API Status workflow occasionally hangs and times out

[JstnMcBrd]

About once every week or two, the API Status action will fail because vitest fails to terminate after running api.test.ts. The test passes successfully, but vitest does not exit, so the workflow eventually times out with an error after 6 hours. This causes the badge in the README to say "API Status: Failing".

Here is an example run.

Run npm run test-api

> delphi-ai@2.0.0 test-api
> vitest api

 RUN  v1.0.1 /home/runner/work/delphi-ai/delphi-ai

 ✓ test/api.test.ts  (1 test) 657ms

 Test Files  1 passed (1)
      Tests  1 passed (1)
   Start at  00:38:52
   Duration  11.03s (transform 41ms, setup 0ms, collect 29ms, tests 657ms, environment 0ms, prepare 71ms)

close timed out after 10000ms
Failed to terminate worker while running /home/runner/work/delphi-ai/delphi-ai/test/api.test.ts.
Tests closed successfully but something prevents Vite server from exiting
You can try to identify the cause by enabling "hanging-process" reporter. See https://vitest.dev/config/#reporters
Error: The operation was canceled.

After some digging, I found this comment in vitest#2008, which suggested that node's fetch API is the culprit. Tests that use node's fetch or undici will occasionally hang. The root issue is nodejs/undici#2026.

This is currently being investigated, so while I wait for the issue to be fixed, my only workaround is to simply re-run the workflows when they fail.

Duplicate issue with JstnMcBrd/dectalk-tts#9

[JstnMcBrd]

This issue will (probably) be fixed in vitest v2, as per vitest-dev/vitest#2008 (comment). Waiting on vitest-dev/vitest#5047 to be merged.

Once that update is released, the --pool=forks workaround can be removed.