vercel/ai · error · APICallError

${response.error.message}

Error message

${response.error.message}

What it means

After calling the Hugging Face Responses API in doGenerate, if the response body contains an error object, the model throws APICallError with the API's error message, a fabricated 400 status code, and the raw response body attached. This means the remote API rejected or failed the request; the SDK surfaces the upstream error text verbatim.

Source

Thrown at packages/huggingface/src/responses/huggingface-responses-language-model.ts:206

    const {
      value: response,
      responseHeaders,
      rawValue: rawResponse,
    } = await postJsonToApi({
      url,
      headers: combineHeaders(this.config.headers?.(), options.headers),
      body,
      failedResponseHandler: huggingfaceFailedResponseHandler,
      successfulResponseHandler: createJsonResponseHandler(
        huggingfaceResponsesResponseSchema,
      ),
      abortSignal: options.abortSignal,
      fetch: this.config.fetch,
    });

    if (response.error) {
      throw new APICallError({
        message: response.error.message,
        url,
        requestBodyValues: body,
        statusCode: 400,
        responseHeaders,
        responseBody: rawResponse as string,
        isRetryable: false,
      });
    }

    const content: Array<LanguageModelV4Content> = [];

    // Process output array
    for (const part of response.output) {
      switch (part.type) {
        case 'message': {
          for (const contentPart of part.content) {
            content.push({

View on GitHub (pinned to 69428b1f8b)

Solutions

1. Read the error message/responseBody in the thrown APICallError (and its cause/headers) to see the exact upstream failure.
2. Verify the model ID exists and your HF token has access to it (check status/quota on huggingface.co).
3. Fix the request payload per the message (e.g. unsupported parameters, oversized input) and retry.

Example fix

// before
const result = await generateText({ model: huggingface.responses('nonexistent/model'), prompt });
// after
const result = await generateText({ model: huggingface.responses('meta-llama/Llama-3.1-8B-Instruct'), prompt });

if (!modelId || !modelId.includes('/')) {
  throw new Error(`Invalid Hugging Face model id: '${modelId}'. Use 'namespace/model' format.`);
}
if (!process.env.HF_TOKEN && !apiKey) {
  console.warn('No Hugging Face token configured; API calls may fail.');
}

import { APICallError } from '@ai-sdk/provider';
function isHfApiError(e: unknown): e is APICallError {
  return APICallError.isInstance(e) && typeof e.responseBody === 'string' && e.responseBody.includes('error');
}

import { APICallError } from '@ai-sdk/provider';
try {
  return await generateText({ model, prompt });
} catch (e) {
  if (APICallError.isInstance(e) && e.statusCode === 400) {
    console.error('Hugging Face API error:', e.message, e.responseBody);
    // inspect responseBody; do not blindly retry non-retryable 400s
  }
  throw e;
}

When it happens

Trigger: Any generateText/streamText call where the Hugging Face Responses API returns a JSON body with an error field — e.g. invalid model ID, malformed input, exceeded quota, or content policy rejection.

Common situations: Wrong or deprecated model ID; missing/invalid HF token causing auth errors reported in the body; prompt payload violating API constraints; Hugging Face service-side errors during outages.

Related errors

• Video generation timed out after ${timeoutMs}ms.
• The response body is empty.
• Incomplete Amazon Bedrock event-stream frame: ${buffer.lengt
• Failed to fetch the response.
• The response body is empty.

AI-assisted analysis of vercel/ai@69428b1f8b (2026-08-30). Data as JSON: /api/errors/1fad8c49a6c40cee. Report an issue: GitHub.