vitest-server

Boot once per session, then rerun cheaply. First run davstack check to confirm the daemon is running.

Once the daemon is up, drive it with the per-daemon CLI:

vitest-server run <file>           # warm rerun (~3-15s); JSON RunResult on stdout

Exit 0 = pass, 1 = fail. The full RunResult shape (per-test entries, error stacks, run-level errors) is on stdout — pipe to jq if you only want failing tests.

Don't restart the daemon between runs — the warm pool is the entire point.

If results look wrong ((0 test), suite errors, post-config-edit weirdness), the failure is usually in the daemon, not your code — check troubleshooting.md before iterating, or fall back to vitest run.

Reference

• README.md — overview, install, why
• docs/setup.md — config file, runtime matrix, primeFile semantics
• docs/usage.md — CLI verbs, HTTP API, RunResult shape, agent TDD loop
• docs/troubleshooting.md — the daemon-vs-code triage tree

CLI reference

vitest-server — Long-lived Vitest daemon + CLI client for fast story/unit reruns.

• vitest-server serve — Boot the long-lived Vitest daemon
• vitest-server run <file> — Rerun a file against the running daemon
• vitest-server refresh — Flush vitest transform cache + vite-node module cache and re-read config without restarting (keeps the warm vitest instance alive). Pass --hard for a full shutdown + detached re-serve when soft refresh is insufficient.
• vitest-server health — Daemon liveness check
• vitest-server shutdown — Stop the running daemon
• vitest-server doctor — Validate local install (node, peer dep, config, daemon liveness)

Run vitest-server <command> --help for the full flags and options of any command.