mastra-ai/mastra · error · Error

structuredOutput object is undefined

Error message

structuredOutput object is undefined

What it means

generateWithStructuredOutput calls the language model requesting structured output and asserts the result contains an `object`. When the model response's structured output field is undefined — despite resolving without throwing — this error is raised instead of returning undefined.

Source

Thrown at packages/memory/src/processors/observational-memory/extraction-runner.ts:81

## Extractors

${extractorInstructions}${priorLines.length > 0 ? `\n\n## Prior Extracted Values\n\n${priorLines.join('\n\n')}` : ''}`;

  const values: Record<string, unknown> = {};
  const failures: Array<{ slug: string; error: string }> = [];

  const generateWithStructuredOutput = async (jsonPromptInjection?: boolean | 'system' | 'inline') => {
    const output = await opts.agent.generate(prompt, {
      structuredOutput: { schema, ...(jsonPromptInjection ? { jsonPromptInjection } : {}) },
      ...(opts.memory ? { memory: opts.memory } : {}),
      ...(opts.abortSignal ? { abortSignal: opts.abortSignal } : {}),
      ...(opts.requestContext ? { requestContext: opts.requestContext } : {}),
      ...opts.observabilityContext,
    });

    if (output.object === undefined) {
      throw new Error('structuredOutput object is undefined');
    }

    return output.object;
  };

  let object: Record<string, unknown>;
  let retryEmptyObject = false;
  try {
    object = await generateWithStructuredOutput();
    retryEmptyObject = shouldRetryEmptyStructuredObject(object, structuredExtractors);
  } catch (error) {
    if (isAbortError(error, opts.abortSignal)) {
      throw error;
    }

    try {
      const fallbackJsonPromptInjection = coreFeatures.has('json-prompt-injection:inline') ? 'inline' : true;
      object = await generateWithStructuredOutput(fallbackJsonPromptInjection);

View on GitHub (pinned to 75dd419e61)

Solutions

1. Switch to a model that supports structured output reliably (or enable JSON mode) in the extraction options
2. Simplify the extraction schema / instructions so the model returns valid JSON
3. Check for truncation (raise maxOutputTokens) and verify the provider response shape; add a retry around extraction

Example fix

// before
const values = await extractStructuredValues({ model: smallLocalModel, ... });

// after
const values = await extractStructuredValues({
  model: 'gpt-4o', // supports structured output
  maxOutputTokens: 4096,
  ...
});

null

try {
  object = await generateWithStructuredOutput(opts);
} catch (err) {
  if (err instanceof Error && err.message === 'structuredOutput object is undefined') {
    object = await generateWithStructuredOutput({ ...opts, model: structuredOutputCapableModel });
  } else {
    throw err;
  }
}

When it happens

Trigger: The LLM call completes but `output.object` is undefined: model ignored the JSON-schema instruction, the provider returned an empty/malformed structured result, or the response was truncated/aborted in a way that still resolved.

Common situations: Using a model that does not reliably support native structured output/JSON mode; very small or weak models with complex extraction schemas; provider API changes returning a different response shape; low maxOutputTokens truncating the JSON before it parses.

Related errors

• Structured output returned no object
• CursorSDKAgent does not support structuredOutput because the
• @mastra/livekit: reply generation runs through the Mastra ag
• Failed to rename session (${res.status})
• No result received from agent execution on iteration ${itera

AI-assisted analysis of mastra-ai/mastra@75dd419e61 (2026-08-30). Data as JSON: /api/errors/10c359bf6a0e142b. Report an issue: GitHub.