n8n-io/n8n · error · UnexpectedError

Unexpected score type: ${typeof score}

Error message

Unexpected score type: ${typeof score}

What it means

An UnexpectedError thrown while mapping hybridSearch results: each document's metadata.score must be a number, and if it is not (string, undefined, object) the mapping aborts. This fires only on the hybrid search branch (args.hybridQuery) where returnMetadata may be ['explainScore']. It signals the Weaviate response shape differs from what n8n expects.

Source

Thrown at packages/@n8n/nodes-langchain/nodes/vector_store/VectorStoreWeaviate/VectorStoreWeaviate.node.ts:82

		if (args.hybridQuery) {
			const options = {
				limit: k ?? undefined,
				autoLimit: args.autoCutLimit ?? undefined,
				alpha: args.alpha ?? undefined,
				vector: query,
				filter: filter ? parseCompositeFilter(filter as WeaviateCompositeFilter) : undefined,
				queryProperties: args.queryProperties
					? args.queryProperties.split(',').map((prop) => prop.trim())
					: undefined,
				maxVectorDistance: args.maxVectorDistance ?? undefined,
				fusionType: args.fusionType,
				returnMetadata: args.hybridExplainScore ? ['explainScore'] : undefined,
			};
			const content = await super.hybridSearch(args.hybridQuery, options);
			return content.map((doc) => {
				const { score, ...metadata } = doc.metadata;
				if (typeof score !== 'number') {
					throw new UnexpectedError(`Unexpected score type: ${typeof score}`);
				}
				return [
					new Document({
						pageContent: doc.pageContent,
						metadata,
					}),
					score,
				] as [Document, number];
			});
		}
		return await super.similaritySearchVectorWithScore(
			query,
			k,
			filter ? parseCompositeFilter(filter as WeaviateCompositeFilter) : undefined,
		);
	}
}

View on GitHub (pinned to 5ac6606e81)

Solutions

1. If you enabled hybridExplainScore, disable it for the retrieval call that needs numeric scores (this branch only exists because of the hybrid path).
2. Update n8n to the latest release in case the metadata-shape handling was patched.
3. Check the Weaviate server version against the version n8n's bundled client expects; align them.
4. If you need explain output, consume it separately from the scoring path so the numeric score mapping is unaffected.

Example fix

// before: options.hybridExplainScore = true  (score becomes explain object)
// after:  options.hybridExplainScore = false (metadata.score is a number again)

// guard the mapping yourself before relying on numeric scores
const score = (doc.metadata as any).score;
if (typeof score !== 'number') { /* fallback to similaritySearchVectorWithScore, or disable hybridExplainScore */ }

function hasNumericScore(m: unknown): m is { score: number } { return typeof (m as any)?.score === 'number'; }

try { results = await store.hybridSearch(...) } catch (e) { if (/Unexpected score type/.test(e.message)) { results = await store.similaritySearchVectorWithScore(query, k); } else throw e; }

When it happens

Trigger: hybridSearch returns documents whose metadata.score is not a number — e.g. when hybridExplainScore is enabled and Weaviate returns an explanation object/string instead of a numeric score, or when a weaviate-ts/langchain version change alters metadata.score's type.

Common situations: Enabling hybridExplainScore changes the returned metadata so score is no longer numeric; weaviate server version returns score under a different key; langchain Document.metadata typing drift after an upgrade.

Related errors

• Filter operator "${operator}" on key "${key}" requires array
• Filter operator "${operator}" on key "${key}" requires a str
• Metadata key "${CONTENT_KEY}" is reserved for the document c
• Metadata value for key "${key}" is unsupported: Pinecone onl
• Filter operator "${operator}" on key "${key}" requires all a

AI-assisted analysis of n8n-io/n8n@5ac6606e81 (2026-08-12). Data as JSON: /api/errors/61533a3a956027e9. Report an issue: GitHub.