Measurements #

URL: /explore

A chart-based view of every measurement Litmus has recorded — filter, group, and plot. The table-based Results detail Measurements tab is scoped to one run; this view spans the whole index, lets you filter on any facet, and renders the result as a chart instead of a table.

Use it to ask questions like "what did output_voltage look like across all production runs on bench-3 last week?", "did this measurement drift when we changed lots?", or "how is the limit margin distributed for iq across all DUTs?".

Filter section #

A row of facet widgets, each populated from values actually present in the index. Changing any facet re-fetches all the others' options (so the Product list only shows products that exist under your other current filters).

Facet	Filters by	Widget
Run outcome	The run's overall outcome (passed, failed, ...)	Multi-select enum
Measurement outcome	The measurement row's own outcome	Multi-select enum
Limit comparator	How the measurement is checked (GELE, LE, ...)	Multi-select enum
Product	DUT part number	Multi-select
Station	Station hostname or identifier	Multi-select
Test phase	The free-form test_phase value stamped on the run. The conventional values are validation, production, and characterization, but any string a project chooses to use will appear here.	Multi-select
Step	The test/step name the measurement belongs to	Multi-select
Measurement	The named measurement (vout, iq, ...)	Multi-select
DUT serial	A specific unit	Multi-select
Date range	Run start date Since / Until	Date range

A count below the row tells you how many measurement rows the current filter set matches.

Before any measurements are recorded, the page shows a quick-start card pointing at the verify and logger.measure calls that populate the index.

Plot controls #

The PLOT card sits above the chart and decides what gets drawn.

Control	What it does
Y axis	Which numeric column to plot vertically. Required. Both fixed schema columns (measurement_value, run_started_at, ...) and dynamic input/output columns (in_*, out_* recorded by context.configure() and context.observe()) appear in the dropdown.
X axis	Which numeric column to plot horizontally. Same column choices as Y. Required for scatter / line / bar; ignored for histogram.
Chart	One of scatter (default), line, bar, histogram.
Group by	Optional. Plots one series per distinct value of the chosen column — e.g. group by dut_serial to see one line per DUT.
Bins	Histogram bin count. 2–200, default 30.
Limit	Maximum row count fetched. 10–100,000, default 5,000. The chart only ever shows up to this many rows.
Refresh	Force a re-fetch. Filter and control changes already auto-refresh; use this to pick up new runs that arrived while the page was open.

When Y is unset (or X is unset for non-histogram charts), the chart area shows a placeholder until you pick a column.

Chart #

Interactive chart of the filtered rows. The chart kind matches the PLOT control:

• Scatter — one point per measurement row at (x, y)
• Line — points sorted by X and connected
• Bar — bars at each X with height Y
• Histogram — distribution of the Y value alone (X is ignored), bucketed by Bins

When the result set exceeds Limit, only the first Limit rows are returned (ordered by run id, not by date). Tighten the filter set or raise Limit if the chart looks truncated; very large limits make the chart slow to render.

Live updates #

The view watches the event log for run.ended events and re-fetches the active chart when a run finishes. No manual reload needed.

Bookmarkable URL state #

Every facet, axis, and chart setting lives in the URL, so a URL captures the exact view:

Parameter	Meaning
Per-facet	Each multi-value facet repeats its column name as a query key (e.g. ?product=PN-100&product=PN-200)
since, until	Date range, YYYY-MM-DD (omitted when blank)
y, x	Selected axes (omitted when blank, and x is omitted on histogram charts where it's ignored)
chart_type	scatter (default — omitted), line, bar, or histogram
group_by	Selected group-by column (omitted when blank)
bins	Histogram bin count (omitted at default 30)
limit	Max row count fetched (omitted at default 5,000)

Defaults are stripped from the URL to keep it short.

Underlying data #

The view reads from the same measurements index the Results detail page reads from, just unfiltered across all runs. The CLI equivalents:

• litmus measurements ... for parametric queries (when wired)
• The Python query class MeasurementsQuery is what the page calls under the hood

Each measurement row carries the values shown in vout, iq, etc. — see Parquet schema for the column list.

Common tasks #

• Distribution of one measurement — set the Measurement facet to one value, Y to measurement_value, Chart to histogram, and read the spread. Limit-bearing measurements show their pass/fail via the Measurement outcome facet.
• One measurement across DUTs — set the Measurement facet, X to run_started_at, Y to measurement_value, Group by dut_serial. One line per unit, time-series view.
• Compare lots — Group by something derived from dut_serial, or stack Lot + Date Range filters and re-query.

See also #

• Results detail → Measurements tab — per-run measurement view
• Metrics → Cpk — process capability ranking across the same data
• Parquet schema — the measurement row columns the facets and axes pull from