Reference
Arcane.ai.models()
Lists raw local model inventory for Settings, Terminal, and Shell diagnostics. Application model selectors use the admitted Arcane.localAI.status().models.ollama catalog instead.
This focused page is derived from the mechanically checked full member inventory.
Syntax
Arcane.ai.models()
Parameters
None
Return value
Description
Lists raw local model inventory for Settings, Terminal, and Shell diagnostics. Application model selectors use the admitted Arcane.localAI.status().models.ollama catalog instead.
Overview
Arcane.ai.models() reads normalized installed-model diagnostics from the
Arcane-managed Ollama service. Use it in diagnostic UI, not to populate an
application model selector. Application inference choices must come from the
admitted Arcane.localAI.status().models.ollama catalog.
Parameters
The method takes no arguments.
Return value
It resolves to LocalAIInventory: {provider: "arcane-ollama", models}. The
bounded models array has at most 512 records with name, modifiedAt,
sizeBytes, digest, family, parameterSize, and quantization fields.
Availability
This is a desktop Core diagnostic API. It requires ai.models.read and is
admitted only to Settings, Shell, and Terminal. Browser previews and the
current Android projection do not expose it.
Errors and recovery
An unavailable or malformed local provider can reject with a local-Ollama
request or LOCAL_OLLAMA_INVALID_RESPONSE error. Recheck service health before
retrying. METHOD_NOT_ALLOWED requires the correct admitted diagnostic app;
retrying from another app does not add authority.
Streaming, cancellation, and events
This is one non-streaming inventory read. It emits no method event and exposes
no per-call AbortSignal. The renderer's ordinary request timeout or page
teardown can stop waiting without promising to cancel host work.
Example
const inventory = await globalThis.Arcane.ai.models();
for (const model of inventory.models) {
console.info(model.name, model.parameterSize ?? 'unknown size');
}