object: inconsistent vector lengths found: dimensions=%d and

Error message

object: inconsistent vector lengths found: dimensions=%d and object=%d

What it means

vectorsToMatrix flattens the object vector and its neighbors' vectors into one dense matrix of width dims. Before copying, it checks the object's vector length equals the expected dims derived from params; otherwise the flattened buffer would be misaligned. The error reports both lengths.

Source

Thrown at modules/text2vec-contextionary/additional/sempath/builder.go:165

	return nn[0].Neighbors, nil
}

// TODO: document behavior if it actually stays like this
func (pb *PathBuilder) vectorsToMatrix(obj search.Result, allObjects []search.Result, dims int,
	params *Params, searchNeighbors []*txt2vecmodels.NearestNeighbor,
) (*mat.Dense, []*txt2vecmodels.NearestNeighbor, error) {
	items := 1 // the initial object
	var neighbors []*txt2vecmodels.NearestNeighbor
	neighbors = pb.extractNeighbors(allObjects)
	neighbors = append(neighbors, searchNeighbors...)
	neighbors = pb.removeDuplicateNeighborsAndDollarNeighbors(neighbors)
	items += len(neighbors) + 1 // The +1 is for the search vector which we append last

	// concat all vectors to build gonum dense matrix
	mergedVectors := make([]float64, items*dims)
	if l := len(obj.Vector); l != dims {
		return nil, nil, fmt.Errorf("object: inconsistent vector lengths found: dimensions=%d and object=%d", dims, l)
	}

	for j, dim := range obj.Vector {
		mergedVectors[j] = float64(dim)
	}

	withoutNeighbors := 1 * dims
	for i, neighbor := range neighbors {
		neighborVector := neighbor.Vector

		if l := len(neighborVector); l != dims {
			return nil, nil, fmt.Errorf("neighbor: inconsistent vector lengths found: dimensions=%d and object=%d", dims, l)
		}

		for j, dim := range neighborVector {
			mergedVectors[withoutNeighbors+i*dims+j] = float64(dim)
		}
	}

View on GitHub (pinned to 75aa4b6d11)

Solutions

Re-vectorize the object so its dimensions match params.Dims
Ensure all objects in the collection use the same vectorizer/model settings (fix moduleConfig)
Check where params dimensions are set for semanticPath and make them consistent with the stored vectors

Defensive patterns

Strategy: validation

Validate before calling

dims := len(params.Vector)
if len(obj.Vector) != dims {
  return fmt.Errorf("object vector dim %d != expected %d; re-vectorize", len(obj.Vector), dims)
}

Type guard

func hasDims(v []float32, dims int) bool { return len(v) == dims }

Prevention

Never change vectorizer/model settings without re-vectorizing the whole collection
Check all stored vector lengths are equal after migrations
Use one dimensionality per collection

When it happens

Trigger: semanticPath computation where the search result's object vector length differs from params.Dims (or similar) — e.g. the collection was re-vectorized with a different model producing different dimensions than the params/neighbor vectors assume.

Common situations: Mixed-dimension vectors in one collection after changing vectorizer settings; legacy objects from an older contextionary version alongside new ones.

Related errors

AI-assisted analysis of weaviate/weaviate@75aa4b6d11 (2026-09-04). Data as JSON: /api/errors/888ef1c33a46a533. Report an issue: GitHub.