Could vitest default to "run" instead of "watch" when run non-interactively?

[kentonv]

Currently, if you run vitest with no specific command, it defaults to watch mode: when tests finish, it stops and watches for changes, then re-runs them. You can press "q" to exit.

This is true even when standard input is not a TTY. So if you do:

vitest < /dev/null

It still stops and waits, but now there's no way to exit except to ctrl+C or otherwise kill the process.

If I opened a PR to change the behavior to default to "run" mode when standard input is not a TTY, would this be accepted?

(Or, alternatively solution: If standard input reaches EOF, can we consider that equivalent to pressing q?)

Here's my specific situation motivating this:

I've been experimenting with Claude Code, an AI assistant which is, uh, surprisingly good. I had written a new library, and I asked Claude to write tests for it. It wrote a beautiful test suite(!), added "scripts": {"test": "vitest"} to my package.json, and then tried to run the tests with npm run test. It then hung forever waiting for the tests to complete. Similarly, as I ask Claude to iterate on stuff, it frequently wants to run the tests using npm run test.

My work-around is to change package.json to use vitest run, but when I actually do run the tests interactively, I'd prefer the "watch" behavior. I can of course use npx vitest instead of npm run test when running interactively, but it seems like defaulting to "run" when run non-interactively would make sense, and would make vitest more friendly to AI assistants in general. People could of course still specify vitest watch explicitly even in non-interactive environments if that's what they really want.

[sheremet-va]

This makes a lot of sense to me. Feel free to open a PR, the team will discuss this in our next meeting

[kentonv]

Done: #7673