fix(artnet): drive each effect with its own parameters, and pair Plasma with plasma

[BernardJen]

Groundwork for better Art-Net visuals, and a fix for a bug 3.1.0 shipped.

effect:plasma never actually worked

Every effect on the relay declares its own parameter set, and the relay ignores anything it doesn't recognise — silently, with a 200. 3.1.0 sent the spot's shape (r,g,b,cx,cy,diameter,softness) to whichever effect was configured. But plasma takes scale, speed, brightness — and no colour at all. So it received seven keys it had no use for, ran at its defaults forever, and never responded to the picture. It looked wired up and wasn't; nothing failed anywhere.

EFFECT_DRIVERS now maps each of the seven field effects to what it actually accepts:

effect	driven by
spot, ripple	colour + centroid position
plasma, aurora	scale from spread, speed and brightness from frame level (no colour — they generate their own)
blobs	colour + integer count and size from spread
sweep	colour + width
tunnel	brightness + integer rings/arms

The other ten effects (fire, police, sparkle, chase, …) aren't field-based, so the relay can't nudge them — their parameters only change by re-POSTing, which resets the animation phase. Those are now started once and left alone rather than nudged pointlessly every second.

Default pairings

Plasma → plasma, Metaballs → blobs, Wave Tank → ripple. Only where the effect genuinely mirrors the screen — a mismatched effect puts the room out of step with the wall, which is worse than the dominant colour. Everything else stays reactive, and Reactive is selectable to opt out.

Mutation testing found a bug in that pairing

The settings UI stores "Reactive" by deleting the entry. For a paired saver that meant:

• the dropdown read Reactive while the pairing was silently in effect — stating the opposite of what the room was doing
• choosing Reactive deleted the key, fell back through to the pairing, and opting out was impossible

The list now shows the effective mode and writes 'reactive' explicitly where a pairing exists. modeForSaver uses hasOwnProperty for the same reason — a truthiness check let a deliberately-cleared entry fall through.

Also removed a genuinely unreachable branch: every driver is field-based, a test enforces it, so the non-field guard could never fire.

Verification

• 784 tests green (33 new), lint clean.
• All 19 mutations caught. Three escaped first, each a real gap: integer parameters sampled only at a spread that lands on a whole number by luck, the unreachable guard, and the opt-out bug above.
• Declared parameter sets are pinned in the test from the live GET /effects, so drift in our table is detectable.

Still unverified against the rig (#274) — this is all parameter shaping, checked against the published schema rather than against light on a wall.

Next: the spatial part of the ask — mapping specific colours on specific parts of the wall onto the strips nearest them, using POST /strips/{name}/pixels.

🤖 Generated with Claude Code