Encountering issue while executing planner due to parameter parsing failure

[jais001]

Hi,
I am new to the semantic kernel using python. I am currently stuck due to an error and also i have few questions.

1. I created a sample email plugin and was able to execute it using the kernel. But while using the kernel in Sequence planner I am not able to execute it. I am receiving an error which is related to the parsing of input arguments failure to the plugin. My code snippet is:

import semantic_kernel as sk
from typing import Annotated, List
from semantic_kernel.functions.kernel_function_decorator import kernel_function
from semantic_kernel.connectors.ai.open_ai import OpenAIChatCompletion
from semantic_kernel.connectors.ai.open_ai.prompt_execution_settings.open_ai_prompt_execution_settings import (
    OpenAIChatPromptExecutionSettings,
)
from semantic_kernel.connectors.ai.function_choice_behavior import FunctionChoiceBehavior

service_id = "function_call"

# Define Semantic Kernel instance
kernel = sk.Kernel()

chat_completion = OpenAIChatCompletion(
      service_id=service_id,
      ai_model_id="gpt-4o-mini",
      api_key=""
   )
kernel.add_service(chat_completion)

class EmailPlugin:
    def __init__(self) -> None:
        pass

    @kernel_function(
            description="Takes recipient as list, cc as list  as input and sends email",
            name="send_email"
    )
    def send_email(
        self,
        # content: Annotated[str, "The input is the content for sending the mail"],
        recipients: Annotated[List[str], "The input is recipients as list"],
        cc: Annotated[List[str], "The input is cc as list"]
        ) -> Annotated[str, "The output is a string"]:
        """Send an email
        Args:
            recipients (list): The recipients as list
            cc (list): The cc in the email as list
        Returns:
            str: Success message
        """        
        # print(f"The content for the email is {content}")
        print(f"The recipients of the email is {recipients}")
        print(f"The cc of the email is {cc}")
        # Code to send the email
        return f"Email sent to successfully"

email = EmailPlugin()
email_plugin = kernel.add_plugin(email, plugin_name="send_email")


settings: OpenAIChatPromptExecutionSettings = kernel.get_prompt_execution_settings_from_service_id(
        service_id=service_id
    )
settings.function_choice_behavior = FunctionChoiceBehavior.Auto(filters={"included_plugins": ["send_email"]})

recipients = ["[email protected]", "[email protected]"]
cc = ["[email protected]"]
ask = f"""
Send an email 
to the recipients:
{recipients} with cc: {cc}, with the following content:
```The report has been generated   ```, convert the content to the suitable email format
The recipients, and cc should be used as input.
"""

result1 = await kernel.invoke_prompt(
            function_name="prompt_test1",
            plugin_name="report_test1",
            prompt=ask,
            settings=settings,
        )

print(result1)

from semantic_kernel.planners import SequentialPlanner

planner = SequentialPlanner(service_id=service_id, kernel=kernel)

plan = await planner.create_plan(goal=ask)

# ask the sequential planner to execute the identified function.
result = await plan.invoke(kernel=kernel)
print(result)

The only difference I am finding is that kernel takes settings as argument in invoke_prompt() function while planner doesn't.
Iam getting the error:

Function failed. Error: Parameter recipients is expected to be parsed to typing.List[str] but is not.
Something went wrong in plan step send_email.send_email:'Parameter recipients is expected to be parsed to typing.List[str] but is not.'

2. Also, how can we add a response object to the planner so that I will get the results in a Pydantic structure?I was able to add response object to the kernel through settings, but couldn't do that for planner. I couldn't find that in the documentation either.

3. Is it possible to create a multi-agent system using a semantic kernel? I could only find examples of single agents and found in a blog that semantic kernel is for single agent systems.

Thanks for the response in advance...

[moonbox3]

@jais001, as a short response to this, we will be deprecating the planners soon. We recommend using auto function calling instead -- as you mention you were doing already before moving to the planner.

[jais001]

@moonbox3 Thanks for the information