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
// Writers Room per-scene image-gen defaults + style discriminators.
//
// Mirror of server/lib/writersRoomStylePresets.js — kept in sync manually
// since client and server are separate bundles. Curated preset ids come from
// the /api/image-gen/style-presets endpoint at runtime; only the two
// special discriminators live here.
import { IMAGE_GEN_MODE, LOCAL_IMAGEGEN_DEFAULT_MODEL, installLocalModelId } from './imageGenBackends';
export const STYLE_ID = Object.freeze({ NONE: 'none', CUSTOM: 'custom' });
export const EMPTY_IMAGE_STYLE = Object.freeze({ presetId: STYLE_ID.NONE, prompt: '', negativePrompt: '' });
// Defaults for the per-scene image gen pipe. Klein-4B is the fastest FLUX.2
// variant on Apple Silicon, and 768×512 is a 3:2 aspect that suits scene work.
// steps + seed are stored as strings (or empty) so the form inputs can bind
// directly. Empty string = "use model default" for steps, "random per render"
// for seed. Parsed to numbers at the generateImage boundary in SceneCard.
export const WR_IMAGE_DEFAULTS = Object.freeze({
modelId: LOCAL_IMAGEGEN_DEFAULT_MODEL,
mode: IMAGE_GEN_MODE.LOCAL,
width: 768,
height: 512,
steps: '',
seed: '',
});
// Build the /image-gen/generate request body for a Writers Room scene render.
// Folds the prompt + the per-scene imageCfg (parsing the string steps/seed
// inputs to numbers, dropping empties so the model uses its own defaults) into
// the payload shape both SceneCard and the live render preview send. Keeping
// the steps/seed parsing in one place means the two render entry points can't
// drift on what "use the model default" means.
//
// `writersRoom` ({ workId, analysisId, sceneId }), when provided, tags the
// render so the server-side completion hook files it onto the analysis snapshot
// durably — sparing the client a generate-then-attach round-trip on the async
// local/Codex lanes (#1363).
export function buildSceneRenderPayload({ prompt, negativePrompt = '', imageCfg = WR_IMAGE_DEFAULTS, writersRoom = null }) {
const stepsNum = imageCfg.steps ? Number(imageCfg.steps) : undefined;
const seedNum = imageCfg.seed && Number(imageCfg.seed) >= 0 ? Number(imageCfg.seed) : undefined;
return {
prompt,
negativePrompt,
modelId: imageCfg.modelId,
mode: imageCfg.mode,
width: imageCfg.width,
height: imageCfg.height,
...(Number.isFinite(stepsNum) ? { steps: stepsNum } : {}),
...(Number.isFinite(seedNum) ? { seed: seedNum } : {}),
...(writersRoom ? { writersRoom } : {}),
};
}
// Resolve the per-scene render config. When the user hasn't pinned a Writers
// Room mode, prefer Codex if enabled — cloud models render storyboard scenes
// more reliably than local diffusion when both are available. If the stored
// mode points at a backend that's no longer available (codex disabled, local
// pythonPath cleared), fall back to the first available backend so the form
// always reflects something that will actually run.
export function readWrImageSettings(settings, availableBackends = null) {
const stored = settings?.writersRoom?.imageGen || {};
const codexEnabled = settings?.imageGen?.codex?.enabled === true;
const grokEnabled = settings?.imageGen?.grok?.enabled === true;
const agyEnabled = settings?.imageGen?.agy?.enabled === true;
const defaultMode = codexEnabled ? IMAGE_GEN_MODE.CODEX
: grokEnabled ? IMAGE_GEN_MODE.GROK
: agyEnabled ? IMAGE_GEN_MODE.AGY : WR_IMAGE_DEFAULTS.mode;
let mode = stored.mode || defaultMode;
if (Array.isArray(availableBackends) && availableBackends.length > 0
&& !availableBackends.some((b) => b.id === mode)) {
mode = availableBackends[0].id;
}
return {
modelId: stored.modelId || installLocalModelId(settings),
mode,
width: Number.isFinite(stored.width) ? stored.width : WR_IMAGE_DEFAULTS.width,
height: Number.isFinite(stored.height) ? stored.height : WR_IMAGE_DEFAULTS.height,
steps: stored.steps != null && stored.steps !== '' ? String(stored.steps) : '',
seed: stored.seed != null && stored.seed !== '' ? String(stored.seed) : '',
};
}