# Watching a Run

```bash
# Follow the newest run live
uv run python -m plesty.pow_pl.viz monitor

# Replay a stored run, two rows per refresh, and record the window
uv run python -m plesty.pow_pl.viz monitor --replay 2 --record run.mp4

# Render a stored run to <run-id>.mp4 offscreen — no window needed
uv run python -m plesty.pow_pl.viz render --run <run_id>
```

The manifest names this viewer (`viz: plesty.pow_pl.viz`), so the PLESTY
bench opens it on the runs it starts, in its Visualization pane or in a window
of its own. Everything around the views — which run, the share, live, replay,
render, record — is `plesty.lib.monitor.Viz`, the same for every experiment.

Four views, two columns:

| View | Shows |
|---|---|
| **Spectrum** | the newest spectrum, as recorded |
| **Spectral map** | every spectrum so far against the wave-plate angle, in counts per second (so the exposures of one angle compare) and without cosmic rays |
| **Excitation power** | the power read during each exposure, against the angle |
| **Integrated emission against power** | each spectrum's emission summed over `analysis.region_nm`, on log–log axes, with the running exponent *k* of $I \propto P^k$ |

The last view uses the same reduction as the analysis after the sweep, so its
*k* is the one the report gives for the integrated emission. A run without the
powermeter block has no power, and those two views stay empty.

Watching is passive: the views read the records the experiment already
committed and the spectra they point at, so a viewer can be opened, closed and
reopened mid-run without disturbing the measurement. Each spectrum file is read
once (`read_spectrum`) and shared by the views. The panel arrangement is
remembered for the next run.

`render` draws the same views on Qt's offscreen platform and encodes one video
next to the run directory, stopping when the last record is drawn; it works
headless, for any run.

Runs are looked for under `--run-root` (default `pow-pl` under `PLESTY_DATA`)
by every run-id generation (`pow_pl_*`, and `pol_pl_*` and `qd_fss_*` from
before the rename). A record names its spectrum relative to the shared disk,
which the viewer joins to this machine's `PLESTY_DATA`.

## The report

After the sweep, one analysis step per dot (`analysis.enabled`) saves
`analysis/<qd>.report.json` in the run directory. When a run started from the
bench ends, the bench compiles the run — configuration, instruments, this
report and the raw records — into `report.html` beside it; *Export HTML* writes
the same file elsewhere. A finished run is analysed again with
`python -m plesty.pow_pl.analysis <run_dir> <qd>`.