Fix: name the webui_test hook timeouts and give the teardown an explicit budget

[AmaadMartin]

Please ensure you have read the contribution guide before creating a pull request.

Link to Issue or Description of Change

1. Link to an existing issue (if applicable):
   Closes: #issue_number
   Related: #issue_number
2. Or, if no issue exists, describe the change:

Problem: tests/integration/adk_web/webui_test.ts has two timeout defects in its server-lifecycle hooks.

• beforeAll ends }, 20000);. The literal is 40 lines below the thing it budgets and carries no explanation of what 20 seconds pays for (booting a server — for the CLI-backed case by spawning the ADK CLI as a child process and waiting for it to listen).
• afterAll passes no timeout at all, so its budget is whatever the runner supplies rather than a value this suite chose. stop() for AdkApiServer is http.Server.close() (dev/src/server/adk_api_server.ts:966-985), whose callback fires only once every open connection has drained — close() does not force keep-alive sockets shut. A connection that never closes therefore does not fail fast; it parks the run for the full inherited budget and reports with a generic hook-timeout message.

This went unnoticed because the file looks like it has a timeout policy: the outer describe passes 20000 as its third argument. That argument is a suite test timeout — Vitest applies it to the it() bodies and it never reaches hooks. Measured, by setting it to 1: the failure is Test timed out in 1ms, and the hooks run unaffected.

The inherited teardown budget is about to get much worse. #548 ("set project-wide integration hookTimeout/testTimeout in vitest.config.ts") is merged upstream and adds hookTimeout: 120000 to the integration project that owns this file. That value is sized for beforeAll hooks that run npm install per fixture (build_setup_test.ts:29, app_loader_test.ts:31,78, agent_dirname_test.ts:28, skills/script_js/agent_test.ts:33); it is a nonsensical ceiling for a socket close.

Solution: declare the two hook budgets as named, commented module-level constants and pass both to the hooks they govern.

const SERVER_START_TIMEOUT = 20000; // was a bare `20000` on beforeAll
const SERVER_STOP_TIMEOUT = 10000; // afterAll had no argument at all

SERVER_START_TIMEOUT keeps beforeAll's existing value — this names it, it does not retune it. SERVER_STOP_TIMEOUT pins the 10s the untimed hook effectively had, so the passing path and the failure path are both unchanged on this base, while the budget becomes a property of what the teardown does instead of a project-wide knob tuned for other suites' work.

Nothing else changes: no assertion, no test name, no HTTP behaviour, no dependency, no export, and vitest.config.ts is untouched.

Two deliberate deviations from the task spec, both because a spec claim did not survive verification:

1. The spec said afterAll currently inherits 120000ms. It does not — on this branch's base it inherits 10000ms. The base (main @ 1210acc7) is 23 commits behind upstream and predates Chore: enforce the node: protocol for Node built-in imports in src (no new deps) #548, so vitest.config.ts here sets no hookTimeout and Vitest's 10s node default applies. Verified by injecting a never-settling await into afterAll on the unmodified file: Hook timed out in 10000ms. SERVER_STOP_TIMEOUT is therefore 10000, not the 20000 the spec proposed — 20000 would have loosened teardown 2x against this base while the description claimed a 6x tightening. 10000 is the zero-regression value on the base and still a 12x tightening once Chore: enforce the node: protocol for Node built-in imports in src (no new deps) #548 arrives.
2. The suite-level 20000 on describe is intentionally left as a literal. The spec put naming it in scope on the grounds that it "costs one line and zero risk". It costs neither: Prettier hugs a test callback only when the third argument is a numeric literal, so substituting an identifier de-hugs the call and re-indents the entire 60-line suite body. Measured — that one substitution took the diff from 16 lines to 153 (90 insertions / 63 deletions, of which git diff -w showed only 31 were semantic). A whole-file re-indent that destroys git blame is not worth naming a number that already sits on the construct it governs, and it is a test timeout, not one of the hook timeouts this change is about.

Collision check (required before implementation): gh pr list --repo AmaadMartin/adk-js --state open --limit 100 returned 100 open PRs; none names this file. gh pr diff --name-only on every plausibly adjacent PR (#254, #256, #257, #260, #261, #276, #299, #305, #311, #324, #343, #349 — the timeout, vitest-config and tests/integration PRs) confirms no open PR touches tests/integration/adk_web/webui_test.ts. A PR search for "webui" and a git ls-remote scan for webui/adk_web branches on the fork both came back empty. No collision, nothing to stack on.

Testing Plan

Please describe the tests that you ran to verify your changes. This is required for all PRs that are not small documentation or typo fixes.

Unit Tests:
[x] I have added or updated unit tests for my change.
[x] All unit tests pass locally.

No new test case was added, deliberately. The change adds no core/src, dev/src or integrations/src line, so it contributes nothing to the coverage denominator (vitest.config.ts coverage.include), and a test asserting that a constant equals 10000 would restate the source and protect nothing. The existing suite is the test, and the verification burden is proving the constants are actually wired to the hooks — a rename that silently failed to bind would look identical in review. No existing test was modified, skipped, weakened, or deleted.

npx vitest run --project integration tests/integration/adk_web/webui_test.ts

Test Files 1 passed (1) | Tests 2 passed (2) — both parameterised cases (Run from ADK CLI, Using ADK API server). Suite time 2610ms against a pre-change baseline of 2615ms, confirming the explicit teardown budget did not clip a legitimately slow stop().

Mutation proofs. A timeout change is invisible on the happy path, so each binding was proven on the failure path. Every mutation was reverted and git diff re-checked clean afterwards.

#	Mutation	Result	Proves
1	SERVER_STOP_TIMEOUT → 1500, never-settling await appended to afterAll	Error: Hook timed out in 1500ms.	The constant reaches afterAll. Before this change the identical hang reported 10000ms no matter what the file declared.
2	SERVER_START_TIMEOUT → 1	Error: Hook timed out in 1ms. (tests skipped)	The constant reaches beforeAll and is not merely declared.
3	Suite timeout → 1	Error: Test timed out in 1ms. (hooks unaffected)	The describe argument governs tests, not hooks — the asymmetry that let the untimed teardown hide.
4	hookTimeout: 120000 added to the integration project, never-settling await in afterAll, with this change	Error: Hook timed out in 10000ms. — wall time 29.15s	The explicit budget overrides a project-level hookTimeout.
5	Same config, without this change (untimed afterAll)	Error: Hook timed out in 120000ms. — wall time 248.78s	The regression this prevents, measured: 4m10s of stall and a budget chosen by a config block written for npm install.

Mutations 4 and 5 are the before/after pair for the state of the tree once #548 lands here: 248.78s → 29.15s, and a hook-timeout message reporting a number this file chose.

Manual End-to-End (E2E) Tests:

Full validation gate, run on the exact pushed commit, mirroring .github/workflows/validation.yaml:

npm run lint          # exit 0
npm run format:check  # exit 0 — "All matched files use Prettier code style!"
npm run build         # exit 0
npx vitest run --project integration tests/integration/adk_web/webui_test.ts   # 2 passed

npm run ts:check reports 308 errors in 48 files both with and without this change (verified by stashing the diff and re-running). All are pre-existing on main and none is in webui_test.ts; this change adds zero. The full repo suite was not run — this change cannot affect any file other than the one it edits.

To reproduce the teardown behaviour manually: append await new Promise(() => {}); to the afterAll body and run the command above. It now fails in ~10s per case with Hook timed out in 10000ms. Revert the injected line afterwards.

Checklist

[x] I have read the CONTRIBUTING.md document.
[x] I have performed a self-review of my own code.
[x] I have commented my code, particularly in hard-to-understand areas.
[x] I have added tests that prove my fix is effective or that my feature works.
[x] New and existing unit tests pass locally with my changes.

No suppressions of any kind were added: git diff main -U0 matches no @ts-expect-error, @ts-ignore, eslint-disable, as any, as never, or coverage-ignore pragma. No console.log. One file changed; no lockfile, package.json, or CHANGELOG.md churn.