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 { useState, useRef, useEffect, useMemo } from 'react';
import { Link, useNavigate } from 'react-router';
import { Sparkles, Pencil, ChevronDown, ChevronRight } from 'lucide-react';
import { generateImage, listUniverseStyles } from '../services/api';
import { RESOLUTIONS, MAX_IMAGE_EDGE, MAX_IMAGE_PIXELS } from '../lib/imageGenResolutions';
import { composeStyledPrompt } from '../lib/composeStyledPrompt';
import { universeStylePreset } from '../lib/universeStylePreset';
import { useLocalStoragePersisted } from '../hooks/useLocalStorageBool';
import useMounted from '../hooks/useMounted';
import ResolutionField from './media/ResolutionField';
import MediaJobThumb from './pipeline/MediaJobThumb';
import AutoSizeTextarea from './ui/AutoSizeTextarea';
import toast from './ui/Toast';
// Universal presets only (no `compatible` gate) so the quick widget can offer
// a resolution dropdown without knowing the active backend/model — these sizes
// render on every image-gen mode. The full Image Gen page keeps the
// backend-filtered list via ImageGenControls.
const QUICK_RESOLUTIONS = RESOLUTIONS.filter((r) => !r.compatible);
export default function QuickImagePrompt() {
const [prompt, setPrompt] = useState('');
const [negativePrompt, setNegativePrompt] = useState('');
const [showNegative, setShowNegative] = useState(false);
const [width, setWidth] = useState(1024);
const [height, setHeight] = useState(1024);
const [universeStyles, setUniverseStyles] = useState([]);
const [universeId, setUniverseId] = useLocalStoragePersisted('dashboard.quickImage.universeId', '');
const mountedRef = useMounted();
// The current/last submission. `jobId` set → async backend (local/codex):
// render the live MediaJobThumb so the user sees the diffusion spinner →
// final image, exactly like the Universe asset slots. `filename` only →
// sync external backend returned a completed render directly.
const [job, setJob] = useState(null);
const [isSubmitting, setIsSubmitting] = useState(false);
const submittingRef = useRef(false);
const navigate = useNavigate();
// Every row is selectable — the endpoint already drops universes with no
// style tokens. Failure is non-fatal: the picker stays hidden and the widget
// behaves exactly as it did before.
useEffect(() => {
listUniverseStyles({ silent: true })
.then((rows) => { if (mountedRef.current) setUniverseStyles(Array.isArray(rows) ? rows : []); })
.catch(() => { /* non-fatal — widget degrades to un-styled quick renders */ });
}, [mountedRef]);
// The selected universe's `{ prompt, negativePrompt }` preset, or null when
// nothing is selected — including when the saved id no longer resolves, so a
// deleted universe can't keep silently styling renders.
const stylePreset = useMemo(() => {
const match = universeId ? universeStyles.find((u) => u.id === universeId) : null;
return match ? universeStylePreset(match) : null;
}, [universeId, universeStyles]);
// Same composition the Universe Builder's canon renders use: the universe's
// embrace tokens PREFIX the user prompt (diffusion weights earlier tokens
// heaviest) and its avoid tokens append to the negative.
const composeForSubmit = (text) => composeStyledPrompt(text, negativePrompt, stylePreset);
const handleResolution = (w, h) => { setWidth(w); setHeight(h); };
const handleGenerate = async (e) => {
e?.preventDefault();
const text = prompt.trim();
if (!text || submittingRef.current) return;
submittingRef.current = true;
setIsSubmitting(true);
// Omit `mode` so the server falls back to the user's saved
// `settings.imageGen.mode` default. Async backends (local/codex) respond
// with { jobId, status, position } — sync external responds with the
// generation result ({ filename, path }). Send the negative prompt as-is
// (empty string is an intentional clear; the server treats present-but-
// empty as "no negative prompt"). Preserve the input on failure so the
// user doesn't have to retype after a server error (the API helper toasts
// on its own).
const styled = composeForSubmit(text);
const result = await generateImage({
prompt: styled.prompt,
negativePrompt: styled.negativePrompt,
width,
height,
}).catch(() => null);
submittingRef.current = false;
setIsSubmitting(false);
if (result) {
// Only clear if the textarea still holds the submitted text — the user
// can keep typing while the request is in flight and we don't want to
// wipe out new input on resolve.
setPrompt((current) => (current === text ? '' : current));
if (result.jobId) {
setJob({ jobId: result.jobId, filename: null });
toast.success('Image queued');
} else if (result.filename) {
setJob({ jobId: null, filename: result.filename });
toast.success('Image generated');
}
}
};
const handleOpenInEditor = (e) => {
e?.preventDefault();
const text = prompt.trim();
// ImageGen page's remix-param effect reads ?prompt=…&width=…&height=… on
// mount and strips them from the URL, so the widget can hand off the
// current prompt + size + negative prompt without coupling to the form's
// internal state. Hand off the STYLED prompt so the editor opens with what
// Generate would have submitted — the Image Gen page has no universe picker
// of its own, so an un-styled handoff would silently drop the selection.
const styled = composeForSubmit(text);
const params = new URLSearchParams();
if (styled.prompt) params.set('prompt', styled.prompt);
if (styled.negativePrompt) params.set('negativePrompt', styled.negativePrompt);
params.set('width', String(width));
params.set('height', String(height));
const qs = params.toString();
navigate(`/media/image${qs ? `?${qs}` : ''}`);
};
const inputCls = 'w-full bg-port-bg border border-port-border rounded-lg px-2 py-2 text-sm text-white focus:outline-none focus:border-port-accent';
return (
<div className="bg-port-card border border-port-border rounded-xl p-4 h-full flex flex-col min-h-0 overflow-y-auto">
<div className="flex items-center justify-between mb-3 shrink-0">
<h3 className="text-sm font-semibold text-white">Quick Image</h3>
<Link to="/media/image" className="text-xs text-gray-500 hover:text-port-accent transition-colors">
Image Gen →
</Link>
</div>
<form onSubmit={handleGenerate} className="flex flex-col gap-2 flex-1 min-h-0">
<label htmlFor="quick-image-prompt" className="sr-only">Image prompt</label>
<AutoSizeTextarea
id="quick-image-prompt"
placeholder="A neon-lit alley at dusk, cinematic, 50mm..."
value={prompt}
onChange={(e) => setPrompt(e.target.value)}
onKeyDown={(e) => {
if (e.key === 'Enter' && (e.metaKey || e.ctrlKey)) handleGenerate(e);
}}
rows={3}
className="w-full min-h-[60px] px-3 py-2 bg-port-bg border border-port-border rounded-lg text-white text-sm"
/>
{/* Universe styling. Hidden entirely when no universe carries style
tokens — an empty dropdown is noise on a compact widget. */}
{universeStyles.length > 0 && (
<div className="shrink-0">
<label htmlFor="quick-image-universe" className="block text-xs font-medium text-gray-400 mb-1">
Universe style
</label>
<select
id="quick-image-universe"
value={universeId}
onChange={(e) => setUniverseId(e.target.value)}
disabled={isSubmitting}
className={`${inputCls} disabled:opacity-50`}
>
<option value="">None</option>
{universeStyles.map((u) => (
<option key={u.id} value={u.id}>{u.name}</option>
))}
</select>
{stylePreset?.prompt ? (
<p className="mt-1 text-[11px] text-gray-500 line-clamp-2" title={stylePreset.prompt}>
Prefixed with: {stylePreset.prompt}
</p>
) : null}
</div>
)}
{/* Shared preset dropdown + "Custom…" width/height inputs. Reuses the
same control as the full Image Gen page so the quick widget offers
custom sizes and (unlike the old inline <select>) never shows a
blank dropdown for an off-preset/remix size — ResolutionField
auto-opens the custom inputs and displays the W×H instead. The
image route accepts any integer edge in [64, 3840] with total
pixels ≤ 8.29M, so step=1 keeps hand-typed non-multiple-of-8 edges
from tripping native form validation. */}
<div className="shrink-0 flex flex-col gap-2">
<ResolutionField
presets={QUICK_RESOLUTIONS}
width={width}
height={height}
onChange={handleResolution}
min={64}
max={MAX_IMAGE_EDGE}
step={1}
maxPixels={MAX_IMAGE_PIXELS}
disabled={isSubmitting}
inputClassName={`${inputCls} disabled:opacity-50`}
/>
</div>
<div className="shrink-0">
<button
type="button"
onClick={() => setShowNegative((v) => !v)}
className="flex items-center gap-1 text-xs font-medium text-gray-400 hover:text-white transition-colors"
aria-expanded={showNegative}
>
{showNegative ? <ChevronDown size={14} /> : <ChevronRight size={14} />}
Negative prompt
</button>
{showNegative && (
<>
<label htmlFor="quick-image-negative" className="sr-only">Negative prompt</label>
<AutoSizeTextarea
id="quick-image-negative"
value={negativePrompt}
onChange={(e) => setNegativePrompt(e.target.value)}
placeholder="Things to avoid…"
rows={2}
className="mt-1 w-full px-3 py-2 bg-port-bg border border-port-border rounded-lg text-white text-sm min-h-[44px]"
/>
</>
)}
</div>
<div className="flex items-center gap-2 shrink-0">
<button
type="submit"
disabled={!prompt.trim() || isSubmitting}
className="flex-1 flex items-center justify-center gap-2 px-3 py-2 bg-port-accent/20 hover:bg-port-accent/30 text-port-accent rounded-lg text-sm transition-colors disabled:opacity-50 min-h-[40px]"
title="Generate with these settings"
>
<Sparkles size={14} />
{isSubmitting ? 'Submitting…' : 'Generate'}
</button>
<button
type="button"
onClick={handleOpenInEditor}
disabled={isSubmitting}
className="flex items-center justify-center gap-2 px-3 py-2 border border-port-border text-gray-300 hover:text-white hover:bg-port-border/50 rounded-lg text-sm transition-colors disabled:opacity-50 min-h-[40px]"
title="Open in editor with these settings"
>
<Pencil size={14} />
Edit
</button>
</div>
{job && (
// Live render preview — async jobs stream the diffusion spinner /
// step counter / latent frame and resolve to the final image;
// sync results render the completed image directly (fallbackFilename
// short-circuits the live subscription). Mirrors the Universe asset
// slot via the shared MediaJobThumb.
<div className="mt-1 shrink-0">
<MediaJobThumb
jobId={job.jobId || job.filename}
fallbackFilename={job.jobId ? null : job.filename}
size="fill"
label="Quick image"
/>
</div>
)}
</form>
</div>
);
}