vercel/ai @ai-sdk/google@2.0.103

on GitHub

Patch Changes

• 7c28815: feat(google): add gemini-4-argon model id and long decode support: requests that reach the per-request cap (CONTINUATION) are continued automatically with usage summed across requests and the first request's prompt token count exposed as providerMetadata.google.initialPromptTokenCount, and the continuationToken provider option / provider metadata resume responses that used up their maxOutputTokens budget