Console & fleet

Profiles for a fleet

A profile is a named configuration set you assign to devices, so a whole fleet follows one policy instead of being set device by device. This page lists the fields a profile carries, how it is assigned, and the per-mode default. Profiles are an Enterprise capability.

~6 min read · Console & fleet

Profiles are an Enterprise capability: creating, listing, editing and assigning them all require an Enterprise account.

What a profile is

A profile is a reusable bundle of settings with a name. You assign it to devices; a device with none assigned follows the default profile for its mode.

Profile fields

FieldMeaning
nameThe display name, unique per account
modeThe device egress topology: cloud or private
modelVariantThe local model size the device runs (small, large, xlarge or the GPU-only ultra)
secondOpinionThe second-check source: none, a frontier provider, your on-prem endpoint, or a hosted check
dualPolicySet with an on-prem second opinion: the local model decides first, the on-prem endpoint acts as an independent second check
A private profile may not use a cloud-based second opinion: private mode forbids the frontier and hosted sources, so content stays on the machine.

The second opinion

The second-opinion source is where a profile carries the most detail:

  • None: the local model alone.
  • Frontier: a hosted provider and model, with a dual mode and a list of categories the local model is trusted on without a recheck.
  • On-prem: your own endpoint by URL, pinned by its TLS certificate hash, with authentication supplied as a reference to a stored secret rather than inline.

Managing profiles

  • List the profiles on the account, with the current defaults.
  • Create or update a profile in one upsert operation.
  • Assign a profile to a device; the profile mode must match the device mode.
  • Delete a profile, unless it is still assigned to devices or is the last default for its mode.

The per-mode default

There is one default profile per mode, a cloud default and a private default. Each ships as the local model only, with no second opinion, until you assign something else. A device with no explicit profile follows the default for its mode, so a new device is never left unconfigured.

Drive assignment across a fleet from a script: the management API. What the mode field changes: cloud or private. The on-prem second opinion in depth: the on-prem model.

Was this helpful?

One policy, every device.

Bundle mode, model and second opinion into a profile and assign it across the fleet. Set up in minutes.

Sign me up