vercel/ai · error

Not supported. You must use a /sync/v1 endpoint for chat mod

Error message

Not supported. You must use a /sync/v1 endpoint for chat models.

What it means

Baseten chat models must be served through a /sync/v1 (OpenAI-compatible) endpoint. If a custom modelURL contains '/predict', the provider throws, because /predict endpoints have a non-OpenAI request/response shape unsupported for chat.

Source

Thrown at packages/baseten/src/baseten-provider.ts:209

      return `${customURL || baseURL}${path}`;
    },
    headers: getHeaders,
    fetch: options.fetch,
  });

  const createChatModel = (modelId?: BasetenChatModelId) => {
    const customURL = options.modelURL;
    if (customURL) {
      if (customURL.includes('/sync/v1')) {
        return new OpenAICompatibleChatLanguageModel(modelId ?? 'placeholder', {
          ...getCommonModelConfig('chat', customURL),
          errorStructure: basetenErrorStructure,
          // Or stream_options.include_usage is omitted and streams report no usage.
          includeUsage: true,
          supportsStructuredOutputs: true,
        });
      } else if (customURL.includes('/predict')) {
        throw new Error(
          'Not supported. You must use a /sync/v1 endpoint for chat models.',
        );
      }
    }

    return new OpenAICompatibleChatLanguageModel(modelId ?? 'chat', {
      ...getCommonModelConfig('chat'),
      errorStructure: basetenErrorStructure,
      includeUsage: true,
      supportsStructuredOutputs: true,
    });
  };

  const createEmbeddingModel = (modelId?: BasetenEmbeddingModelId) => {
    const customURL = options.modelURL;
    if (!customURL) {
      throw new Error(
        'No model URL provided for embeddings. Please set modelURL option for embeddings.',

View on GitHub (pinned to 69428b1f8b)

Solutions

1. Change modelURL to the deployment's /sync/v1 endpoint, e.g. https://model-<id>.api.baseten.co/production/sync/v1.
2. Deploy the model as an OpenAI-compatible sync deployment on Baseten.
3. Omit modelURL to use the default chat base URL.

Example fix

// before
const baseten = createBaseten({ apiKey, modelURL: 'https://model-abc.api.baseten.co/production/predict' });
// after
const baseten = createBaseten({ apiKey, modelURL: 'https://model-abc.api.baseten.co/production/sync/v1' });

if (options.modelURL && options.modelURL.includes('/predict')) {
  throw new Error('Chat requires a /sync/v1 modelURL on Baseten');
}

try {
  const model = baseten('gpt-4o');
} catch (e) {
  if (e instanceof Error && e.message.includes('/sync/v1 endpoint')) {
    // fix modelURL config
  }
}

When it happens

Trigger: createBaseten({ modelURL: 'https://.../models/<id>/predict' }) then calling provider() or provider.chatModel — any chat model creation with a /predict URL.

Common situations: Reusing the URL of a Baseten Model API (which uses /predict) for chat; copying a model-page URL instead of the sync inference URL.

Related errors

• Not supported. You must use a /sync or /sync/v1 endpoint for
• No model URL provided for embeddings. Please set modelURL op
• Invalid argument for parameter model: model ${model.provider
• Unsupported output: ${_exhaustiveCheck}
• Invalid argument for parameter output: Invalid output type.

AI-assisted analysis of vercel/ai@69428b1f8b (2026-08-30). Data as JSON: /api/errors/77b3ba43a48bd915. Report an issue: GitHub.