ODOCK.AI
Security & GuardrailsTutorials

Configure Model Policies

Add resource-specific policy limits to a model.

Configure Model Policies

Use model policies when a model needs a stronger runtime envelope because it is expensive, sensitive, slow, or intended for a narrow workload.

Open Models in your organisation.

Open the model detail page.

Confirm the model name, slug, provider, provider key, pricing, and access grants are correct.

For model concepts, see Models.

Find the Policies card and click Edit.

Set model-specific limits.

Common model controls:

FieldWhy use it
Tokens per minuteProtects high-cost or limited-capacity models.
Max tokens per requestPrevents unusually large generations.
Requests per minuteSmooths traffic to slower or rate-limited providers.
Max concurrencyPrevents too many simultaneous calls to this model.

Save policies.

Test through an API key that has model access.

The request must pass the organisation, team if present, API key, and model policies. If any one scope blocks, the request stops before or during gateway enforcement.

Why This Works

Model policies are added after Odock resolves the requested model. That means they are useful for resource-specific limits, but they are not a replacement for API key access grants.

For calling models through Odock, see Call a model through Odock.

On this page