n8n-io/n8n · error · NodeOperationError

Index ${mongoVectorIndexName} not found

Error message

Index ${mongoVectorIndexName} not found

What it means

NodeOperationError (with remediation description) thrown when building the Atlas vector store client: the node lists search indexes on the collection and the user-supplied vector index name is not among them. MongoDB Atlas Vector Search requires a pre-created `vectorSearch` index before queries can run.

Source

Thrown at packages/@n8n/nodes-langchain/nodes/vector_store/VectorStoreMongoDBAtlas/VectorStoreMongoDBAtlas.node.ts:330

	sharedFields,
	async getVectorStoreClient(context, _filter, embeddings, itemIndex) {
		const client = await createMongoClient(context, context.getNode().typeVersion);
		try {
			const db = await getDatabase(context, client);
			const collectionName = getCollectionName(context, itemIndex);
			const mongoVectorIndexName = getVectorIndexName(context, itemIndex);
			const embeddingFieldName = getEmbeddingFieldName(context, itemIndex);
			const metadataFieldName = getMetadataFieldName(context, itemIndex);

			const collection = db.collection(collectionName);

			// test index exists
			const indexes = await collection.listSearchIndexes().toArray();

			const indexExists = indexes.some((index) => index.name === mongoVectorIndexName);

			if (!indexExists) {
				throw new NodeOperationError(context.getNode(), `Index ${mongoVectorIndexName} not found`, {
					itemIndex,
					description: 'Please check that the index exists in your collection',
				});
			}
			const preFilter = getFilterValue<IDataObject>(PRE_FILTER_NAME, context, itemIndex);
			const postFilterPipeline = getFilterValue<IDataObject[]>(
				POST_FILTER_NAME,
				context,
				itemIndex,
			);

			return new ExtendedMongoDBAtlasVectorSearch(
				embeddings,
				{
					collection,
					indexName: mongoVectorIndexName, // Default index name
					textKey: metadataFieldName, // Field containing raw text
					embeddingKey: embeddingFieldName, // Field containing embeddings

View on GitHub (pinned to 5ac6606e81)

Solutions

1. In the Atlas UI, create a Search Index of type `vectorSearch` on the target collection with the configured name.
2. Match the node's `vectorIndexName` field exactly to the index name shown in Atlas.
3. Wait for the index to finish building (status `ACTIVE`) before running the node.
4. Confirm the right database and collection are selected so the index lookup hits the right place.

Example fix

// before
const indexExists = indexes.some((index) => index.name === mongoVectorIndexName);
if (!indexExists) {
  throw new NodeOperationError(context.getNode(), `Index ${mongoVectorIndexName} not found`, {
    itemIndex,
    description: 'Please check that the index exists in your collection',
  });
}

// after: tell the user which indexes DO exist so they can pick the right one
const known = indexes.map((i) => i.name).filter(Boolean) as string[];
if (!indexExists) {
  throw new NodeOperationError(context.getNode(), `Index ${mongoVectorIndexName} not found`, {
    itemIndex,
    description: known.length
      ? `Available indexes: ${known.join(', ')}. Create a 'vectorSearch' index named '${mongoVectorIndexName}' in Atlas if missing.`
      : 'No search indexes exist on this collection. Create a vectorSearch index in Atlas.',
  });
}

// Pre-flight: confirm the index exists (and is ready) before the workflow runs.
async function ensureVectorIndex(collection: Collection, indexName: string): Promise<void> {
  const indexes = await collection.listSearchIndexes().toArray();
  const match = indexes.find((i) => i.name === indexName);
  if (!match) throw new Error(`Index ${indexName} not found. Create a vectorSearch index in Atlas.`);
  if (match.status && !/ready|active/i.test(String(match.status))) {
    throw new Error(`Index ${indexName} is not ready (status: ${match.status}).`);
  }
}

function hasNamedIndex(indexes: { name?: string }[], name: string): boolean {
  return indexes.some((i) => i.name === name);
}

const indexes = await collection.listSearchIndexes().toArray();
if (!hasNamedIndex(indexes, mongoVectorIndexName)) {
  throw new NodeOperationError(context.getNode(), `Index ${mongoVectorIndexName} not found`, {
    itemIndex,
    description: `Available: ${indexes.map((i) => i.name).filter(Boolean).join(', ') || 'none'}`,
  });
}

When it happens

Trigger: Running the node in Retrieve / Load mode before the vector index has been created in the Atlas UI; the index was created with a different name than what is configured in the node; the index is still building (`status: 'BUILDING'`) and not yet listed as active.

Common situations: New Atlas cluster where the user created the collection but forgot the search index; index named `default` but the node expects `vector_index`; index recently deleted; wrong database/collection selected so the lookup runs against an empty collection.

Related errors

• Index ${index} not found
• Error: ${error.message}
• Parameter ${key} must be a string
• Error: No JSON string provided.
• Redis client not initialized

AI-assisted analysis of n8n-io/n8n@5ac6606e81 (2026-08-12). Data as JSON: /api/errors/03c405fbec3ce630. Report an issue: GitHub.