What causes "Dataset not found" in mastra-ai/mastra?

Verify the dataset id exists (list datasets) and matches the organizationId/projectId you pass.

mastra-ai/mastra · error · MastraError

DATASET_NOT_FOUND

DATASET_NOT_FOUND

Error message

Dataset not found

What it means

manager.get looks up a dataset by id (scoped by optional organizationId/projectId filters) via store.getDatasetById and throws DATASET_NOT_FOUND when no record matches. Raised in get and its callers (finalizeExperiment, resultA, resultB) when the referenced dataset doesn't exist or isn't visible in the given scope.

Solutions

Verify the dataset id exists (list datasets) and matches the organizationId/projectId you pass.
Catch DATASET_NOT_FOUND and surface a user-facing 'dataset missing' message or trigger re-creation.
Check scope: pass the same organizationId/projectId used at creation time.
For finalize/result flows, guard dataset lifetime so datasets aren't deleted while experiments reference them.

Example fix

// before
const ds = await datasets.get({ id: datasetId, organizationId, projectId }); // throws if absent
// after
let ds;
try {
  ds = await datasets.get({ id: datasetId, organizationId, projectId });
} catch (err) {
  if (err?.id === 'DATASET_NOT_FOUND') {
    ds = await datasets.create({ name: 'evals', organizationId, projectId });
  } else throw err;
}

Defensive patterns

Strategy: try-catch

Validate before calling

const existing = await datasets.list({ organizationId, projectId, /* name filter if available */ });
if (!existing.some(d => d.id === datasetId)) throw new Error(`Dataset ${datasetId} not found in scope`);

Type guard

function isDatasetNotFound(e: unknown): e is { id: 'DATASET_NOT_FOUND'; text: string } {
  return typeof e === 'object' && e !== null && (e as any).id === 'DATASET_NOT_FOUND';
}

Try / catch

try {
  return await datasets.get({ id, organizationId, projectId });
} catch (err) {
  if (isDatasetNotFound(err)) return null; // or re-create / report to user
  throw err;
}

Prevention

Pass the same organizationId/projectId scope used when the dataset was created.
Look up dataset ids dynamically (by name) instead of hardcoding across environments.
Prevent deleting datasets that active experiments reference (check finalizeExperiment flows).
Handle DATASET_NOT_FOUND as a first-class state in multi-tenant UIs.

When it happens

Trigger: Fetching a dataset with a wrong/deleted id; id from another project/organization passed without matching scope filters; dataset removed between creating an experiment and finalizing it.

Common situations: Hardcoded dataset ids drifting across environments; multi-tenant apps where projectId scoping hides the dataset; race with dataset deletion; using an experiment's datasetId after the dataset was purged.

Understand the failure class

Background: 'Could not be found', 'does not exist', 'not found in database': the resource-not-found family when an ID, slug, key, or URI lookup comes back empty — this error's family across 20 libraries.

Related errors

AI-assisted analysis of mastra-ai/mastra@75dd419e61 (2026-08-30). Data as JSON: /api/errors/3c6f74263ca662b8. Report an issue: GitHub.

Appendix: source

Thrown at packages/core/src/datasets/manager.ts:166

    const createScope = scopeFromArgs(input);
    return new Dataset(result.id, this.#mastra, createScope);
  }

  /**
   * Get an existing dataset by ID, optionally scoped to a tenant.
   *
   * When `organizationId` / `projectId` are provided, the read is scoped to
   * that tenancy: a dataset row that exists but belongs to a different tenant
   * returns NOT_FOUND (same 404 as a truly missing row) rather than leaking
   * cross-tenant existence. The returned {@link Dataset} handle carries the
   * scope forward on all subsequent reads and item mutations.
   */
  async get(args: { id: string; organizationId?: string; projectId?: string }): Promise<Dataset> {
    const store = await this.#getDatasetsStore();
    const scope = scopeFromArgs(args);
    const record = await store.getDatasetById({ id: args.id, filters: scope });
    if (!record) {
      throw new MastraError({
        id: 'DATASET_NOT_FOUND',
        text: 'Dataset not found',
        domain: 'STORAGE',
        category: 'USER',
      });
    }
    return new Dataset(args.id, this.#mastra, scope);
  }

  /**
   * List all datasets with pagination.
   *
   * Supports optional tenancy and candidate-identity filters. When omitted, all
   * datasets visible to the configured storage instance are returned.
   */
  async list(args?: {
    page?: number;
    perPage?: number;

View on GitHub (pinned to 75dd419e61)