Arcane.ollama.create()

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

Syntax

Arcane.ollama.create(request, options?)

Parameters

request: OllamaCreateRequest; options?: OllamaStreamControls or OllamaChunkCallback

Return value

Promise<OllamaCreateResponse>

Description

Creates a package-owned verified alias only after from, normalized system, and parameters match policy and files/adapters are absent; other admitted fields remain provider-native.

Overview

Arcane.ollama.create(request, streamOptions?) exposes a tightly restricted verified model-definition create path. Core accepts creation only for an exact app-owned model and rechecks the base model and resource admission. Prefer the managed selection or createBrain() workflow unless the application package explicitly owns this definition.

Parameters

The closed request is at most 8 MiB and requires a valid model. The direct parser recognizes from, files, adapters, template, license, system, parameters, messages, and quantize. The managed verified boundary requires from, normalized system, and parameters to match the package definition exactly and forbids files and adapters. Do not add other provider-native controls unless the package contract expressly owns them. Reserved Arcane and rollback aliases are denied. Stream controls are a callback or {onChunk, signal, timeoutMs}.

Return value

It resolves to Ollama's bounded provider-native create envelope. With onChunk, callbacks receive (chunk, {operation: "create", streamId}), and the promise resolves to the final chunk. This raw response does not include an ArcaneOperation.

Availability

This is an exclusive desktop Core mutation requiring ai.models.manage, verified-only package policy, and an exact app-owned definition. Unverified mode, Android, browser previews, and reserved aliases cannot use it.

Errors and recovery

MODEL_NOT_APP_VERIFIED, MODEL_DEFINITION_VERIFICATION_FAILED, managed- workflow, resource-admission, and unverified-mutation errors require the package-owned workflow rather than edited renderer input. Provider failure can leave uncertain local creation state; refresh model inventory and verification before retrying.

Streaming, cancellation, and events

Create chunks are filtered by the wrapper's stream ID. The default renderer timeout is 50 minutes. A supplied signal rejects renderer observation, but Core does not cooperatively cancel this raw create mutation; verify final local state after timeout or teardown.

Example

The helper is intentionally not invoked. Its definition must come from immutable package-owned policy, never user-edited JSON.

async function createVerifiedDefinition(definition) {
    function reportCreateChunk(chunk) {
        console.info(chunk.status ?? 'Create is running.');
    }

    return globalThis.Arcane.ollama.create({
        model: definition.name,
        from: definition.from,
        system: definition.system,
        parameters: definition.parameters
    }, {
        onChunk: reportCreateChunk,
        timeoutMs: 3000000
    });
}

Reference group

AI and Ollama