Get the status of an NLME job

The lifecycle field is state (not status): one of running, succeeded, failed, dead, unknown, or not_found. The artifacts directory is artifact_dir (not artifacts_dir) and equals file.path(run_dir, "artifacts"). This is a single-shot progress snapshot - it reads the latest status once and returns immediately:

Usage

get_nlme_job_status(
  job_id,
  project_dir = ".",
  wait_seconds = 0,
  poll_interval = 2
)

Arguments

job_id

Job id from start_nlme_job().

project_dir

Project root.

wait_seconds

Optional server-side wait: block (polling every poll_interval seconds) until the job reaches a terminal state (succeeded/failed/dead/not_found) or this many seconds elapse, then return the latest status. Default 0 (single-shot, non-blocking). Capped at 600 so a tool handler never blocks indefinitely. For agent use prefer wait_for_nlme_job() instead of tuning this directly.

poll_interval

Seconds between internal polls while waiting (default 2, minimum 0.5). Ignored when wait_seconds <= 0.

Value

A list with job_id, state, timing (started/finished/ exit_code), pid, run_dir, artifact_dir, live progress, and (for a terminal job) fit_health. This is the raw snapshot primitive; the agent-facing next_action directive is added by wait_for_nlme_job().

Details

st <- get_nlme_job_status(job_id, project_dir)
artifact_dir <- st$artifact_dir

To watch a long fit without burning agent turns, prefer wait_for_nlme_job(), which blocks server-side using the launch-configured watch budget. The wait_seconds/poll_interval arguments here are retained for backward compatibility and advanced manual R use; wait_for_nlme_job() is the agent-facing watcher.

See also

wait_for_nlme_job() for the preferred long-job watcher.