Scale

Local Scheduler

Run manuscript evaluation as a manageable local queue.

What it does

The scheduler layer can run batches manually, drain the queue once, or keep a background worker alive. Status and stop controls make it practical to process a larger project over time.

Why it matters

A serious manuscript has too many scenes for one-off prompting to be useful. The queue makes repeated, structured evaluation feel like local tooling instead of a pile of manual chores, while still evaluating each scene with its own bounded context.

  • Queue jobs from the writing environment.
  • Check status while the local worker runs.
  • Run configured built-in or custom metrics across many scenes.
  • Stop immediately or after the current job finishes.