[Python] Passing request context from FastAPI to Microsoft Semantic Kernel Plugin for OpenAI...

I am integrating Microsoft Semantic Kernel with OpenAI in my FastAPI application. I have a chat/ endpoint where I receive a session_id from the request, and I need to pass this session_id to a plugin along with the openai_client. However, I'm unsure how to properly pass the session_id from the FastAPI request to the plugin inside the kernel's execution process.

Here is the relevant code for setting up the kernel and plugin:

# Kernel and service setup
kernel = Kernel()

execution_settings = AzureChatPromptExecutionSettings(tool_choice="auto")
execution_settings.function_choice_behavior = FunctionChoiceBehavior.Auto(filters={})

openai_client = OpenAI(api_key=api_key)
chat_completion_service = OpenAIChatCompletion(
ai_model_id=model_id,
api_key=api_key,
service_id=service_id
)

# Add services and plugins
kernel.add_service(chat_completion_service)
kernel.add_plugin(MovesPlugin(openai_client), plugin_name='MovesPlugin')


Inside my FastAPI endpoint, I want to pass the session_id to the plugin when calling the kernel for a chat response:

# Inside FastAPI endpoint
@app.post("/chat/")
async def chat(request: Request):
session_id = await request.json().get('session_id')

# Get chat completion service
_chat_completion_service = kernel.get_service(type=ChatCompletionClientBase)

# Get chat completion response
response = await _chat_completion_service.get_chat_message_content(
chat_history=chat_history,
kernel=kernel,
settings=execution_settings
)

return response


How can I pass the request context (session_id) from the FastAPI request to the MovesPlugin and make sure it is used correctly along with the openai_client inside the Semantic Kernel execution?

Any guidance or suggestions would be greatly appreciated!

Continue reading...