weaviate/weaviate · error

the object is invalid, as weaviate could not extract any…

Error message

the object is invalid, as weaviate could not extract any contextionary-valid words from it. This is the case when you have set the options 'vectorizeClassName: false' and 'vectorizePropertyName: false' in this class' schema definition and not a single property's value contains at least one contextionary-valid word. To fix this, you have several options:

1.) Make sure that the schema class name or the set properties are a contextionary-valid term and include them in vectorization using the 'vectorizeClassName' or 'vectorizePropertyName' setting. In this case the vector position will be composed of both the class/property names and the values for those fields. Even if no property values are contextionary-valid, the overall word corpus is still valid due to the contextionary-valid class/property names.

2.) Alternatively, if you do not want to include schema class/property names in vectorization, you must make sure that at least one text/string property contains at least one contextionary-valid word.

3.) If the word corpus weaviate extracted from your object (see below) does contain enough meaning to build a vector position, but the contextionary did not recognize the words, you can extend the contextionary using the REST API. This is the case	when you use mostly industry-specific terms which are not known to the common language contextionary. Once extended, simply reimport this object.

The following words were extracted from your object: %v

To learn more about the contextionary and how it behaves, check out: https://www.semi.technology/documentation/weaviate/current/contextionary.html

Original error: %v

What it means

Object() calls the contextionary client's VectorForCorpi and, when the returned error matches the typed ErrNoUsableWords, replaces it with this long user-facing explanation. It means vectorization failed because no extracted word was contextionary-valid, typically when both vectorizeClassName and vectorizePropertyName are false and no property value contains a valid word. The original error (with the extracted words) is appended via 'Original error: %v'.

Solutions

  1. Set vectorizeClassName or vectorizePropertyName to true in the class schema so valid schema words contribute to the corpus
  2. Make sure at least one text/string property value contains at least one common English word known to the contextionary
  3. Extend the contextionary via the REST API with your industry-specific terms, then re-import the object

Example fix

// before
{"vectorizer": {"text2vec-contextionary": {"vectorizeClassName": false, "vectorizePropertyName": false}}}
// after
{"vectorizer": {"text2vec-contextionary": {"vectorizeClassName": true, "vectorizePropertyName": true}}}
Defensive patterns

Strategy: type-guard

Validate before calling

// before creating the object, ensure at least one vectorized field has plausible words
func hasUsableText(props map[string]interface{}, vectorizedProps []string) bool {
	for _, p := range vectorizedProps {
		if s, ok := props[p].(string); ok {
			for _, w := range strings.Fields(s) {
				if len([]rune(w)) >= 3 { return true }
			}
		}
	}
	return false
}

Type guard

func isInvalidObjectNoWords(err error) bool {
	var target error
	return strings.Contains(err.Error(), "could not extract") &&
		strings.Contains(err.Error(), "Original error") ||
		errors.As(err, &target)
}

Try / catch

obj, err := batch.ObjectsBatchCreate().WithObject(o).Do(ctx)
if err != nil {
	if strings.Contains(err.Error(), "could not extract any contextionary-valid words") {
		// fix schema settings or enrich the object text
	}
}

Prevention

When it happens

Trigger: A Weaviate object import (batch or REST/GraphQL create) into a class using the text2vec-contextionary module where: vectorizeClassName=false, vectorizePropertyName=false, and every property value consists solely of words unknown to the contextionary.

Common situations: Users importing objects with only IDs/numbers/foreign-language or domain jargon text while disabling name vectorization; misconfigured schema where all textual content was expected to be vectorized but no valid words exist.

Related errors


AI-assisted analysis of weaviate/weaviate@75aa4b6d11 (2026-09-04). Data as JSON: /api/errors/471d4b165a120ca8. Report an issue: GitHub.

Appendix: source

Thrown at modules/text2vec-contextionary/vectorizer/vectorizer.go:111

	return vec, additional, nil
}

func (v *Vectorizer) object(ctx context.Context, object *models.Object, overrides map[string]string,
	cfg moduletools.ClassConfig,
) ([]float32, []txt2vecmodels.InterpretationSource, error) {
	icheck := NewIndexChecker(cfg)
	corpi, isEmpty := v.objectVectorizer.Texts(ctx, object, icheck)
	if isEmpty {
		// don't vectorize empty text
		return nil, nil, nil
	}

	vector, ie, err := v.client.VectorForCorpi(ctx, []string{corpi}, overrides)
	if err != nil {
		switch {
		case errors.As(err, &ErrNoUsableWords{}):
			return nil, nil, fmt.Errorf("the object is invalid, as weaviate could not extract "+
				"any contextionary-valid words from it. This is the case when you have "+
				"set the options 'vectorizeClassName: false' and 'vectorizePropertyName: false' in this class' schema definition "+
				"and not a single property's value "+
				"contains at least one contextionary-valid word. To fix this, you have several "+
				"options:\n\n1.) Make sure that the schema class name or the set properties are "+
				"a contextionary-valid term and include them in vectorization using the "+
				"'vectorizeClassName' or 'vectorizePropertyName' setting. In this case the vector position "+
				"will be composed of both the class/property names and the values for those fields. "+
				"Even if no property values are contextionary-valid, the overall word corpus is still valid "+
				"due to the contextionary-valid class/property names."+
				"\n\n2.) Alternatively, if you do not want to include schema class/property names "+
				"in vectorization, you must make sure that at least one text/string property contains "+
				"at least one contextionary-valid word."+
				"\n\n3.) If the word corpus weaviate extracted from your object "+
				"(see below) does contain enough meaning to build a vector position, but the contextionary "+
				"did not recognize the words, you can extend the contextionary using the "+
				"REST API. This is the case	when you use mostly industry-specific terms which are "+
				"not known to the common language contextionary. Once extended, simply reimport this object."+

View on GitHub (pinned to 75aa4b6d11)