AALLM Docs

Explore the deployment registry

Use the registry to move from a vague model name to the exact provider deployment your product can evaluate and govern.

In the app:RegistryOpen screen

Why the registry matters

A model release can be exposed by several providers, regions, and API surfaces. Those deployments can differ in price, availability, limits, and lifecycle. The registry gives you the deployment-level identity needed for reliable cost and compatibility decisions.

Choose a deployment

  1. Open Registry and use Filter records with a model, provider, region, lifecycle status, or capability value.
  2. Confirm the canonical model_id and the provider-specific provider_model_id.
  3. Check region, mode, context window, lifecycle status, and the presence of a rate card.
  4. Copy the deployment public ID. It is the stable input used by quotes, policies, lockfiles, probes, and API calls.

The app shows the first 250 catalog deployments

Use the public registry or API filters for a larger or more targeted catalog query.

What to inspect

model_id
canonical identity
The lab and model release, independent of the serving provider.
provider_id
serving identity
The cloud, lab, or gateway exposing the model.
region
constraint
Where the deployment is exposed; verify this against residency requirements.
lifecycle_status
risk
Whether the deployment is available, deprecated, retired, or otherwise constrained.
context_tokens
capacity
The published input context limit used in compatibility decisions.
rate_card
cost evidence
The public rates and source evidence available for deterministic quotes.

Use the deployment ID

Calculate a representative quote, test the deployment against a compatibility policy, then include it in a lockfile only after the decision is approved.

Building the same workflow in code?