v0.7.0

[0.7.0] — 2026-07-03

The first big step toward v1.0: making cross-entity styling a first-class,
discoverable feature instead of something only possible by hand-typing an
entity_id, letting one set of threshold rules drive more than one visual
property at once, and adding a genuine smooth-fade alternative to discrete
step rules.

Added

• Searchable entity picker everywhere. Every entity field in the panel
  (Threshold's entity, Animation's custom trigger entity, and every
  "controlled by" field below) uses HA's own <ha-entity-picker> (search by
  name, domain icons, autocomplete) via a new shared cms-entity-picker
  component, instead of a bare text input you had to get the entity_id
  exactly right in.
• Icon Color, Background, Filter, and Accent Color can all be controlled
  by a different entity than the card's own — the same capability
  Threshold and Animation already had, generalized to every conditional
  module. Styling one card's appearance off a different entity's state
  (e.g. a toggle card whose icon color reflects a separate status sensor,
  not the toggle entity itself) is now a first-class option everywhere, not
  just something the card's own entity could drive. Available regardless of
  whether the card's own entity has an on/off state at all — a card like
  button (whose entity has none) still offers conditional coloring bound
  to a different, toggleable entity, with a clear inline warning if neither
  the card's own entity nor a picked one is toggleable.
• Accent Color gained the same conditional/entity-binding capability
  every other module already had — previously static-color-only. Also
  dropped the --accent-color CSS-variable name and its explanation from
  the panel; no other module exposes its underlying CSS variable name this
  way, and the code editor is there for anyone who wants to see it.
• Threshold rules can drive multiple properties at once — e.g. icon
  color and accent color changing together off one shared rule set,
  instead of duplicating the same rules per property. "Apply to" is a set
  of checkboxes; round-trip parsing recognises matching threshold blocks
  across properties and merges them into one module state (a genuine
  mismatch is left alone rather than silently merged, and preserved in
  Advanced CSS instead of dropped).
• Threshold Colors "Fade" mode — a genuine alternative to discrete step
  rules: define value→color points (e.g. 0→gray, 150→orange, 220→red) and
  the color blends smoothly between them, clamped at the ends, with a live
  gradient-bar preview and per-point ▲/▼ swap buttons for reordering
  colors without recomputing values by hand. Internally approximated as
  ~32 closely-spaced step rules — HA's sandboxed Jinja2 has no way to build
  a color string from interpolated numbers, so true continuous color math
  isn't reasonably expressible there — but your actual points, not the ~32
  generated ones, come back correctly when reopening the editor, via a
  small marker alongside the real rules in the generated CSS.

Fixed

• ha-state-icon's color property could be silently claimed by the Icon
  Color recognizer even when it didn't understand the value, permanently
  blocking Threshold (and Advanced CSS) from ever reading it on save —
  reachable whenever a card had a threshold-driven icon color alongside a
  different, unrelated threshold-driven property. Icon Color now only
  claims the property in branches where it actually recognises the value.
• Gradient mode's colors could fail to apply against real card-mod
  entirely, with no error anywhere. Root cause: real card-mod's own
  style-string parsing — not this project's — silently drops an entire
  style block the instant a {/} character appears in any declaration's
  value, even safely inside a quoted string a spec-compliant CSS tokenizer
  would treat as inert. The gradient marker's first encoding was JSON,
  which hit exactly that. Confirmed directly against a live card-mod
  instance by isolating single-character-class variants; fixed by
  switching to a brace-free value:color,value:color,... encoding, and
  re-verified end-to-end (Studio UI → generated CSS → real <hui-card>
  render → correct getComputedStyle color) against both real card-mod
  and a real UIX install independently, rather than trusted from source
  reading alone.
• Typing a new value into a gradient point could scramble a different
  point's value mid-edit — the point list re-sorts by value on every
  keystroke, and rows weren't keyed, so Lit's DOM diffing could reuse an
  input element positionally instead of per-point the instant a partial
  value crossed another point's position. Fixed by committing the value on
  blur/Enter instead of every keystroke, and keying the row list by point
  id so a genuine reorder can't steal a focused, in-progress edit.

What's Changed

• Claude/cardmod feature request htidxc by @DerTrolli in #24

Full Changelog: v0.6.2...v0.7.0