Arcane.ollama.pull()

This focused page is derived from the mechanically checked full member inventory.

Syntax

Arcane.ollama.pull(model, options?, streamOptions?)

Parameters

model: OllamaModelName; options?: OllamaPullOptions; streamOptions?: OllamaStreamControls or OllamaChunkCallback

Return value

Promise<never>

Description

Direct application pull fails closed; Arcane's managed workflow owns integrity-bound preflight and admission.

Overview

Arcane.ollama.pull(model, options?, streamOptions?) is intentionally denied to applications. A raw mutable-tag pull lacks integrity-bound pre-download evidence and native resource admission. Use an admitted managed selection or application-model workflow; do not retry the raw method or call Ollama's loopback endpoint directly.

Parameters

The wrapper accepts a model name, optional {insecure}, and optional stream controls. Model names use the 1-256 character Ollama pattern and insecure would be provider-native registry control. These arguments do not relax the Core policy denial.

Return value

For application callers the method returns Promise<never>: it rejects before issuing a provider pull and therefore has no success envelope.

Availability

The vocabulary is present only at desktop Core's ai.models.manage boundary, but every admitted application call is denied. Verified-only policy returns MANAGED_MODEL_WORKFLOW_REQUIRED; unverified-model policy returns UNVERIFIED_MODEL_MUTATION_FORBIDDEN. Android and browser previews do not gain model-management authority.

Errors and recovery

Both policy errors are terminal for this raw method. Choose a model through Arcane.ollama.select() or another package-bound managed workflow that performs registry evidence, integrity verification, and resource admission. Do not weaken transport security with insecure as a workaround.

Streaming, cancellation, and events

Although the wrapper signature accepts stream controls, Core rejects before a provider request, so there are no chunks or operation events to cancel.

Example

This safe example demonstrates the required denial without starting a download.

try {
    await globalThis.Arcane.ollama.pull('gemma3:latest');
    throw new Error('Raw Ollama pull unexpectedly succeeded.');
} catch (error) {
    const expectedCodes = [
        'MANAGED_MODEL_WORKFLOW_REQUIRED',
        'UNVERIFIED_MODEL_MUTATION_FORBIDDEN'
    ];
    if (!expectedCodes.includes(error.code)) {
        throw error;
    }
    console.info('Use the managed Arcane model workflow.');
}

Reference group

AI and Ollama