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 { useCallback, useEffect, useRef, useState } from 'react';
import { getLocalLlmStatus, installLocalLlmBackend } from '../services/api';
import socket from '../services/socket';
import toast from '../components/ui/Toast';
import { localLlmBackendLabel } from '../lib/localLlmBackends.js';
/**
* The installed-model status both local-LLM views read, plus the one
* busy/toast/refresh path every action on either view routes through.
*
* The Runtimes and Model Library views never mount together (the tab renders
* exactly one), so each mounted view owning this hook still leaves exactly ONE
* subscriber on `localLlm:progress` — the single-subscriber rule holds without
* threading status down as props from a parent that renders neither view.
*
* @param {object} [options]
* @param {Function} [options.onRefresh] Extra loads to fire alongside a status
* refresh (the Runtimes view pulls llama.cpp / MTPLX / Slotstream here).
* @param {Function} [options.onReload] Extra reload after an action or a
* `complete` progress frame (the Library view re-queries its catalog).
* BOTH must be stable (`useCallback`) — they are effect dependencies.
*/
export default function useLocalLlmStatus({ onRefresh, onReload } = {}) {
const [status, setStatus] = useState(null);
const [loading, setLoading] = useState(true);
const [selected, setSelected] = useState('ollama');
const [actionInProgress, setActionInProgress] = useState(null);
const [progressMsg, setProgressMsg] = useState('');
const [confirmAction, setConfirmAction] = useState(null);
const progressTimer = useRef(null);
const statusRequestId = useRef(0);
const selectedInitialized = useRef(false);
const loadStatus = useCallback(() => {
const requestId = ++statusRequestId.current;
setLoading(true);
onRefresh?.();
return getLocalLlmStatus({ silent: true })
.then((s) => {
if (requestId !== statusRequestId.current) return;
setStatus(s);
// Default the model-management view to the active backend on first load.
if (!selectedInitialized.current && s?.backend) {
setSelected(s.backend);
selectedInitialized.current = true;
}
})
.catch(() => {
if (requestId === statusRequestId.current) toast.error('Failed to load local LLM status');
})
.finally(() => {
if (requestId === statusRequestId.current) setLoading(false);
});
}, [onRefresh]);
useEffect(() => { loadStatus(); }, [loadStatus]);
useEffect(() => {
const handleProgress = (data) => {
// `localLlm:progress` is a shared channel. Measurement frames (`assessment`,
// `assessment-sweep`) belong to the Performance tab and say nothing about
// what is installed here — and an overnight sweep emits a `complete` frame
// per model, so answering them would reload the status AND re-query the
// Hugging Face catalog once per measured model, all night. This tab owns
// the unscoped install/migrate/upgrade frames only.
if (data?.scope === 'assessment' || data?.scope === 'assessment-sweep' || data?.scope === 'security-guard' || data?.scope === 'jev') return;
clearTimeout(progressTimer.current);
setProgressMsg(data.message || '');
if (data.event === 'complete') {
progressTimer.current = setTimeout(() => setProgressMsg(''), 3000);
loadStatus();
onReload?.();
}
if (data.event === 'error') {
progressTimer.current = setTimeout(() => setProgressMsg(''), 5000);
}
};
socket.on('localLlm:progress', handleProgress);
return () => {
socket.off('localLlm:progress', handleProgress);
clearTimeout(progressTimer.current);
};
}, [loadStatus, onReload]);
const runAction = useCallback((key, fn, successMsg, options = {}) => {
const { onError, clearConfirm = true, ollamaService = false } = options;
if (clearConfirm) setConfirmAction(null);
setActionInProgress(key);
return fn()
.then((result) => {
if (successMsg) toast.success(typeof successMsg === 'function' ? successMsg(result) : successMsg);
// Optimistic repaint for the Ollama service controls only. Every runtime
// start/stop result carries `running` — llama-server's and MTPLX's too —
// so the CALLER declares this, rather than it being inferred from the
// response shape; otherwise stopping MTPLX would paint Ollama as stopped
// until the refetch lands.
if (ollamaService && typeof result?.running === 'boolean') {
setStatus((prev) => prev ? ({
...prev,
ollama: {
...prev.ollama,
installed: true,
available: result.running
}
}) : prev);
}
loadStatus();
onReload?.();
return result;
})
.catch((err) => {
// Caller-handled errors (e.g. OLLAMA_OUTDATED → offer to upgrade) ask us
// to skip the default toast and run their own handler instead. The error
// toast from apiCore has already fired unless the caller passed {silent}
// through fn — onError just gets to consume the structured code/context.
if (typeof onError === 'function') onError(err);
})
.finally(() => setActionInProgress(null));
}, [loadStatus, onReload]);
// Both views offer "install the missing backend" — the Runtimes card as a row
// action, the Library as the blocker on an empty catalog — so the action lives
// here rather than as two copies that could drift on their toast or key.
const installBackend = useCallback((backend) => runAction(
`runtime-install-${backend}`,
() => installLocalLlmBackend(backend),
(r) => r?.note ? `Installed ${localLlmBackendLabel(backend)} — ${r.note}` : `Installed ${localLlmBackendLabel(backend)}`
), [runAction]);
return {
status,
loading,
selected,
setSelected,
loadStatus,
runAction,
installBackend,
actionInProgress,
busy: actionInProgress != null,
progressMsg,
setProgressMsg,
confirmAction,
setConfirmAction,
};
}