Skip to content

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

fix: dedicated model host startup failures log a bounded cause (#13024)

[._.]/b9735eb79704
diff --git a/server/index.js b/server/index.js
index b644c71009..0a19cbc6c6 100644
--- a/server/index.js
+++ b/server/index.js
@@ -599,8 +599,10 @@ if (!isSmokeBoot()) {
// Reminders for scheduled human action steps (Brain threads); zero provider calls.
import('./services/humanActionReminders.js').then(({ initHumanActionReminders }) => initHumanActionReminders())
.catch((err) => console.error(`❌ Human action reminders could not start: ${err.message}`));
- import('./services/fleetLlmHost.js').then(({ startFleetLlmHost }) => startFleetLlmHost())
- .catch(() => console.error('❌ Dedicated model host listener could not start; open AI Providers → Model host setup.'));
+ import('./services/fleetLlmHost.js').then(({ startFleetLlmHost, describeFleetHostError }) => startFleetLlmHost().catch((err) => {
+ console.error(`❌ Dedicated model host listener could not start (${describeFleetHostError(err)}); open AI Providers → Model host setup.`);
+ }))
+ .catch((err) => console.error(`❌ Dedicated model host listener could not load: ${err.message}`));
// Local file scan only (no provider calls); keeps the federated Claude Code
// token history current even when the usage page is never opened.
import('./services/claudeCodeTranscriptUsage.js').then(({ startTranscriptUsageRefresh }) => startTranscriptUsageRefresh())
diff --git a/server/services/fleetLlmHost.js b/server/services/fleetLlmHost.js
index 230aedc3b9..8c80602379 100644
--- a/server/services/fleetLlmHost.js
+++ b/server/services/fleetLlmHost.js
@@ -58,6 +58,12 @@ async function readHostEnv() {
return { project, env, apiKey };
}
+/** Bounded, secret-free cause for a listener failure: code/syscall/port, else a truncated first message line. */
+export function describeFleetHostError(err) {
+ if (err?.code) return [err.code, err.syscall, err.port].filter(Boolean).join(' ');
+ return String(err?.message ?? err).split('\n')[0].slice(0, 200);
+}
+
export async function startFleetLlmHost() {
if (gateway) return;
const enabled = parseEnvContents((await tryReadFile(PORTOS_ENV_PATH)) || '').get(ENABLED_KEY) === '1';
@@ -72,7 +78,7 @@ export async function startFleetLlmHost() {
next.server.once('error', reject);
next.server.listen(PORTS.FLEET_LLM, '0.0.0.0', resolve);
});
- next.server.on('error', () => console.error('❌ Fleet inference listener failed'));
+ next.server.on('error', (err) => console.error(`❌ Fleet inference listener failed (${describeFleetHostError(err)})`));
gateway = next;
noteFleetHostChanged();
}
diff --git a/server/services/fleetLlmHost.test.js b/server/services/fleetLlmHost.test.js
index 77dbc4c43c..368e8ed4a7 100644
--- a/server/services/fleetLlmHost.test.js
+++ b/server/services/fleetLlmHost.test.js
@@ -3,7 +3,7 @@ vi.mock('./fleetLlmStartup.js', () => ({
isFleetHostLoginTaskInstalled: async () => true,
removeFleetHostLoginTask: async () => { state.loginTaskRemoved = true; return { success: true }; },
}));
-import { beforeEach, describe, expect, it, vi } from 'vitest';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
const state = vi.hoisted(() => ({ providers: [], writes: [], enabled: '', loginTaskRemoved: false, specs: { platform: 'win32', cuda: { gpus: [{ name: 'NVIDIA RTX 3090', vramGb: 24 }] } }, peers: [], peerResponses: new Map() }));
// The usage ledger has its own suite and its own disk; here it would only pull
// `readJSONFile` into the narrow `fileUtils` double this file already declares.
@@ -28,7 +28,7 @@ vi.mock('../lib/systemCapabilities.js', () => ({ detectSystemCapabilities: async
vi.mock('../lib/cudaCapability.js', () => ({ getCudaUtilization: async () => ({ gpus: [{ memoryUsedMib: 1000 }] }) }));
vi.mock('../lib/tailscale.js', () => ({ getTailscaleStatus: async () => ({ running: true, dnsName: 'host-XXXX.example.ts.net' }) }));
vi.mock('../lib/fileUtils.js', () => ({ PATHS: { root: '/example' }, atomicWrite: async (path, value) => state.writes.push({ path, value }), tryReadFile: async () => state.enabled }));
-vi.mock('node:fs/promises', () => ({ readFile: async () => 'VLLM_API_KEY=example-api-key-with-32-characters\n' }));
+vi.mock('node:fs/promises', () => ({ readFile: async () => { if (state.readError) throw state.readError; return 'VLLM_API_KEY=example-api-key-with-32-characters\n'; } }));
vi.mock('../lib/portosEnv.js', async (original) => ({ ...(await original()), upsertPortosEnvLine: async (key, value) => { state.enabled = `${key}=${value}`; } }));
vi.mock('../lib/commandExists.js', () => ({ commandOutput: async (_cmd, args) => args.includes('config') ? JSON.stringify({ services: { single: { image: 'example-runtime' } } }) : args.includes('inspect') ? 'sha256:' + 'a'.repeat(64) : '1.0' }));
vi.mock('../lib/vllmQwenProject.js', () => ({ inspectVllmQwenProject: async () => ({ dir: '/example', composeFile: '/example/compose.yaml', hasWeights: true }) }));
@@ -36,8 +36,8 @@ vi.mock('../lib/openAiModelsProbe.js', () => ({ probeOpenAiModels: async () => (
vi.mock('../lib/streamingSpawn.js', () => ({ runStreamingCommand: vi.fn(async () => ({ success: true })) }));
vi.mock('./vllmQwenManager.js', () => ({ ensureVllmProjectDir: async () => null, provisionVllmQwenProject: async () => ({ success: true }) }));
vi.mock('./providers.js', () => ({ getAllProviders: async () => ({ providers: state.providers }), createProvider: async (record) => state.providers.push(record), updateProvider: async (id, patch) => Object.assign(state.providers.find(p => p.id === id), patch) }));
-vi.mock('./fleetLlmGateway.js', () => ({ createFleetLlmGateway: () => ({ server: { once() {}, on() {}, listen(_port, _host, done) { done(); } }, status: () => ({ active: 0, queued: 0 }), close: async () => {} }) }));
-import { configureFleetLlmHost, disableFleetLlmHost, getFleetLlmHostStatus, getFleetPeerHosts, revealFleetPeerHostKey, stopFleetLlmHost } from './fleetLlmHost.js';
+vi.mock('./fleetLlmGateway.js', () => ({ createFleetLlmGateway: () => ({ server: { once(_evt, fail) { this.fail = fail; }, on() {}, listen(_port, _host, done) { if (state.listenError) this.fail(state.listenError); else done(); } }, status: () => ({ active: 0, queued: 0 }), close: async () => {} }) }));
+import { configureFleetLlmHost, describeFleetHostError, disableFleetLlmHost, startFleetLlmHost, getFleetLlmHostStatus, getFleetPeerHosts, revealFleetPeerHostKey, stopFleetLlmHost } from './fleetLlmHost.js';
import { runStreamingCommand } from '../lib/streamingSpawn.js';
beforeEach(async () => { await stopFleetLlmHost(); state.writes = []; state.enabled = ''; state.loginTaskRemoved = false; state.providers = []; state.peers = []; state.peerResponses.clear(); vi.clearAllMocks(); });
describe('dedicated host setup workflow', () => {
@@ -160,3 +160,26 @@ describe('fleet peer host discovery', () => {
});
});
+
+describe('dedicated host startup failures', () => {
+ beforeEach(() => {
+ state.enabled = 'PORTOS_FLEET_LLM_ENABLED=1';
+ });
+ afterEach(() => {
+ state.readError = null;
+ state.listenError = null;
+ vi.restoreAllMocks();
+ });
+
+ it('rejects with a port-conflict cause the boot log can name', async () => {
+ state.listenError = Object.assign(new Error('listen EADDRINUSE: address already in use 0.0.0.0:1'), { code: 'EADDRINUSE', syscall: 'listen', port: 1 });
+ const err = await startFleetLlmHost().catch((e) => e);
+ expect(describeFleetHostError(err)).toBe('EADDRINUSE listen 1');
+ });
+
+ it('describes a rejected env read without echoing the API key', async () => {
+ state.readError = new Error('permission denied\nVLLM_API_KEY=example-api-key-with-32-characters');
+ const err = await startFleetLlmHost().catch((e) => e);
+ expect(describeFleetHostError(err)).toBe('permission denied');
+ });
+});