n8n-io/n8n · error · GuardrailError

Guardrail validation failed: ${error instanceof Error ? erro

Error message

Guardrail validation failed: ${error instanceof Error ? error.message : 'Unknown error'}

What it means

Catch-all GuardrailError thrown by the Guardrails model helper for any error during LLM guardrail validation that is not a LangChain parser error. The message is 'Guardrail validation failed: ' plus the underlying error.message (or 'Unknown error' for non-Error throws). It is the fall-through branch after the parser-error check.

Source

Thrown at packages/@n8n/nodes-langchain/nodes/Guardrails/helpers/model.ts:126

				return content[0].text as string;
			}
			throw new Error('Invalid content type');
		};

		const text = extractText(result.content);
		const { confidenceScore, flagged } = await outputParser.parse(text);

		// Validate output consistency
		if (typeof confidenceScore !== 'number' || typeof flagged !== 'boolean') {
			throw new GuardrailError(name, 'Invalid output format', 'Expected number and boolean fields');
		}

		return { confidenceScore, flagged };
	} catch (error) {
		if (isLangChainParserError(error)) {
			throw new GuardrailError(name, 'Failed to parse output', MODEL_OUTPUT_PARSER_ERROR_MESSAGE);
		}
		throw new GuardrailError(
			name,
			`Guardrail validation failed: ${error instanceof Error ? error.message : 'Unknown error'}`,
			error?.description,
		);
	}
}

export async function runLLMValidation(
	name: string,
	inputText: string,
	{ model, prompt, threshold, systemMessage }: LLMConfig,
): Promise<GuardrailResult> {
	try {
		const result = await runLLM(name, model, prompt, inputText, systemMessage);
		const triggered = result.flagged && result.confidenceScore >= threshold;
		return {
			guardrailName: name,
			tripwireTriggered: triggered,

View on GitHub (pinned to 5ac6606e81)

Solutions

1. Read the appended error.message to distinguish auth/network/rate-limit causes.
2. For rate limits, add backoff/retry at the model client or workflow level.
3. Verify the model credentials and endpoint configuration in the connected Chat Model node.
4. Wrap the guardrail in retry logic and fail closed only after retries are exhausted.

function isGuardrailValidationFailure(e) {
  return e instanceof GuardrailError && e.message.startsWith('Guardrail validation failed:');
}

try { ... } catch (e) {
  if (e instanceof GuardrailError && e.message.startsWith('Guardrail validation failed:')) {
    // inspect e.message suffix for auth/rate-limit/network; retry or fail-closed
  } else throw e;
}

When it happens

Trigger: Inside the try block around model.invoke/parse, any thrown error that is not an OutputParserException reaches the second throw branch: model API failures, network errors, auth errors, unexpected exceptions from the model client, or non-Error rejections.

Common situations: Model provider rate limit or 401 auth error; network timeout reaching the LLM endpoint; langchain model client bug; a custom model client that throws plain objects; revoked API credentials.

Related errors

• error?.description || error?.message
• error?.description || error?.message || 'Unknown error'
• Invalid output format
• Failed to parse output
• Error executing tool: ${(error as Error).message}

AI-assisted analysis of n8n-io/n8n@5ac6606e81 (2026-08-12). Data as JSON: /api/errors/985c70076f200207. Report an issue: GitHub.