AM Pilot | Ops

Health Center

Health Center

API

Checking Waiting for /api/health

Feed

Checking Waiting for internal health payload

Signals

Pending Awaiting the first admin or customer app signal

Last sync

Updated Pending

Tenants

— Loading

Critical

— Checking

Incidents 24h

— Waiting

Active tenants 7d

— Checking

Attention

Needs action

Waiting for tenant posture data.

Health issues will appear here.

Incidents

Live app failures

Waiting for client events.

Runtime and API failures will appear here.

Tenant matrix

Admin and customer app posture

Tenant	Admin app	Customer app	Activity	Issues

Customer workspaces

Provisioning

Workspaces, domains, first-admin access, and platform fee settings.

Workspaces

Customer list

Provision and manage customer workspaces for the platform.

0 active workspaces

Create

New workspace

Create a tenant workspace and invite the first administrator in one pass.

Tenant slug * Company name

Communication email Brand logo URL

Quote domain Admin domain

Admin email * Admin full name

Platform fee (%)

My account

Password

Update your platform-owner password for the provisioning surface.

Estimator quality

Estimator Lab

Keep this surface focused on shadow slicing, component errors, confidence bands, and release gates. It should explain whether the estimator is trustworthy enough to ship.

Shadow mode first Target fidelity: 90%

Shadow labels

Capture ML vs native slice deltas

Per upload, persist the resolved recipe, ML prediction, native result, slicer version, and estimator version.

Error segmentation

Break time and grams by component

Model time, support time, travel time, part grams, and support grams need separate quality visibility.

Release gates

Promote only when regimes improve

Validate by printer, material, layer height, support usage, and job size before any model promotion.

Before charts

Data model that has to exist

Resolved recipe snapshot shared by ML and native slicing
Per-job labels for print time and grams from both engines
Versioning for estimator, native slicer, and config bundle
Offline retraining flow instead of direct online learning

Boundary

What does not belong here

Generic runtime incidents should stay in Health Center
Tenant onboarding should stay in Provisioning
Release actions should wait until shadow validation data is real