Arcane.ai.saveProviderSettings()

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

Syntax

Arcane.ai.saveProviderSettings(settings)

Parameters

settings: AIProviderSettingsInput

Return value

Promise<AIProviderSettingsResult>

Description

Saves provider settings; validates the selected model when OpenAI is selected and validates a supplied replacement credential before persistence. Ollama-selected settings may retain a dormant OpenAI model for later revalidation.

Overview

Arcane.ai.saveProviderSettings(settings) applies a provider-settings patch. Use it only from Settings after explicit user intent. Arcane protects OpenAI credentials and attempts to roll back credential and settings changes together when a save fails.

Parameters

The patch can contain provider, openAIModel, responseLength, token, and removeToken; omitted keys preserve current values. provider is "ollama" or "openai". openAIModel is 1-128 characters matching [A-Za-z0-9][A-Za-z0-9._:-]{0,127}. responseLength is "low", "medium", or "high". A replacement token is trimmed, begins with sk-, and is 20-512 characters. Only removeToken: true requests deletion, and a supplied token becomes the final credential. Arcane performs account-model validation when OpenAI is selected or a replacement token is supplied.

Return value

It resolves to the credential-free settings shape {provider, openAIModel, openAIConfigured, responseLength} after persistence.

Availability

This is an exclusive desktop Core mutation requiring ai.settings.manage and the Settings app identity. OpenAI validation uses the network and protected credential store. It is not projected to Android.

Errors and recovery

Correct INVALID_AI_PROVIDER_SETTINGS, provider, model, response-length, or token errors before retrying. OpenAI account/model errors require a valid credential and accessible model. On CREDENTIAL_STORE_UNAVAILABLE, leave the prior settings in place. CREDENTIAL_ROLLBACK_FAILED requires diagnostics and manual Settings review before another mutation. Never log a supplied token.

Streaming, cancellation, and events

The method does not stream or emit operation events. The wrapper waits up to 130 seconds and exposes no signal. Core does not cooperatively cancel this mutation, so closing the page or timing out can stop observation while the save continues; read providerSettings() before deciding whether to retry.

Example

This credential-free example selects local inference. Read a new OpenAI token from a protected user input at call time rather than embedding it in source.

document.querySelector('#confirm-provider-settings')?.addEventListener(
    'click',
    async function handleConfirmedProviderSettings() {
        const saved = await globalThis.Arcane.ai.saveProviderSettings({
            provider: 'ollama',
            responseLength: 'medium'
        });
        console.info(saved.provider, saved.responseLength);
    }
);

Reference group

AI and Ollama