microsoft/semantic-kernel · error · AgentInitializationException

Could not generate schema for the type {response_format}.

Error message

Could not generate schema for the type {response_format}.

What it means

Raised by configure_response_format() on the plain-Python-class branch when KernelJsonSchemaBuilder.build(parameter_type=response_format, structured_output=True) returns None. The builder returns None when it cannot derive a JSON schema for the given type, so the method cannot construct a structured output envelope.

Source

Thrown at python/semantic_kernel/agents/open_ai/openai_responses_agent.py:801

                f"Encountered unexpected response_format type: {resp_type}. Allowed types are `json_object` "
                " and `json_schema`."
            )
        if isinstance(response_format, type):
            if issubclass(response_format, BaseModel):
                interim_format = type_to_text_format_param(response_format)
                if interim_format["type"] != "json_schema":
                    raise AgentInitializationException("Only 'json_schema' is allowed from that helper.")
                configured_format = {
                    "type": "json_schema",
                    "name": interim_format.get("name", response_format.__name__),
                    "schema": interim_format.get("schema"),
                    "strict": interim_format.get("strict", True),
                }
            else:
                # Build a schema from a plain Python class
                generated_schema = KernelJsonSchemaBuilder.build(parameter_type=response_format, structured_output=True)
                if generated_schema is None:
                    raise AgentInitializationException(f"Could not generate schema for the type {response_format}.")
                configured_format = {
                    "type": "json_schema",
                    "name": response_format.__name__,
                    "schema": generated_schema,
                    "strict": True,
                }
        else:
            raise AgentInitializationException(
                "response_format must be a dictionary, a subclass of BaseModel, a Python class/type, or None"
            )

        return {"format": configured_format}

    # endregion

    # region Invocation Methods

    @trace_agent_get_response

View on GitHub (pinned to c028a0c7dc)

Solutions

1. Convert the class to a pydantic BaseModel (preferred — uses the BaseModel branch and the openai helper).
2. Ensure all fields have type annotations the builder understands.
3. Build the JSON schema yourself and pass it via the dict path with type 'json_schema'.
4. Log KernelJsonSchemaBuilder.build(...) in isolation to see why it returns None.

Example fix

// before
class Out:
    value = None  # no annotation
cfg = OpenAIResponsesAgent.configure_response_format(Out)

// after
from pydantic import BaseModel
class Out(BaseModel):
    value: str
cfg = OpenAIResponsesAgent.configure_response_format(Out)

from semantic_kernel.schema.kernel_json_schema_builder import KernelJsonSchemaBuilder
schema = KernelJsonSchemaBuilder.build(parameter_type=response_format, structured_output=True)
if schema is None:
    raise ValueError('Cannot derive schema; convert to BaseModel or supply dict')

def type_has_schema(t: type) -> bool:
    from semantic_kernel.schema.kernel_json_schema_builder import KernelJsonSchemaBuilder
    return KernelJsonSchemaBuilder.build(parameter_type=t, structured_output=True) is not None

from semantic_kernel.exceptions.agent_exceptions import AgentInitializationException
try:
    cfg = OpenAIResponsesAgent.configure_response_format(MyClass)
except AgentInitializationException as e:
    if 'generate schema' in str(e):
        from pydantic import BaseModel
        class MyModel(BaseModel):
            ...  # mirror MyClass
        cfg = OpenAIResponsesAgent.configure_response_format(MyModel)
    raise

When it happens

Trigger: Passing a plain Python class/type (not a pydantic BaseModel) whose annotations or structure the KernelJsonSchemaBuilder cannot introspect — e.g. a class with no type-annotated fields, dynamic/Generated types, or types the builder explicitly skips.

Common situations: Using a dataclass or TypedDict the schema builder doesn't support, a class with untyped attributes, or passing a builtin/abstract type. The builder yields None rather than a partial/invalid schema.

Related errors

• If response_format has type 'json_schema', 'json_schema' mus
• Encountered unexpected response_format type: {resp_type}. Al
• Only 'json_schema' is allowed from that helper.
• The option keys 'name' and 'type' are required for a paramet
• The provided response format '{formatString}' is not support

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/e4d2ecef52033eb8. Report an issue: GitHub.