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
/**
* Align the shipped MTPLX provider presets with the model id emitted by the
* managed MTPLX server.
*
* MTPLX serves one checkpoint under its generated model id. The old provider
* seed used the friendly alias `mtplx`, which made readiness report a mismatch
* and caused OpenCode to request a model the endpoint did not advertise. Only
* the untouched sentinel configuration is migrated; a user who selected a
* different checkpoint or model list keeps that choice.
*/
import { readProvidersDoc, writeJsonAtomic } from './_lib.js';
const OLD_MODEL = 'mtplx';
const NEW_MODEL = 'mtplx-qwen38-27b-optimized-speed';
const PROVIDER_IDS = ['mtplx', 'opencode-mtplx', 'opencode-mtplx-tui'];
const isUntouchedSentinel = (provider) =>
Array.isArray(provider?.models)
&& provider.models.length === 1
&& provider.models[0] === OLD_MODEL
&& provider.defaultModel === OLD_MODEL;
export default {
async up({ rootDir }) {
const doc = await readProvidersDoc({ rootDir });
if (!doc.ok) return { ok: false, reason: doc.reason, updated: 0 };
let updated = 0;
for (const id of PROVIDER_IDS) {
const provider = doc.providers[id];
if (!isUntouchedSentinel(provider)) continue;
provider.models = [NEW_MODEL];
provider.defaultModel = NEW_MODEL;
updated += 1;
}
if (!updated) return { ok: true, reason: 'already-current-or-custom', updated: 0 };
await writeJsonAtomic(doc.path, doc.config);
console.log(`📝 ${doc.path}: aligned ${updated} MTPLX provider model preset${updated === 1 ? '' : 's'} with the served model`);
return { ok: true, reason: 'updated', updated };
},
};