the
Keycloak admin API update (realm-export.json needs a new entry) and
update the three PUBLIC_* URLs. Consider a Tailscale magic-DNS name
as a stable replacement for the prod-deploy sprint.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add a 4-layer test stack covering RLS, triggers, and UI flows for Fases 0–2a,
then restructure every plan file so future fases start with tests and end with
a verification gate.
Test suite
- packages/test-utils: Vitest integration tests signing HS256 JWTs via jose so
each test acts as a specific seed user (createClientAs + createAdminClient)
- supabase/tests: pgTAP for accept_invitation(), item_frequency trigger, and
promote-on-admin-leave; each file self-installs pgtap extension
- apps/web/tests: Playwright E2E with live Keycloak login per test (storageState
caching doesn't rehydrate Supabase's session state reliably)
- just test-all chains the three suites; test-db forwards POSTGRES_PASSWORD
as PGPASSWORD with ON_ERROR_STOP=1 so failures abort the chain
Supabase auth gotcha
- PostgREST queries inside onAuthStateChange deadlock on GoTrue's navigator.locks
auth lock (getAccessToken → getSession → initializePromise waits on the same
lock that's held during event dispatch). Fix is two defenses: a pass-through
lock in $lib/supabase, and a setTimeout(0) defer in root +layout.svelte to
push loadUserCollectives out of the callback's microtask chain. Either alone
is insufficient; both together unblock the Playwright suite.
Env key rotation
- apps/web/.env.development had a stale demo anon key signed with a different
secret than root .env; Vite inlined that into the browser bundle so Kong (which
uses the root .env value) rejected every request with 401. Aligned the two
files and added a memory entry to flag this for the next rotation.
Plan restructure (TDD)
- Every fase now opens with X.0 Tests primero and closes with X.Z Verificación
final. Completed fases (0, 1, 2a) show the pattern retroactively with the
tests that currently cover them; pending fases (2b, 3, 4) list the tests to
write before implementation.
Docs
- CLAUDE.md status line reports 54 + 12 + 15 = 81 green tests, adds gotchas
#11 (auth-lock deadlock) and #12 (no storageState caching)
- README.md adds a TDD methodology section and the test-all command
- .gitignore excludes Playwright's generated reports and auth state
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
