Python: [Bug]: BaseChatClient.as_agent() passes function_invocation_configuration to Agent() which doesn't accept it

[Laende]

Description

File: python/packages/core/agent_framework/_clients.py — BaseChatClient.as_agent()

Description:

BaseChatClient.as_agent() accepts a function_invocation_configuration parameter and forwards it into Agent(**agent_kwargs). However, Agent.__init__() (in _agents.py) does not accept this keyword argument, causing a TypeError at runtime whenever function_invocation_configuration is provided.

TypeError: Agent.__init__() got an unexpected keyword argument 'function_invocation_configuration'

Root cause:

function_invocation_configuration is consumed by FunctionInvocationLayer, which is in the client's MRO (OpenAIChatCompletionClient → FunctionInvocationLayer → ...), not in the Agent's MRO (Agent → AgentMiddlewareLayer → AgentTelemetryLayer → RawAgent → BaseAgent → ...).

Lines 86-87 of as_agent() conditionally add it to agent_kwargs, and line 89 passes it to Agent():

if function_invocation_configuration is not None:
    agent_kwargs["function_invocation_configuration"] = function_invocation_configuration

return Agent(**agent_kwargs)

Since Agent.__init__ has no **kwargs and no function_invocation_configuration parameter, this always raises TypeError.

Reproduction:

from agent_framework import FunctionInvocationConfiguration
from agent_framework.openai import OpenAIChatCompletionClient

client = OpenAIChatCompletionClient(model="gpt-4o", api_key="test")

# This raises TypeError
agent = client.as_agent(
    name="test",
    instructions="You are helpful.",
    function_invocation_configuration=FunctionInvocationConfiguration(
        include_detailed_errors=True,
    ),
)

Expected behavior:

as_agent() should either apply function_invocation_configuration to the client before creating the Agent (since it's a client-level concern), or Agent.__init__ should accept and forward it appropriately.

Workaround:

Set function_invocation_configuration at client construction time instead:

client = OpenAIChatCompletionClient(
    model="gpt-4o",
    api_key="test",
    function_invocation_configuration=FunctionInvocationConfiguration(
        include_detailed_errors=True,
    ),
)
agent = client.as_agent(name="test", instructions="You are helpful.")

Code Sample

Error Messages / Stack Traces

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[3], [line 6](vscode-notebook-cell:?execution_count=3&line=6)
      2 from agent_framework.openai import OpenAIChatCompletionClient
      3 
      4 
      5 # This raises TypeError
----> [6](vscode-notebook-cell:?execution_count=3&line=6) agent = client.as_agent(
      7     name="test",
      8     instructions="You are helpful.",
      9     function_invocation_configuration=FunctionInvocationConfiguration(

File ...\agent_framework\_clients.py:652, in BaseChatClient.as_agent(self, id, name, description, instructions, tools, default_options, context_providers, middleware, require_per_service_call_history_persistence, function_invocation_configuration, compaction_strategy, tokenizer, additional_properties)
    649 if function_invocation_configuration is not None:
    650     agent_kwargs["function_invocation_configuration"] = function_invocation_configuration
--> [652](file:///.../agent_framework/_clients.py:652) return Agent(**agent_kwargs)

TypeError: Agent.__init__() got an unexpected keyword argument 'function_invocation_configuration'

Package Versions

agent-framework-core: 1.0.0

Python Version

Python 3.14

Additional Context

No response

[eavanvalkenburg]

this is a oversight, function invocation happens in the chat client, so we should not have that param on the as_agent method, so the fix is to remove it.