Run 35B MoE models in ~2.5 GB of RAM. Weights stream from SSD; a trained router predicts loads a step ahead. Mac, iPhone, Android, Windows supported
Python
import { invoke } from "@tauri-apps/api/core";
import { useCallback, useEffect, useState } from "react";
export interface ServiceHealth {
daemon_version: string;
host_app_version: string | null;
scope: string;
started_by: string;
uptime_s: number;
pid: number;
}
export interface ServicePhase {
phase: string;
health?: ServiceHealth;
code?: number | null;
detail?: string;
base?: string;
spawnedByUs?: boolean;
}
export interface ServiceState {
phase: ServicePhase;
appVersion: string;
versionMismatch: boolean;
spawnWasDev: boolean;
pinned: boolean;
}
export function isLoopback(bound: string): boolean {
return bound === "::1" || bound.startsWith("127.");
}
export function useServiceState(): [ServiceState | null, () => void] {
const [svc, setSvc] = useState<ServiceState | null>(null);
const refresh = useCallback(() => {
invoke<ServiceState>("service_state")
.then(setSvc)
.catch(() => undefined);
}, []);
useEffect(() => {
refresh();
const poll = setInterval(refresh, 2000);
return () => clearInterval(poll);
}, [refresh]);
return [svc, refresh];
}