Python: Bug: [copilot_studio_skill sample] Copilot Studio skill call fails on argument mismatch

[sergey124]

Describe the bug
The copilot_studio_skill sample fails to get a response from the skill app.

To Reproduce
Steps to reproduce the behavior:

1. deploy Copilot Studio agent with Semantic Kernel skill, following the guide in README - steps 1 to 6
2. in Copilot Studio agent -> Topics -> Greeting, add Action Invoke Semantic Kernel skill after the Trigger
3. click Test button to open the chat interface
4. type and send Hi
5. see error message in response

The skill encountered an error or bug.
To continue to run this bot, please fix the bot source code.
Hello, how can I help you today?

See the error log in the skill Container App:

  File "/usr/local/lib/python3.12/site-packages/teams/app.py", line 903, in _start_long_running_call
    return await func(context)
           ^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/teams/app.py", line 763, in _on_turn
    is_ok, matches = await self._on_activity(context, state)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/teams/app.py", line 884, in _on_activity
    if not await route.handler(context, state):
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/app/bot.py", line 62, in on_message
    sk_response = await agent.get_response(history=chat_history, user_input=user_message)
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/semantic_kernel/utils/telemetry/agent_diagnostics/decorators.py", line 69, in wrapper_decorator
    return await get_response_func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/semantic_kernel/agents/chat_completion/chat_completion_agent.py", line 275, in get_response
    async for response in self._inner_invoke(
                          ^^^^^^^^^^^^^^^^^^^
TypeError: ChatCompletionAgent._inner_invoke() got multiple values for argument 'history'
INFO:aiohttp.access:100.100.0.93 [28/Apr/2025:07:25:07 +0000] "POST /api/messages HTTP/1.1" 200 3597 "-" "Microsoft-BotFramework/3.1 BotBuilder/4.23.0.0 .NETCoreAppVersion/v8.0"

Expected behavior
The skill replies with a joke-related phrase.

Screenshots
Error message in Copilot Studio test chat agent reply:

Error log in the skill Container App:

Platform

• Language: Python
• Source: main branch of repository
• AI model: OpenAI:GPT-4o-mini(2024-07-18)

Additional context
I managed to fix the issue with this change, which

• updates semantic-kernel version to 1.28.1
• updates agent.get_response usage to new contract
• uses openAiApiKey parameter provided in copilot_studio_skill/.azure/SKCopilot/.env as AZURE_OPENAI_API_KEY
• also I had to specify in the SKCopilot/.env the AZURE_OPENAI_MODEL="my-gpt4o-deployment-name", as the sample expects specific deployment name gpt-4o not existing in our OpenAI resource.

[sergey124]

@unsafecode , at the end of the issue description there's a fix suggestion that worked for me:
sergey124@13f387b
There in main.bicep:103 you put

openAiApiKey: '' // Force ManId, otherwise set openAI.listKeys().key1

which I populated with a param from .env, as I couldn't implement the comment - maybe you could implement.

Screenshot:

[moonbox3]

This issue here is that this demo hasn't been updated to use the thread abstractions we introduced in v1.27.0+. The get_response method no longer takes a chat history, but rather messages and an optional thread.

[sergey124]

@moonbox3 , also:

1. openAiApiKey has to be provided by bicep or the user - main.bicep:103
2. the sample expects deployment naming AZURE_OPENAI_MODEL="gpt-4o", it would be useful to mention, to avoid errors
3. AZURE_OPENAI_API_VERSION="2024-08-01-preview" is quite old, you suggested in another issue to update - aca.bicep:7, main.parameters.json:27