> ## Documentation Index
> Fetch the complete documentation index at: https://aidocs.ethanbragdon.icu/llms.txt
> Use this file to discover all available pages before exploring further.

# Models API Reference

> Reference the Starfire AI models resource and the capability metadata applications should use before choosing a route.

# Models API reference

The models resource lets a developer discover the model catalog exposed to the current credential and account context.

## Resource

The Alpha v1 architecture defines:

```http theme={null}
GET /v1/models
```

Use the active OpenAPI/Developer Portal to confirm availability and the exact response schema.

## What a model record can describe

A model response can expose public metadata such as:

* model ID
* display name
* context/output limits
* reasoning support
* streaming support
* vision support
* tool support
* structured-output support
* research/FORGE eligibility where relevant

Private provider credentials, internal costs, and proprietary routing rules are not part of the public model resource.

## Choose by capability

Do not hardcode a model for a workflow solely because its name looks familiar. Check the capabilities required by the request.

For example, an image workflow needs vision eligibility and a FORGE build needs a route allowed for build execution.

## Availability is contextual

A credential can see a different effective catalog from another user because of plan, organization policy, rollout, or administrator configuration.

## Cache carefully

The model catalog can change. Cache for performance when appropriate, but do not assume a model that was available indefinitely remains enabled forever.

<Note>
  If a saved model becomes unavailable, handle the API error or refresh the catalog rather than silently replacing it with a capability-incompatible route.
</Note>