Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS
JavaScript ai-agentsdashboarddeveloper-toolsdigital-twinquantified-selfsecond-brainself-hostedtailscale
/**
* The client suite's time budgets, defined in ONE place so the async bound and
* the budgets enclosing it can never be edited apart.
*
* `ASYNC_UTIL_TIMEOUT_MS` bounds every Testing Library async utility —
* `waitFor`, `findBy*`, and the `settledInput.js` helpers built on them.
* `TEST_TIMEOUT_MS` bounds the whole test, and the same value bounds a hook.
* The ORDER matters more than any of the numbers: an inner bound that reaches
* its enclosing budget can never report its own failure, because the test dies
* first with a bare "test timed out" naming nothing. That trap has been hit
* twice here — `WordplayTrainer`'s 5s drill bound against vitest's 5s default,
* and `BeeperTab`'s `{ timeout: 15000 }`, which could never wait more than 5s
* and so never helped anything.
*
* 5000ms async: Testing Library defaults to 1000ms; this suite ran at 3000
* (#3474) because several views debounce at 500ms and some wait on a debounce
* plus a retry. `vitest.config.js` already documented 3s as marginal on the
* 2-vCPU public runner — that comment is why the client worker cap is 2 — and
* the `await`-a-mock-call assertions that went red there passed locally every
* time (#7448). A loaded runner should make a test SLOWER, not red.
*
* 15000ms enclosing, as 3x: headroom, not latency. It is not a guarantee that
* the async bound always gets to speak — three sequential near-budget waits
* still exhaust it — but it is comfortably clear of the one-wait case that
* actually bites, and a real hang still reports in ~5s from the async utility,
* naming what it was waiting on.
*
* Raising the async budget is NOT a substitute for fixing an ordering race: it
* buys a slow runner room, and cannot rescue an assertion waiting on a call
* that already happened with the wrong argument. See `settledInput.js` and
* `pageLoadBarrier.js` for that half.
*/
export const ASYNC_UTIL_TIMEOUT_MS = 5000;
export const TEST_TIMEOUT_MS = ASYNC_UTIL_TIMEOUT_MS * 3;