Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS
JavaScript ai-agentsdashboarddeveloper-toolsdigital-twinquantified-selfsecond-brainself-hostedtailscale
import useImageRenderSettings from './useImageRenderSettings';
import useLocalImageRuntime from './useLocalImageRuntime';
import { deckCardSize } from '../lib/decks';
import { IMAGE_GEN_MODE, IMAGE_RUNTIME_READINESS, RENDER_TARGET, modeLabel } from '../lib/imageGenBackends';
/**
* "What will this deck's next card render actually run on, and can it run at
* all" — the backend pin ladder resolved for `RENDER_TARGET.DECK`, the card
* trim, the local runtime's verdict, and a one-line `summary` naming the
* options, in one object.
*
* A hook rather than state inside the render bar, because the deck page has two
* consumers of the same answer: the render bar (which shows the options and
* queues batches) and the card grid (whose per-card re-render button must queue
* against those same options, name them, and stand down for the same reason).
* Resolving it once also means one settings fetch and one runtime probe per
* page rather than one per consumer.
*/
export default function useDeckRenderTarget(deck) {
const { imageCfg, backends } = useImageRenderSettings({ record: deck, target: RENDER_TARGET.DECK });
const size = deckCardSize(deck);
// `backends` is empty until the settings fetch lands, and until then imageCfg
// is the UNRESOLVED placeholder (local + the install default) — probing on
// that asks about the wrong model, and about a local runtime a cloud-pinned
// deck never touches. Wait for the real answer.
const resolved = backends.length > 0;
const rendersLocally = resolved && imageCfg.mode === IMAGE_GEN_MODE.LOCAL;
const localRuntime = useLocalImageRuntime(rendersLocally ? (imageCfg.modelId || null) : null);
// Queueing against a runtime the server will refuse is the reported complaint
// one step earlier than the error message, so every render affordance stands
// down while it is unavailable. An unknown verdict does NOT block — a probe
// that could not answer must not be able to lock the deck out of rendering.
const blocked = localRuntime.runtime?.readiness === IMAGE_RUNTIME_READINESS.UNAVAILABLE;
// What this deck will actually render on, in the order a sentence wants it.
// Written once: the render bar's collapsed header, its expanded note and the
// grid's per-card button all say the same three facts, and a fourth would
// otherwise have to be added to each. Empty until the fetch lands, which is
// also how a caller knows not to claim anything about the render yet.
const summary = resolved
? [modeLabel(imageCfg.mode), imageCfg.mode === IMAGE_GEN_MODE.LOCAL && imageCfg.inheritedBackend
? 'Inherited local model' : imageCfg.cloudModel || imageCfg.modelId, `${size.width}×${size.height}`].filter(Boolean).join(' · ')
: '';
return { backends, size, summary, localRuntime, blocked };
}