mastra-ai/mastra · error · MastraError

MASTR_SCORER_FAILED_TO_RUN_WORKFLOW_FAILED

MASTR_SCORER_FAILED_TO_RUN_WORKFLOW_FAILED

Error message

Scorer Run Failed: ${workflowFailure.message}

What it means

When a MastraScorer's underlying workflow run fails, run() inspects the failure state and wraps the workflow's error message in a MASTR_SCORER_FAILED_TO_RUN_WORKFLOW_FAILED MastraError (SCORER domain, USER category), including completed steps and the failed step in details. It indicates a step inside the scorer's internal pipeline threw, rather than a config problem.

Source

Thrown at packages/core/src/evals/base.ts:1075

          cause: workflowFailure,
        });
      }
      throw error;
    }

    if (workflowResult.status === 'failed') {
      const workflowFailure = getErrorFromUnknown(workflowResult.error, {
        fallbackMessage: 'Scorer workflow failed',
      });
      const failedJudgeExecution = takeFailedJudgeExecution(workflowFailure);
      const failedStepFromError = takeFailedScorerStep(workflowFailure);
      const failureState = this.getWorkflowFailureState(workflowResult);
      const failedStep = failedStepFromError ?? failureState.failedStep;
      const { completedSteps, latestSuccessfulOutput } = failureState;
      evalSpan?.error({ error: workflowFailure, endSpan: true });

      if (!failedStep) {
        throw new MastraError(
          {
            id: 'MASTR_SCORER_FAILED_TO_RUN_WORKFLOW_FAILED',
            domain: ErrorDomain.SCORER,
            category: ErrorCategory.USER,
            text: `Scorer Run Failed: ${workflowFailure.message}`,
            details: {
              scorerId: this.config.id ?? this.config.name,
              steps: this.steps.map(s => s.name).join(', '),
            },
          },
          workflowFailure,
        );
      }

      const finalStepResult = failedJudgeExecution
        ? this.appendFailedJudgeExecution(latestSuccessfulOutput, failedStep, failedJudgeExecution)
        : latestSuccessfulOutput;
      const result = this.hasScorerResultFields(finalStepResult)

View on GitHub (pinned to 75dd419e61)

Solutions

1. Read the wrapped workflowFailure.message and the failedStep in error.details to find which step threw, then fix that step's root cause.
2. Verify LLM credentials, model availability, and rate limits if the failed step calls a model.
3. Harden custom steps (preprocess/generateScore) with input validation and try/catch so transient issues don't fail the whole run.
4. Catch this error at the call site to mark the evaluation run as failed instead of crashing the batch.

Example fix

// before
const result = await scorer.run({ input, runId }); // crashes batch on step failure
// after
try {
  const result = await scorer.run({ input, runId });
} catch (e) {
  logger.error('Scorer step failed', { scorerId: 's1', cause: e.message, details: e.details });
}

// validate inputs/credentials each step depends on before running
if (!process.env.OPENAI_API_KEY) throw new Error('Missing OPENAI_API_KEY required by scorer steps');

function isScorerWorkflowFailure(e: unknown): e is MastraError & { id: 'MASTR_SCORER_FAILED_TO_RUN_WORKFLOW_FAILED' } {
  return e instanceof MastraError && e.id === 'MASTR_SCORER_FAILED_TO_RUN_WORKFLOW_FAILED';
}

try {
  const result = await scorer.run({ input, runId });
} catch (e) {
  if (isScorerWorkflowFailure(e)) {
    logger.error(`Scorer step failed: ${e.details?.failedStep}`, { completedSteps: e.details?.completedSteps, cause: e.message });
    return; // mark run failed, continue batch
  }
  throw e;
}

When it happens

Trigger: Calling scorer.run(...) where any workflow step (preprocess, generateScore, etc.) throws — the workflow returns a failure and there is a resolvable failedStep (explicitly or from the failure state) — so the raw step error is rethrown with 'Scorer Run Failed: <message>'.

Common situations: A generateScore step calling an LLM that errors (bad API key, rate limits, model outage); a preprocess step throwing on unexpected input data; a custom step with a bug; network failures during score generation.

Related errors

• RUN_EXPERIMENT_SCORER_FAILED_TO_SCORE_STEP_RESULT
• RUN_EXPERIMENT_SCORER_FAILED_TO_SCORE_WORKFLOW_TRAJECTORY
• MASTR_SCORER_FAILED_TO_CREATE_MISSING_ID
• MASTR_SCORER_FAILED_TO_RUN_MISSING_GENERATE_SCORE
• RUN_EXPERIMENT_TARGET_FAILED_TO_GENERATE_RESULT

AI-assisted analysis of mastra-ai/mastra@75dd419e61 (2026-08-30). Data as JSON: /api/errors/4d59fd7009e68e02. Report an issue: GitHub.