## [0.10.0] - 2026-06-16

- **Map diagrams can normalise their fields.** A new *Normalize*
control rewrites each diagram field as an expression so fields on very
different scales become comparable on every feature, instead of one
big-number field swamping the rest. **Min–max (0–1)** is the safe
default for pie/bar; **Z-score** standardises (great for bars; can be
negative); **Log** compresses heavy tails. Each field's
min/max/mean/std are computed from the (optionally selected) features
and baked into the diagram's category expressions — **nothing is
written to your data**. A live hint warns when Z-score is paired with
a pie/stacked diagram (which can't draw negative angles).
- **Labels are now expression-driven.** Beyond a single field you can
**round** numbers, add a **thousands separator**, stack a **second
field on its own line** (a real newline via `char(10)`), add a
**prefix/suffix** or units, change **case**, **word-wrap**, or type
**any QGIS expression** in an advanced box that overrides the
controls. A live **preview** shows the exact expression and a sample
from the first feature, and numeric formatting is applied **only to
numeric fields**, so a name on a second line is never wrapped in
`round()`. Built on `QgsPalLayerSettings` expressions
(`core/expressions.py`).
- **A built-in, offline user guide (❔ Guide).** A designed,
self-contained HTML guide covers every surface — engines and the 17
chart types, animation, Explore, map-diagram **normalisation** (with
worked examples), **copy-ready label-expression recipes**, the smart
assistant, troubleshooting — plus **copy-ready prompts for an external
AI assistant** (ChatGPT/Claude) for those who want extra help. No
JavaScript, so it renders on every QGIS web backend and exports/opens
anywhere. When opened from the dock it includes a personalised *“For
your current layer”* section.
- **Smart suggestions (💡 Suggest), fully offline.** The dock reads the
active layer's fields and picks the most insightful chart — a
correlated-pair scatter with a trend line, a mean-by-category bar, a
count bar or a histogram — configures the controls and renders it,
telling you *why*. No account, no internet, no new dependencies
(`core/assistant.py`).
- **Explore, levelled up.** The dashboard adds a collapsible
**field-summary table** (type · missing % · distinct · range/top per
field), an overall **completeness** KPI, a **normalised box plot**
putting every numeric field on one 0–1 axis (the same comparability
idea as the map-diagram normaliser), and deeper **insights** —
dominant-category share, range, **skew with a log-transform hint**,
**outliers** (1.5×IQR), near-constant and mostly-empty fields, and a
couple more notable correlations.
- Verified end-to-end on **QGIS 3.44** and **QGIS 4**: expression
labels evaluate to multi-line, formatted text; all four normalisation
expressions parse and evaluate; the richer profile/dashboard build;
the assistant and guide produce the expected output; and the dock
wiring (label/diagram apply, Suggest, hints) was driven offscreen. No
regression in the 48 static + 12 animated engine pages or the existing
profile/dashboard tests.

None