test(a2a): promote the A2A server batch to @stable (#1349)

[rafaelgiln]

Closes #1349.

What this does

Promotes the A2A server batch — 7 specs, 17 tests, shipped across #1242, #1244 and #1247 — to @stable, and flips the nine [-] bullets in QA-CHECKLIST.md §16.1 to [x].

Until now daily-stable.yml filters on --grep "@stable" and none of these tests carried it, so ten automated checklist bullets watched nothing. An A2A regression between two nightlies was invisible to every scheduled lane; the specs ran only when a PR happened to touch them (import graph) or someone dispatched manual.yml.

Why the tag was withheld, and why that no longer holds

Every spec doc gave the same reason:

No @stable yet: granted only after team validation (CONTRIBUTING.md). New in #1244, no daily history.

CONTRIBUTING.md says the opposite: "every new test enters with @stable" — in the PR itself, with the reviewer confirming the five validation steps on approval. Its four exceptions are inherited-not-yet-reviewed, temporarily-removed-while-failing, utility specs and @destructive. This batch is none of them.

"No daily history" is also circular — without the tag these specs never enter the daily, so they can never accumulate the history their own docs cite as the precondition.

The technical blocker is gone too: LANGFLOW_A2A_ENABLED=true has been on every lane and both start scripts since #1240, and the batch is LLM-free (Chat Input → Chat Output passthrough with a per-run sentinel — no provider, no key, no quota).

Evidence

npx playwright test --grep "@a2a" --repeat-each=3 --retries=0

Against langflowai/langflow-nightly:latest = 1.12.0.dev18 (four nightlies past the dev14 the docs recorded), LANGFLOW_A2A_ENABLED=true, LANGFLOW_WORKERS=1, default Playwright worker count.

Metric	Result
Executions	51 / 51 passed (17 × 3)
Retries	0 — the local default of 3 disabled so a flake could not hide
Wall clock	53.4 s
🚨 Backend Error:	0
Flows before → after	26 → 26 (nothing leaked)
Re-run after the tag conversion	17 / 17, 21.3 s, 0 backend errors

The tag conversion — the larger half of the diff

These specs declared their tags inside the test title. The rest of the suite uses an inline tag: array, and scripts/lib/stable-tests.ts — the shared parser behind the Phase 0 block and the checklist guard — counts only a tag array containing the literal "@stable".

Appending @stable to the title would have put these 17 tests in the daily while leaving them out of every generated count: exactly the split #985 exists to prevent. So the tags moved to tag: arrays and the titles are now clean, matching the convention (stable-tests.ts treats titles as the source of the Phase 0 bullet text).

The describe blocks lose their title tags for the same reason — all 7 title-tagged describes in the entire suite were A2A's.

Parser before → after: 448 → 465 @stable test() calls, zero warnings.

Checklist symbols

• Nine [-] → [x] in §16.1.
• The [~] Try it bullet stays [~]. The gap it names is a product absence, not a coverage gap: agentTab.viewExchange is a dead i18n key — 7 locales, zero call sites, absent from the DOM after a completed turn (Create: A2A Agent tab — publish flow and the Try it panel (§16.1 UI) #1244). The spec still earns the tag; only the bullet symbol is held by the partial.
• The eight [ ] bullets are untouched. §16.2 (A2A Client) needs the loopback-SSRF question answered first (follow-up 4 of the scoping doc); non-owner-publish 403 and the disabled-server state are already recorded as not automatable.

Per #741, the generated blocks are not regenerated here — update-coverage-summary.yml does that on merge. scripts/check-checklist-guard.mjs confirms no generated block was touched.

Docs

All 7 spec docs: the "No @stable yet" bullet is replaced with the tag plus the measured justification, and Last validated bumped 1.12.0.dev14 → 1.12.0.dev18. a2a-coverage-scope.md is deliberately left at dev14 — its Last validated refers to the surface map measurement, which this PR did not redo.

Gates run locally

Gate	Result
npm run typecheck	✅ 0 errors
npm run lint	✅ 0 errors (1319 pre-existing warnings, unchanged)
npm run test:units	✅ 502 pass
npm run test:scripts	✅ 795 pass
npm run check:checklist-coverage	✅ every @stable spec referenced by a Part II bullet
node scripts/check-checklist-guard.mjs	✅ no generated block edited
npx playwright test --grep "@a2a"	✅ 17/17

Risk to watch after merge

The 4 UI tests (agent-tab-*) are the only plausible flake surface: the Agent-tab save has no DOM completion signal (agent-status flips optimistically), so both specs wait on the PATCH /api/v1/flows/{id} the button fires, and agent-card-name is mount-time state that a save does not refresh (asserted after a reload). Both were measured in #1244 and are handled; 3/3 green here. Worth watching in the first week of daily history.

🤖 Generated with Claude Code