Query model catalogs and runtime status¶
Browse model capabilities and prices, inspect an individual model’s health checks, or compare runtime metrics for the last hour. Catalog visibility does not establish calling permission; list callable models before making an inference request.
Prerequisites¶
Prepare a personal access token to query the current user’s model catalog and runtime status. The public model catalog and public overview do not require authentication. For authentication and query scope, see Usage and monitoring.
Call sequence¶
Query the public model catalog to browse public models. If you only need the model count and compatible API entry points, query the public overview instead. These queries can be used independently.
Query the model catalog for information visible to the current user. Query filter options when building catalog filters, or selection options when you only need a candidate list.
To inspect an individual model, use its catalog record identifier to query model health and review the aggregate status and recent channel checks.
To compare recent activity, query model runtime status and inspect call counts, latency, and error rates for the last hour.
API operations¶
Browse model catalogs¶
Task |
API operation |
When to use it or what to confirm |
|---|---|---|
Browse models visible to the current user |
Obtain capabilities, prices, and catalog record identifiers. |
|
Browse public models |
Inspect enabled public models without authentication. |
|
Obtain model candidates |
Populate a model selection control. |
|
Obtain catalog filters |
Inspect counts by model type, family, and pricing status. |
|
Inspect the public overview |
Obtain the public model count and compatible API entry points. |
Inspect health and runtime metrics¶
Task |
API operation |
When to use it or what to confirm |
|---|---|---|
Inspect one model |
Use a catalog record identifier to inspect health checks. |
|
Compare recent activity |
Inspect call counts, latency, error rates, and health for the last hour. |
Confirm the result¶
Locate the target model in List the model catalog, then inspect the same catalog record with Get model health. Runtime metrics describe recent activity; they do not replace checking model authorization for the calling credential.
Limitations and notes¶
A catalog record identifier and the name used for inference serve different purposes. Use the identifier required by each operation.
The catalog, selection options, and filter options do not support identical filters. Do not reuse every filter across them.
Runtime status uses the last hour. Other filters affect metric aggregation, not the outer set of enabled models.