Node Dashboard
Tour of the local node dashboard served by the host agent — run, monitor, pause, and test your node from one command panel.
The node dashboard is the operator command panel for a running node. It is the same local web app you used during setup, opened in its command view, and it is also embedded in the desktop app. Use it to watch readiness, review activity and earnings, pause work, and test runtime and model behavior locally.
After setup completes, the host agent keeps serving
this app from the local listener at http://127.0.0.1:8831. It opens on the
Overview tab.
Command View
The dashboard is a single command view with a row of tabs across the top. Each tab is one job — watch the node, manage models, test inference, mint keys, or adjust settings.
| Tab | Purpose |
|---|---|
| Overview | Node readiness, posture, and a summary of local activity |
| Models | Installed models and their per-profile readiness |
| Playground | Test runtime and model behavior against the local endpoint |
| API | Create, scope, and revoke local API keys, and review local usage |
| Benchmark | Run and review benchmark evidence for a model profile |
| Earnings | Observed earning state and receipt history (shown once the node is enrolled) |
| Settings | Node configuration, including the Runtimes and Serving controls |
| Logs | Local request and service logs, redacted by default |
Runtime install and the serving-mode toggle are Settings sub-tabs rather than top-level tabs. Everything else above is reachable directly from the tab row.
Host & Advanced Controls
A single Host & advanced controls button in the top bar opens one takeover covering node-level state and connectivity — readiness, machine details, claim, and network/tunnel overrides all live inside it:
Tunnel rotate and revoke control the node's public inference route only. Rotating issues a fresh credential; revoking takes the node off the public route until you bring it back.
Pause And Resume Work
Serving has its own dedicated Pause / Resume control — from the tray menu,
or POST /v1/serving/pause and /v1/serving/resume on the local API. Pausing
stops the node from accepting new routed work while any in-flight job finishes;
enrollment, account link, and identity are all kept, so nothing needs
re-setting-up when you resume.
This is the right tool for maintenance, driver updates, or reclaiming the machine for a while. Resume puts it straight back into rotation.
A paused node reports its state through the same readiness surfaces, so it shows as not currently taking routed work rather than as broken.
Separately, Settings → Serving carries the mode choice from setup — Serve inference locally vs Link to your account & earn — for operators who want to change their mind about whether the node reaches outside the machine at all.
Pause is a short, reversible break from routed work. A machine's Bare metal / Earn mode (set from the machine's page in the account dashboard) is a different and more drastic switch: Earn mode removes the machine from B3IQ serving entirely — no SSH, no health telemetry, no inference (a liveness heartbeat continues) — so it can be offered to external GPU marketplaces. Reach for pause first; only flip to Earn mode if you want the machine off B3IQ inference altogether.
Local Playground
The Playground tab runs prompts against the local OpenAI-compatible endpoint
so you can confirm runtime and model behavior before trusting routed traffic.
Pick a model id from the local model list (GET /v1/models), send a prompt, and
watch the response.
Local request logs are redacted by default so prompts and outputs stay out of the log view. Local request content stays local unless you send it to a runtime or explicitly use a routed workflow.
When testing from a script instead of the Playground, mint a local key from the
API tab and keep it out of shell history — read it from an env var or a
chmod 600 file rather than passing it on the command line. See the
Host Agent page for safe local-CLI patterns.
Your Machine In The Account Dashboard
The local panel above is not the whole story once a machine is claimed. Its
primary home becomes the account dashboard at b3iq.org/dashboard — reachable
from any browser, not just the machine itself — where each machine gets its
own detail page with tabs:
The account dashboard is also where fleet-wide surfaces live: a Models page
that aggregates usage across every machine you own, org fleet SSH keys that
apply to every machine an organization owns, automatic health email alerts
(offline, GPU running hot, disk nearly full), a Support tab for filing
tickets, and the account-level b3iq CLI for driving all
of it from a terminal — see Host agent for how that
CLI differs from the local b3iq-host one.
Node State In The Public View
The public node list mirrors the kind of state the dashboard shows for your own node — readiness, model coverage, and whether a node is currently serving.
