Python: [Bug]: issue in agent_framework_anthropic import

[leelakarthik]

Description

When i update Annotation with Annotations, issue re surfaces with Content Import, when digged again it's Contents.

Both I found in _types.py.

Later ImportError: cannot import name 'prepare_function_call_results' from 'agent_framework' (/Users/karthik/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/init.py)

This i could not fix locally as Iam not completely aware of the method signature.

Code Sample

from agent_framework_anthropic import AnthropicClient

Error Messages / Stack Traces

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
Cell In[1], line 1
----> 1 from agent_framework_anthropic import AnthropicClient
      2 from .MAF.codebase_analyst_agent.config.settings import settings

File ~/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework_anthropic/__init__.py:5
      1 # Copyright (c) Microsoft. All rights reserved.
      3 import importlib.metadata
----> 5 from ._chat_client import AnthropicChatOptions, AnthropicClient
      7 try:
      8     __version__ = importlib.metadata.version(__name__)

File ~/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework_anthropic/_chat_client.py:7
      4 from collections.abc import AsyncIterable, MutableMapping, MutableSequence, Sequence
      5 from typing import Any, ClassVar, Final, Generic, Literal, TypedDict
----> 7 from agent_framework import (
      8     AGENT_FRAMEWORK_USER_AGENT,
      9     AIFunction,
     10     Annotation,
     11     BaseChatClient,
     12     ChatMessage,
     13     ChatOptions,
     14     ChatResponse,
     15     ChatResponseUpdate,
     16     Content,
     17     FinishReason,
     18     HostedCodeInterpreterTool,
     19     HostedMCPTool,
     20     HostedWebSearchTool,
     21     Role,
     22     TextSpanRegion,
     23     UsageDetails,
     24     get_logger,
     25     prepare_function_call_results,
     26     use_chat_middleware,
     27     use_function_invocation,
     28 )
     29 from agent_framework._pydantic import AFBaseSettings
     30 from agent_framework.exceptions import ServiceInitializationError

ImportError: cannot import name 'Annotation' from 'agent_framework' (/Users/karthik/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/__init__.py)

Package Versions

1.0.0b251120, 1.0.0b260127

Python Version

3.14

Additional Context

When i update Annotation with Annotations, issue re surfaces with Content Import, when digged again it's Contents.

Both I found in _types.py.

Later ImportError: cannot import name 'prepare_function_call_results' from 'agent_framework' (/Users/karthik/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/init.py)

This i could not fix locally as Iam not completely aware of the method signature.

[leelakarthik]

cannot import name 'use_instrumentation' from 'agent_framework.observability' (/Users/karthik/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/observability.py)
Got this error when i commented the prepare_function_call_results import

[leelakarthik]

under _chat_clients.py
class AnthropicClient(BaseChatClient[TAnthropicOptions], Generic[TAnthropicOptions]):
BaseChatClient is not subscriptable
probable fix: class AnthropicClient(BaseChatClient, Generic[TAnthropicOptions]):

[leelakarthik]

cannot import name 'use_instrumentation' from 'agent_framework.observability' (/Users/karthik/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/observability.py) Got this error when i commented the prepare_function_call_results import

I saw that the function is renamed to use_observability

[leelakarthik]

from anthropic import AsyncAnthropicFoundry
from dotenv import load_dotenv
load_dotenv()
client = AnthropicClient(
    anthropic_client=AsyncAnthropicFoundry(), model_id=settings.CLAUDE_DEPLOYMENT_NAME
)

maths_school_agent = client.create_agent(
    name="maths teacher",
    instructions="You are a maths teacher that helps students with their math homework upto class 10.",
)

result = await maths_school_agent.run("Can you help me solve 2x + 3 = 7?")
print(result)

---

TypeError Traceback (most recent call last)
Cell In[24], line 1
----> 1 result = await maths_school_agent.run("Can you help me solve 2x + 3 = 7?")
2 print(result)
File ~/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/_middleware.py:906, in use_agent_middleware..middleware_enabled_run(self, messages, thread, middleware, **kwargs)
903 return result if result else AgentRunResponse()
905 # No middleware, execute directly
--> 906 return await original_run(self, normalized_messages, thread=thread, **kwargs)
File ~/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/observability.py:996, in _trace_agent_run..trace_run(self, messages, thread, **kwargs)
992 global OBSERVABILITY_SETTINGS
994 if not OBSERVABILITY_SETTINGS.ENABLED:
995 # If model diagnostics are not enabled, just return the completion
--> 996 return await run_func(self, messages=messages, thread=thread, **kwargs)
997 attributes = _get_span_attributes(
998 operation_name=OtelAttr.AGENT_INVOKE_OPERATION,
999 provider_name=provider_name,
(...) 1005 **kwargs,
1006 )
1007 with _get_span(attributes=attributes, span_name_attribute=OtelAttr.AGENT_NAME) as span:
File ~/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/_agents.py:556, in ChatAgent.run(self, messages, thread, frequency_penalty, logit_bias, max_tokens, metadata, model, presence_penalty, response_format, seed, stop, store, temperature, tool_choice, tools, top_p, user, additional_properties, **kwargs)
554 await self._async_exit_stack.enter_async_context(mcp_server)
555 final_tools.extend(mcp_server.functions)
--> 556 response = await self.chat_client.get_response(
557 messages=thread_messages,
558 chat_options=run_chat_options
559 & ChatOptions(
560 model_id=model,
561 conversation_id=thread.service_thread_id,
562 frequency_penalty=frequency_penalty,
563 logit_bias=logit_bias,
564 max_tokens=max_tokens,
565 metadata=metadata,
566 presence_penalty=presence_penalty,
567 response_format=response_format,
568 seed=seed,
569 stop=stop,
570 store=store,
571 temperature=temperature,
572 tool_choice=tool_choice,
573 tools=final_tools,
574 top_p=top_p,
575 user=user,
576 additional_properties=additional_properties or {},
577 ),
578 **kwargs,
579 )
581 await self._update_thread_with_type_and_conversation_id(thread, response.conversation_id)
583 # Ensure that the author name is set for each message in the response.
File ~/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/_tools.py:842, in _handle_function_calls_response..decorator..function_invocation_wrapper(self, messages, **kwargs)
840 fcc_messages: "list[ChatMessage]" = []
841 for attempt_idx in range(max_iterations):
--> 842 response = await func(self, messages=prepped_messages, **kwargs)
843 # if there are function calls, we will handle them first
844 function_results = {
845 it.call_id for it in response.messages[0].contents if isinstance(it, FunctionResultContent)
846 }
File ~/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/observability.py:771, in _trace_get_response..decorator..trace_get_response(self, messages, **kwargs)
768 global OBSERVABILITY_SETTINGS
769 if not OBSERVABILITY_SETTINGS.ENABLED:
770 # If model diagnostics are not enabled, just return the completion
--> 771 return await func(
772 self,
773 messages=messages,
774 **kwargs,
775 )
776 if "token_usage_histogram" not in self.additional_properties:
777 self.additional_properties["token_usage_histogram"] = _get_token_usage_histogram()
File ~/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/_middleware.py:1003, in use_chat_middleware..middleware_enabled_get_response(self, messages, **kwargs)
1001 # If no chat middleware, use original method
1002 if not chat_middleware_list:
-> 1003 return await original_get_response(self, messages, **kwargs)
1005 # Create pipeline and execute with middleware
1006 from ._types import ChatOptions
File ~/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/_clients.py:416, in BaseChatClient.get_response(self, messages, frequency_penalty, logit_bias, max_tokens, metadata, model, presence_penalty, response_format, seed, stop, store, temperature, tool_choice, tools, top_p, user, additional_properties, **kwargs)
413 self._prepare_tool_choice(chat_options=chat_options)
415 filtered_kwargs = self._filter_internal_kwargs(kwargs)
--> 416 return await self._inner_get_response(messages=prepped_messages, chat_options=chat_options, **filtered_kwargs)
TypeError: AnthropicClient._inner_get_response() missing 1 required keyword-only argument: 'options'

I have updated the params of _inner_get_response to take chat_options instead of options

from agent_framework._types import prepend_instructions_to_messages is also a broken import

[leelakarthik]

ImportError: cannot import name 'validate_tool_mode' from 'agent_framework._types' (/Users/karthik/Code bases/Jan/.venv/lib/python3.14/site-packages/agent_framework/_types.py)

[eavanvalkenburg]

make sure you have the same version of core and anthropic, we currently always release all packages with new versions, so this is quite important!

[leelakarthik]

I am using 1.0.0b260127