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 { MessageSquare, Code2, Brain, Eye, Boxes, Wrench, Music } from 'lucide-react';
/**
* The capability icon row, shared by every surface that shows what a model can do.
*
* It started life inside the install tab and is now also the gate the
* Performance page's capability tests key off — so it lives here rather than
* being copied. Two surfaces rendering the same claim with different icons (or,
* worse, a different vocabulary) is how a model ends up looking tool-capable on
* one page and not on another.
*
* `cls` is the icon color; the bordered chip uses the same hue at low opacity —
* the LM Studio style, icons rather than text, because a model with five badges
* would otherwise wrap a card into three lines.
*/
export const CAPABILITY_META = {
chat: { Icon: MessageSquare, label: 'Chat', cls: 'text-gray-400 border-gray-500/50' },
code: { Icon: Code2, label: 'Code', cls: 'text-sky-400 border-sky-400/50' },
reasoning: { Icon: Brain, label: 'Reasoning', cls: 'text-emerald-400 border-emerald-400/50' },
vision: { Icon: Eye, label: 'Vision', cls: 'text-amber-400 border-amber-400/50' },
embeddings: { Icon: Boxes, label: 'Embeddings', cls: 'text-violet-400 border-violet-400/50' },
tools: { Icon: Wrench, label: 'Tool use', cls: 'text-blue-400 border-blue-400/50' },
audio: { Icon: Music, label: 'Audio generation', cls: 'text-pink-400 border-pink-400/50' },
};
/**
* A model's capability badges.
*
* `capabilities` is `null` when NOTHING authoritative reported a badge set — a
* bare llama.cpp endpoint lists model ids and nothing else. That is rendered as
* an explicit "not reported" chip rather than as an empty row, because an empty
* row reads as "this model can do nothing", which is a different and untrue
* claim. `[]` (a runtime that answered and named none) legitimately renders
* nothing.
*/
export default function CapabilityBadges({ capabilities, unknownLabel = 'capabilities not reported' }) {
if (!Array.isArray(capabilities)) {
return (
<span
className="px-1.5 py-0.5 text-[10px] rounded border border-dashed border-port-border text-gray-500"
title="This runtime lists model ids only, so PortOS cannot say what this model claims."
>
{unknownLabel}
</span>
);
}
return capabilities.map((capability) => {
const meta = CAPABILITY_META[capability];
// An unmapped capability still shows — the server's vocabulary can grow
// ahead of this map, and silently dropping one would hide a real claim.
if (!meta) {
return <span key={capability} className="px-1.5 py-0.5 bg-port-border/60 rounded">{capability}</span>;
}
const Icon = meta.Icon;
return (
<span
key={capability}
title={meta.label}
aria-label={meta.label}
className={`inline-flex items-center justify-center w-5 h-5 rounded border ${meta.cls}`}
>
<Icon size={12} />
</span>
);
});
}