Modo "Por tema" no acervo de falas: expansão determinística do tema via
DeepSeek (generateObject, timeout, fallback null) alimenta o motor lexical
existente (searchText/keywords OR), mantendo facetas AND, paginação por fala,
o OR de cortes do C174 e o access control intacto. Resultado de tema mostra
selo e "Por que apareceu" com trecho real (frase contínua), sem score;
degradação honesta para termo exato com aviso e vazio honesto com saídas.
- URL: mode=tema (default omitido; canonicaliza sem q), seletor otimista
- ai/expandSpeechSearchTheme + lib/speechThemeTerms (sanitiza saída do modelo)
- view model com themeMatchTerm + espelho speechMatchesSearchTerm
- UI do card de tema (selos, proveniência, ações) e aviso degradado
- sem migration/Consent/exposição pública; gate de leitura inalterado
- testes unit/int/e2e; design hi-fi estendido e certificado pelo designer
Fase 1: playwright bare (--list, extensão VS Code, runs ad-hoc) quebrava na
coleta — next@15 sem exports não resolve 'next/cache' em ESM nativo e
server-only exige a condição react-server; o script test:e2e carregava ambas
via NODE_OPTIONS (por isso o CI passava). Hook de resolução ESM vitest-free
(e2eEsmResolve.mjs) registrada no main via module.register + propagada aos
workers via NODE_OPTIONS quando invocada bare; path do script intocado.
Fase 2: a sequência persistente campaign_fixture_municipality_alloc nunca é
resetada e um ciclo completo claima ~505 do catálogo de 435 — o wrap em todo
ciclo em DB de worktree fazia dois runs concorrentes receberem o mesmo índice
e purgarem as linhas vivas um do outro. Registry de claims compartilhado
vitest-free (campaignMunicipalityAllocator.ts): INSERT ON CONFLICT DO NOTHING
com skip de slot vivo, steal de claim stale (TTL 2h), release no cleanup dos
dois fixtures, erro claro de exaustão. Guard estático no
testMunicipalityAllocatorConventions + spec int de stress.
Board "Acompanhe de perto" (S1+S2) ganha o feed automático do perfil oficial:
loader irmão src/utilities/socialFeed/instagramFeed.ts (Graph API /media com
children p/ carrossel + username, unstable_cache tag social-feed + revalidate
300, snapshot cru pré-exclusão persistido via SQL direto, refresh de token
automático com retry 1x — refresh endpoint só renova tokens de Instagram
Login, page tokens caem no fail-closed), global SocialFeedSettings estendida
(instagramEnabled/AccessToken/UserId/MaxItems + snapshot hidden; excludedItems
já cobria a plataforma desde a S2). Requisito nº 1 — picker de exclusão no
admin (primeiro componente custom do repo, campo ui + importmap): lista os
posts do snapshot com thumbnail e toggle "Não exibir" que escreve em
excludedItems. Card com coverAspect square (1:1 do rascunho), badge
Instagram, título = legenda com fallback para post sem legenda, clique no
permalink (noopener), link "Seguir no Instagram →" derivado do username.
Os feeds externos mudam para o subfolder socialFeed/ (pin da S2 previa).
Migration 20260818_192948_add_instagram_social_feed_fields (não destrutiva).
Stub IG no webServer (porta dev+2000, helper socialStub.ts) + 3 e2e novos no
describe serial + kill switch estendido + smoke do picker no admin.
Closes#20
ci-pr.yml e ci.yml deixam de ter jobs irmãos: 1 job 'checks' com steps em
série (guards → lint → format → typecheck → knip → cycles → unit → int →
build → e2e). Step vermelho interrompe o job, o run falha e o slot é
liberado — sem API de cancel (Forgejo 9.0.3 não tem), sem watchdog, sem
upgrade do Forgejo. Skips do ci-scope.mjs viram step id 'scope'; e2e em 1
processo com PLAYWRIGHT_WORKERS=4 (substitui 2 shards × 2 workers); 2
services Postgres por run com portas host derivadas do run_id (runs
concorrentes não colidem); contexto 'CI (PR) / checks' preservado (job id
checks — contrato do automerger OPS61); deploy OPS53 gated em needs checks.
Plumbing morto removido (e2eShardConfig/e2e_shards). Docs: AGENT-OPS,
TESTING.md, changelog. Impl: docs/plans/ops62-ci-fail-fast-forgejo-impl.md
(a intenção watchdog+upgrade fica como histórico do gate).
Causa raiz (medida com o binário 1.58.2, config real): commander parseia
posicionais e -g/--list funcionam; o que arrastava a suíte era o closure de
dependências do dev mode (setup<-campaign<-frontend<-admin): filtrar um spec
do frontend puxava campaign(160)+setup(2). run-e2e-affected.mjs agora gera
test:e2e <passthrough verbatim> --no-deps -- <paths> via helpers puros em
scripts/lib/playwright-e2e-args.mjs (unit-pinned); passthrough corrigido
(pnpm consome o -- e o path era descartado); run-e2e-affected.mjs entrou em
HIGH_RISK_EXACT. Docs: receitas de filtro no playwright.config.ts, AGENTS.md,
changelog.
Board "Acompanhe de perto" (S1) ganha o feed automático do canal oficial:
global SocialFeedSettings (admin-only, kill switch + toggle + maxItems +
exclusões por item), loader src/utilities/youtubeFeed.ts (Data API v3
search→statistics, unstable_cache tag social-feed + revalidate 300,
fail-closed com snapshot persistido via SQL direto — revalidateTag dentro
de unstable_cache lança no Next 15.4), card generalizado
(CampaignArticleCard → CampaignContentCard, external com noopener) e merge
por recência com o grande = mais recente elegível. Stub HTTP do YouTube no
webServer do Playwright (porta dev+1000) + 4 e2e novos no describe serial.
Migration 20260818_111748_add_social_feed_settings (não destrutiva).
Closes#18
Bento 1+4 no desktop e carrossel 1/tela no mobile (~390px) lendo o post
collection com isPostVisible fail-closed; seção some inteira sem posts
visíveis. Zero migration/Consent; home vira ISR com tag posts +
revalidate=300 (label relativa não congela).
pnpm worktree next [--go] [--no-migrate] agora provisiona um ambiente
determinístico por worktree: slot = parte numérica do código da Issue
(C15→15; hash sem dígitos; bump em colisão entre worktrees vivos), dev
server em 3100+slot, bancos teqo_wt<slot> / teqo_wt<slot>_test no
container compartilhado (-p teqo) com a cadeia de migrations aplicada;
escreve .env.local + .env.test.local (gitignored, marcador gerado; envs
manuais nunca sobrescritos) e copia PAYLOAD_SECRET/Blob/VAPID do main.
kill remove o worktree e dropa os bancos gerados (WITH FORCE, best-effort).
Suites leem .env.test.local depois de .env.test (vitest.setup /
playwright / db-doctor / gate-ci espelham o layering) e
assertTestDatabase aceita teqo_test ou teqo_<worktree>_test em localhost
só (fail-closed preservado). db:start/db:stop fixam -p teqo (um
container de qualquer diretório); max_connections 200→400. Sem Docker
(Cursor Cloud): degrada ao compartilhado com aviso.
Derivação pura em scripts/lib/worktree-env.mjs (unit-tested); fluxo
validado end-to-end: provision → migrate → 598/598 int no banco do
worktree → kill sem resíduo.
Root cause of the prod-build React #130 on /campanha/municipios: the
md:hidden mobile-cards tree still hydrated ~25 rows x ~5 sheet controls
(~125 eager Drawer/DrawerCloseButton instances) even on wide viewports.
Filtered ?q=... routes failed more often only because they exercise the
list with filters active — the search param itself was not the bug.
CampaignListSheetProvider hosts a single bottom Drawer for the whole
mobile section; CampaignCellEditOverlay sheet cells portal their
body/footer into it while keeping per-cell open state and autosave
scaffolds. Footers with DrawerCloseButton are only created once a sheet
actually opens. Desktop popover path is unchanged; per-cell Drawer
remains the fallback for surfaces without the provider.
Also: DrawerCloseButton (function render prop, avoids base-ui element
render in prod bundles) replaces the DrawerClose render={<Button/>}
pattern across call sites; stale e2e selectors refreshed; e2e affected
script + playwright github/list reporters for CI; use-mobile reverted
(no consumers of the gate remain).
Guardrail filed as issue #52.
The first full e2e run on the parallel pipeline hit the 30-min timeout:
dev-mode cold webpack compiles per route, a __webpack_modules__ TypeError,
and dev-server memory-threshold restarts. CI now builds (NEXT_DIST_DIR
matches the webServer env) and playwright webServer runs `pnpm start`;
local spec development keeps `pnpm dev`.
Fold duplicated campanha access, loaders, list/filter/form shells, and scripts into shared modules; reorganize utilities by domain; stabilize e2e; and document prevention-first audit workflow so the same smells cannot creep back in.
E2E-driven fixes (the suite now passes 9/9; two journeys remain
latency-flaky on cold dev compiles and pass on retry):
- leader home crashed with Forbidden since the remodel:
loadLeaderContactsPageData loaded municipality options through the
staff-scoped loader (leaders are denied municipality reads); their scope
now derives from the engaged leadership's linked municipalities
- leaders saw 'Contato — —' instead of their supporters' names:
getAccessibleContactIds now mirrors canReadSupporter for leaders
(contacts of supporters they created)
- leader nav had two items with the same /campanha href (duplicate React
keys); the leader home IS the contact tool — one entry
- RSC serialization contracts: CampaignListPagination is a server component
again (hrefForPage stays server-side; CampaignTransitionAnchor client leaf
intercepts clicks with serializable props); CampaignSearchForm takes
basePath instead of a function prop
- hasAnyEstimate moved from the 'use client' PledgeEstimateForm to
voteEstimate.ts — server components may not call client-module exports
- playwright.config honors TEQO_TEST_DATABASE_URL (parity with
vitest.setup.ts) and uses a fixed .next/e2e dist dir so Next stops
appending per-PID entries to tsconfig include; report artifacts
eslint-ignored
Adds electoral nuclei, leaderships, updates, and WhatsApp invite flows behind campaign auth, plus the consolidated schema migration that Vercel applies on deploy.
Local dev and the test suite previously defaulted to the live production
Neon database, and running the tests mutated prod data. Add guardrails plus
a proper local Postgres + migrations workflow so production can't be hit by
mistake:
- docker-compose: local Postgres 17 (matches prod) + auto-created teqo_test
- dev guard refuses a non-local DATABASE_URL; tests refuse non-_test databases
- .env.test isolates the suite; vitest/playwright load it and fail closed
- Payload migrations set up (push disabled); build runs `payload migrate`,
so deploys apply pending migrations to prod automatically
- baseline `initial` migration + idempotent consent.text varchar->jsonb fix
- `pnpm db:pull` copies prod content locally, excluding supporter PII
- docs (README, AGENTS.md) + agent skills for the local-db/migration workflow
