chroma-core/chroma · error · ValueError

The model cannot be changed after the embedding function has

Error message

The model cannot be changed after the embedding function has been initialized.

What it means

MistralEmbeddingFunction.validate_config_update rejects any config update containing the key 'model', raising this ValueError. The model determines the embedding dimension and vector space (mistral-embed = 1024 dims), so changing it in place would make existing stored vectors incomparable; it is deliberately immutable. Only api_key_env_var is updatable.

Source

Thrown at chromadb/utils/embedding_functions/mistral_embedding_function.py:80

    def build_from_config(config: Dict[str, Any]) -> "EmbeddingFunction[Documents]":
        model = config.get("model")
        api_key_env_var = config.get("api_key_env_var")

        if model is None or api_key_env_var is None:
            assert False, "This code should not be reached"  # this is for type checking
        return MistralEmbeddingFunction(model=model, api_key_env_var=api_key_env_var)

    def get_config(self) -> Dict[str, Any]:
        return {
            "model": self.model,
            "api_key_env_var": self.api_key_env_var,
        }

    def validate_config_update(
        self, old_config: Dict[str, Any], new_config: Dict[str, Any]
    ) -> None:
        if "model" in new_config:
            raise ValueError(
                "The model cannot be changed after the embedding function has been initialized."
            )

    @staticmethod
    def validate_config(config: Dict[str, Any]) -> None:
        """
        Validate the configuration using the JSON schema.

        Args:
            config: Configuration to validate
        """
        validate_config_schema(config, "mistral")

View on GitHub (pinned to aecdd12c8a)

Solutions

  1. Create a new collection with the new model EF and re-embed your corpus
  2. Strip the immutable key: cfg = {k: v for k, v in new_config.items() if k != 'model'}
  3. Treat model changes as a migration (export documents, new collection, re-import), not a config edit

Example fix

# before
update = ef.get_config()          # {"model": "mistral-embed", "api_key_env_var": "MISTRAL_API_KEY"}
update["model"] = "mistral-embed-v2"
ef.validate_config_update(ef.get_config(), update)  # ValueError

# after
update = {"api_key_env_var": "MISTRAL_API_KEY_PROD"}  # only mutable keys
ef.validate_config_update(ef.get_config(), update)
Defensive patterns

Strategy: validation

Validate before calling

def safe_mistral_update(ef, new_config: dict) -> dict:
    if "model" in new_config:
        raise ValueError(
            "model is immutable; create a new collection and re-embed instead"
        )
    ef.validate_config_update(ef.get_config(), new_config)
    return new_config

Try / catch

try:
    ef.validate_config_update(old, new)
except ValueError as e:
    if "cannot be changed" in str(e):
        new = {k: v for k, v in new.items() if k != "model"}
        ef.validate_config_update(old, new)
    else:
        raise

Prevention

When it happens

Trigger: Applying a config update with {'model': 'mistral-embed-new'} to an existing EF; round-tripping get_config() (which returns model and api_key_env_var) straight into an update; admin tooling PUTting full config objects.

Common situations: Mistral releasing a new embedding model and attempting an in-place upgrade; config-sync scripts echoing stored config back; manually editing serialized EF config JSON.

Related errors

AI-assisted analysis of chroma-core/chroma@aecdd12c8a (2026-08-16). Data as JSON: /api/errors/18a2680a8e217949. Report an issue: GitHub.