Arcane.ollama.push()

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

Syntax

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

Parameters

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

Return value

Promise<OllamaPushResponse>

Description

Pushes only a model authorized by verified application policy.

Overview

Arcane.ollama.push(model, options?, streamOptions?) pushes an exact package-authorized verified model through the local Ollama provider. It can transfer model data to a configured registry. Use it only after explicit user authorization and provider/registry review; ordinary applications should not expose raw model publishing.

Parameters

model is required and must be both a valid Ollama name and an app-owned verified definition. The optional provider object contains only Boolean insecure; keep it false unless an approved development registry explicitly requires otherwise. Stream controls can be a named callback or {onChunk, signal, timeoutMs} and are not forwarded as provider fields.

Return value

It resolves to the bounded provider-native push envelope. With a chunk callback, the callback receives (chunk, {operation: "push", streamId}), and the promise resolves with the final provider chunk.

Availability

This is an exclusive desktop Core mutation requiring ai.models.manage and an exact verified-only package model. Unverified-model mode is inference-only. Android and browser previews do not expose it.

Errors and recovery

Invalid request/model errors require correction. Missing verified policy or an unknown app model is not retryable without a new package. Registry/provider failures can be retried only after confirming that a duplicate or partial remote publication is safe. Never place registry credentials in the request or logs.

Streaming, cancellation, and events

Chunks use wrapper-filtered ollama.chunk events. The default renderer timeout is 50 minutes. A supplied signal stops renderer observation, but push is not in Core's cooperatively cancellable method set; the host/provider mutation can continue. Aborting is not rollback, and no ArcaneOperation receipt is added to the provider-native result.

Example

The helper is deliberately not invoked. Call it only after an admitted UI has obtained explicit publishing confirmation.

async function pushVerifiedModelAfterConfirmation(model) {
    function reportPushChunk(chunk) {
        console.info(chunk.status ?? 'Push is running.');
    }

    return globalThis.Arcane.ollama.push(model, {insecure: false}, {
        onChunk: reportPushChunk,
        timeoutMs: 3000000
    });
}

Reference group

AI and Ollama