Skip to content

Monitoring

The Monitor tab shows everything happening during a reconstruction job in real time.


Stage Stepper

At the top of the monitor, a horizontal stepper shows each pipeline stage with its current status:

StatusAppearance
PendingGray
RunningBlue with spinner
CompleteGreen check
FailedRed X
CachedGreen with cache icon
SkippedGray strikethrough
Timed outAmber clock
GatedAmber pause — waiting for user review

Click any completed stage step to see its duration and output files.


Session Starting Indicator

When a Python, Julia, or MATLAB stage starts for the first time in a job, a spinner banner appears while the interpreter initialises. Subsequent stages on the same backend are near-instant thanks to persistent sessions.


Interactive Gate Banner

When a stage of type interactive reaches the review point, an amber banner appears:

⏸ Awaiting mask review for Stage Name

Options:

  • Open Mask Editor — switch to the Mask Editor tab to review and approve the auto-generated mask.
  • Auto-approve — accept the auto-generated result without review.
  • Skip gate — bypass this stage entirely.

Convergence Charts

The chart area shows real-time metrics emitted by the running module:

  • Error norm — primary convergence metric (how far from the solution).
  • Penalty — regularisation penalty value.
  • Additional named metrics (e.g. step_size) if reported by the module.

Charts update live as new metrics events arrive. Hover over a point to see exact values. The x-axis is iteration number.


Resource Utilisation

Three progress bars show live system resource usage:

BarWhat it measures
CPUAverage across all logical cores
RAMPercentage of total installed RAM
GPUGPU utilisation (NVIDIA nvidia-smi or Metal)

Resource snapshots are taken at a configurable interval (default: 2 seconds, adjustable in Settings → Advanced).


Log Console

A scrolling console shows all log output from the running backend:

  • [info] — normal progress messages.
  • [warn] — non-fatal warnings.
  • [error] — errors (stage may still complete).
  • Raw stdout lines (if the module prints directly).

The console auto-scrolls to the bottom. Scroll up to pause auto-scroll. The last 2 000 lines are kept in memory.


Timeout

Each stage has a configurable timeout (default: 30 minutes, set in Settings → Pipeline). If a stage exceeds its timeout, a dialog appears offering:

  • Extend (+15 min) — add more time without cancelling.
  • Cancel — cancel the job.

Cancel

Click Cancel at any time to stop the job. A confirmation dialog appears. On cancel:

  • The current backend process receives SIGINT, then SIGKILL after 5 s if it doesn't exit.
  • The job is marked cancelled in History.

Error Detail

If a stage fails, a red error banner appears with a summary. Click Details (or click the status bar error indicator) to open the Error Detail Drawer — a slide-out panel showing:

  • Error message and stack trace.
  • Stage metadata (duration, exit code).
  • Output files written before the failure.
  • Log lines from the failed stage.

Stage Caching

If a stage's inputs haven't changed since a previous run, it may be served from cache (indicated by the cache icon in the stepper). Cached stages complete instantly. Caching can be disabled globally in Settings → Pipeline → Enable Stage Caching.

FORGE Studio