mlflow/mlflow · error · ValueError

Invalid route type {route_type}

Error message

Invalid route type {route_type}

What it means

OpenAIProvider.get_endpoint_url maps a gateway route type string ('llm/v1/chat', 'llm/v1/completions', 'llm/v1/embeddings') to an OpenAI path. Any other route_type value raises a plain ValueError. It is an internal invariant error, normally not caused by user API calls but by a mis-registered route.

Source

Thrown at mlflow/gateway/providers/openai.py:268

                result_headers.pop("authorization", None)
                result_headers.pop("api-key", None)
            result_headers = client_headers | result_headers

        return result_headers

    @property
    def adapter_class(self):
        return OpenAIAdapter

    def get_endpoint_url(self, route_type: str) -> str:
        if route_type == "llm/v1/chat":
            route_path = "chat/completions"
        elif route_type == "llm/v1/completions":
            route_path = "completions"
        elif route_type == "llm/v1/embeddings":
            route_path = "embeddings"
        else:
            raise ValueError(f"Invalid route type {route_type}")

        # Append the route path to the base URL. Note that we cannot simply append the route path
        # at the end of the base URL because it has query parameters for the Azure OpenAI case.
        parsed_base_url = urlparse(self.base_url)
        return urlunparse(parsed_base_url._replace(path=f"{parsed_base_url.path}/{route_path}"))

    async def _chat_stream(
        self, payload: chat.RequestPayload
    ) -> AsyncIterable[chat.StreamResponsePayload]:
        from fastapi.encoders import jsonable_encoder

        payload = jsonable_encoder(payload, exclude_none=True)
        self.check_for_model_field(payload)

        # Inject stream_options.include_usage=true to get usage in final chunk
        if payload.get("stream_options") is None:
            payload["stream_options"] = {"include_usage": True}
        elif "include_usage" not in payload["stream_options"]:

View on GitHub (pinned to 6a27f2decc)

Solutions

1. Use exactly one of the route type strings 'llm/v1/chat', 'llm/v1/completions', or 'llm/v1/embeddings'.
2. If passing an EndpointType enum, convert it to its string value (str(endpoint_type)) before calling get_endpoint_url.
3. Create routes via mlflow.gateway.start_server / MlflowException-facing public APIs rather than calling get_endpoint_url directly.
4. Check the MLflow version; route type naming changed historically, so align code examples with your installed version.

Example fix

// before
provider.get_endpoint_url(EndpointType.LLM_V1_CHAT)
// after
provider.get_endpoint_url(str(EndpointType.LLM_V1_CHAT))  # 'llm/v1/chat'

VALID = {"llm/v1/chat", "llm/v1/completions", "llm/v1/embeddings"}
assert route_type in VALID, f"unsupported route type: {route_type}"

def is_valid_route_type(rt) -> bool:
    return isinstance(rt, str) and rt in {"llm/v1/chat", "llm/v1/completions", "llm/v1/embeddings"}

try:
    url = provider.get_endpoint_url(route_type)
except ValueError as e:
    logging.error("Invalid route type: %s", e)
    raise

When it happens

Trigger: Registering or invoking a gateway route whose route_type string is not one of 'llm/v1/chat', 'llm/v1/completions', 'llm/v1/embeddings' (e.g. 'llm/v1/embedding', 'chat', or an EndpointType enum not converted to its string form).

Common situations: Programmatic route registration with a mistyped route type, custom code calling get_endpoint_url directly with an enum instead of its string value, or older/newer MLflow route type names mixed across versions.

Related errors

• Invalid route type {route_type}
• Unsupported route_type '{route_type}' for Databricks provide
• Invalid route type {route_type}
• Endpoint {name!r} is not a chat endpoint.
• Endpoint {name!r} is not a completions endpoint.

AI-assisted analysis of mlflow/mlflow@6a27f2decc (2026-08-29). Data as JSON: /api/errors/475b335679da9aee. Report an issue: GitHub.