mastra-ai/mastra · error · Error

No data source: provide datasetId or data

Error message

No data source: provide datasetId or data

What it means

runExperiment requires exactly one data source: either a storage-backed datasetId or inline `data`. When neither is supplied, this plain Error is thrown from the data-source resolution branch, and the run is marked failed during setup before any items execute.

Source

Thrown at packages/core/src/datasets/experiment/index.ts:296

          domain: 'STORAGE',
          category: 'USER',
        });
      }

      items = versionItems.map(v => ({
        id: v.id,
        datasetVersion: v.datasetVersion,
        input: v.input,
        groundTruth: v.groundTruth,
        expectedTrajectory: v.expectedTrajectory as TrajectoryExpectation | undefined,
        requestContext: v.requestContext,
        metadata: v.metadata,
        toolMocks: v.toolMocks,
        unmockedToolPolicy: v.unmockedToolPolicy,
        scorerIds: v.scorerIds,
      }));
    } else {
      throw new Error('No data source: provide datasetId or data');
    }
  } catch (err) {
    await markFailedOnSetupError(err);
    throw err; // unreachable, but satisfies TS control flow
  }

  // Phase B — Resolve task function
  let execFn: (item: ExperimentItem, signal?: AbortSignal) => Promise<ExecutionResult>;

  try {
    if (config.task) {
      // Inline task path
      const taskFn = config.task;
      execFn = async (item, itemSignal) => {
        try {
          const result = await taskFn({
            input: item.input,
            mastra,

View on GitHub (pinned to 75dd419e61)

Solutions

1. Pass inline `data: [...]` items to runExperiment
2. Or pass a `datasetId` pointing at a stored dataset
3. Validate the config object before calling (ensure data or datasetId is a non-empty value)
4. Fix the upstream code that was supposed to populate `data` (e.g. check the fetch/loading result)

Example fix

// before
const data = await loadData(); // may return undefined
await runExperiment({ ...config, data });
// after
const data = await loadData();
if (!data || data.length === 0) throw new Error('loadData returned no data');
await runExperiment({ ...config, data });

if (!config.datasetId && !(Array.isArray(config.data) && config.data.length > 0)) {
  throw new Error('runExperiment requires datasetId or non-empty data');
}

function hasDataSource(c: { datasetId?: string; data?: unknown[] }): boolean {
  return typeof c.datasetId === 'string' || (Array.isArray(c.data) && c.data.length > 0);
}

try {
  await runExperiment(config);
} catch (err) {
  if (err instanceof Error && err.message.includes('No data source')) {
    console.error('Experiment config missing both datasetId and data');
  } else throw err;
}

When it happens

Trigger: Calling runExperiment (or startExperiment/startExperimentAsync/executeRun) with a config that omits both datasetId and data.

Common situations: Building config programmatically where the data field ends up undefined (failed fetch assigned to data); spreading a partial options object; forgetting to pass data after removing a datasetId.

Understand the failure class

Background: "missing required argument" and "the following required arguments were not provided": what required-argument errors mean and how to fix them — this error's family across 20 libraries.

Related errors

• Unknown target type: ${targetType}
• No task: provide targetType+targetId or task
• AGENT_DURABLE_METHOD_NOT_AVAILABLE
• Unsupported output format: ${format}
• Experiment not found: ${experimentIdA}

AI-assisted analysis of mastra-ai/mastra@75dd419e61 (2026-08-30). Data as JSON: /api/errors/92f2d7fcff2645bf. Report an issue: GitHub.