googleapis/mcp-toolbox · error

embedding model does not exist: %s

Error message

embedding model does not exist: %s

What it means

EmbedParams fails fast when a parameter's configured embedding model name cannot be found in the model manager (pMgr.GetEmbeddingModel returned !ok). It means the tool config references an embedding model that was never registered under that exact name. This is a configuration/name resolution problem, not a runtime API failure.

Source

Thrown at internal/util/parameters/parameters.go:209

		}

		// Get parameter's value to be embedded
		valueStr, ok := paramValues[i].Value.(string)
		if !ok {
			return nil, fmt.Errorf("parameter '%s' is marked for embedding but has a non-string value (type: %T)", p.GetName(), paramValues[i].Value)
		}

		parametersToEmbed[modelName] = append(parametersToEmbed[modelName], ParamToEmbed{
			OriginalValue: valueStr,
			Index:         i,
		})
	}

	// Batch embedding request sent to each model
	for modelName, params := range parametersToEmbed {
		model, ok := pMgr.GetEmbeddingModel(modelName)
		if !ok {
			return nil, fmt.Errorf("embedding model does not exist: %s", modelName)
		}

		// Extract only the string values for the API call
		stringBatch := make([]string, len(params))
		for i, paramStr := range params {
			stringBatch[i] = paramStr.OriginalValue
		}

		embeddings, err := model.EmbedParameters(ctx, stringBatch)
		if err != nil {
			return nil, fmt.Errorf("error embedding parameters with model %s: %w", modelName, err)
		}

		if len(embeddings) != len(stringBatch) {
			return nil, fmt.Errorf("model %s returned %d embeddings for %d inputs", modelName, len(embeddings), len(stringBatch))
		}

		for i, rawVector := range embeddings {

View on GitHub (pinned to 8cc6e09de2)

Solutions

1. Compare the model name in the error with the keys of your embeddingModels config section; fix the name so they match exactly (case-sensitive).
2. Add the missing embedding model definition so it is registered with the model manager before EmbedParams is called.
3. Diff your customized tools config against the upstream/prebuilt version to find the renamed or removed model entry.
4. Dump the registered embedding model keys at startup and validate all references in a unit test.

Example fix

# before (name mismatch)
embeddingModels:
  myEmbedModel:
    kind: gemini
tools:
  search:
    kind: mysqldata-sql
    embeddingModel: my-embed-model

# after (exact match)
embeddingModels:
  myEmbedModel:
    kind: gemini
tools:
  search:
    kind: mysqldata-sql
    embeddingModel: myEmbedModel

for _, name := range referencedEmbeddingModels {
    if _, ok := mgr.GetEmbeddingModel(name); !ok {
        return fmt.Errorf("unknown embedding model %q", name)
    }
}

func embeddingModelExists(mgr *models.Manager, name string) bool {
    _, ok := mgr.GetEmbeddingModel(name)
    return ok
}

res, err := params.EmbedParams(ctx, pMgr, p)
if err != nil {
    if strings.Contains(err.Error(), "embedding model does not exist") {
        // fix config / register model, then retry once
    }
    return err
}

When it happens

Trigger: Calling EmbedParams (directly or via a tool invocation with parameter embeddings enabled) where the embedding model name grouped under parametersToEmbed does not match any key registered in the model manager — e.g. a typo in the model name in the tools YAML, or the embedding model definition was omitted/renamed.

Common situations: Renaming a model in the embeddingModels section without updating parameter references; typos or case mismatches in the model name; users overriding prebuilt configs and dropping the embedding model definition; loading a partial tools file that lacks the model block.

Related errors

• unable to initialize logger: %w
• invalid log level
• unable to initialize tool %q: %w
• unable to retrieve source %q for tool %q
• unable to initialize group %q: %w

AI-assisted analysis of googleapis/mcp-toolbox@8cc6e09de2 (2026-09-05). Data as JSON: /api/errors/5bd110ce86d93bc0. Report an issue: GitHub.