n8n-io/n8n · error · OperationalError

Unexpected embedding response from Bedrock

Error message

Unexpected embedding response from Bedrock

What it means

Thrown by BedrockInvokeModelEmbeddings.embed() when the model response body could not be coerced into a flat array of numbers after every known shape was tried: body.embeddingsByType (Titan typed selectors), body.embeddings as a raw array, and selectEmbeddingType(body.embeddings). It is an OperationalError, signalling a transient/unexpected upstream response shape rather than a user configuration fault.

Source

Thrown at packages/@n8n/nodes-langchain/nodes/embeddings/EmbeddingsAwsBedrock/BedrockInvokeModelEmbeddings.ts:84

					accept: 'application/json',
				}),
			);
			const body = jsonParse<EmbeddingResponseBody>(new TextDecoder().decode(response.body));
			if (Array.isArray(body.embedding)) {
				return body.embedding;
			}
			const titanTyped = selectEmbeddingType(body.embeddingsByType);
			if (titanTyped) {
				return titanTyped;
			}
			const rows = Array.isArray(body.embeddings)
				? body.embeddings
				: selectEmbeddingType(body.embeddings);
			const first = rows?.[0];
			if (Array.isArray(first)) {
				return first;
			}
			throw new OperationalError('Unexpected embedding response from Bedrock');
		});
	}

	async embedDocuments(documents: string[]): Promise<number[][]> {
		return await Promise.all(
			documents.map(async (document) => await this.embed(document, 'search_document')),
		);
	}

	async embedQuery(text: string): Promise<number[]> {
		return await this.embed(text, 'search_query');
	}
}

View on GitHub (pinned to 5ac6606e81)

Solutions

1. Confirm the selected Bedrock model id is an embeddings model (e.g. amazon.titan-embed-text-v2:0, cohere.embed-english-v3).
2. Reproduce the raw InvokeModel response with the AWS CLI (aws bedrock invoke-model) and compare its JSON keys against embeddings / embeddingsByType.
3. If the model uses a new shape, extend BedrockInvokeModelEmbeddings to handle it (add a branch before the throw) or switch to a supported model.
4. Retry once to rule out a truncated streaming response; if intermittent, inspect retries/timeout options on the node.

// Before relying on the embedding, validate the body shape
function extractEmbedding(body: any): number[] | null {
  const typed = selectEmbeddingType(body.embeddingsByType);
  if (typed) return typed;
  const rows = Array.isArray(body.embeddings) ? body.embeddings : selectEmbeddingType(body.embeddings);
  const first = rows?.[0];
  return Array.isArray(first) ? first : null;
}
const vec = extractEmbedding(body);
if (!vec) throw new Error('Embedding response shape not recognised; inspect body');

const isEmbeddingResponse = (b: unknown): b is { embeddings: number[][] } =>
  typeof b === 'object' && b !== null &&
  Array.isArray((b as any).embeddings) &&
  Array.isArray((b as any).embeddings[0]);

try {
  return extractEmbedding(body) ?? throwOperational();
} catch (e) {
  // log raw body for support, then retry once with a known-good model
  logger.warn({ body }, 'Unrecognised Bedrock embedding response');
  throw e;
}

When it happens

Trigger: An AWS Bedrock InvokeModel call returns an embedding payload whose structure differs from every shape the code understands — e.g. a Cohere or new model returning {data: [{embedding: [...]}]} without an embeddings/embeddingsByType field, a model returning a 200 with an empty body, or a region/model mismatch returning a non-embedding JSON object.

Common situations: Selecting a Bedrock model that does not output embeddings (e.g. a text-generation model chosen by mistake); AWS rolling out a response-format change; cross-region inference profile returning a wrapped payload; IAM/network layer truncating the body.

Related errors

• NVIDIA embeddings API returned ${data.length} embeddings for
• Additional Model Request Fields must be valid JSON
• Additional Model Request Fields must be a JSON object
• Cannot embed empty or undefined text
• Documents must be an array

AI-assisted analysis of n8n-io/n8n@5ac6606e81 (2026-08-12). Data as JSON: /api/errors/fc3d2fe8ec7b6ae4. Report an issue: GitHub.