Share via

AIModel Class

Protocol defining the interface for AI models that can generate text responses.

This protocol standardizes how different AI providers (OpenAI, Azure OpenAI, etc.) integrate with the Teams AI framework. Implementations should handle message processing, function calling, and optional streaming.

Constructor

AIModel(*args, **kwargs)

Methods

generate_text

Generate a text response from the AI model.

Note

Implementations should handle function calling recursively - if the returned

ModelMessage contains function_calls, they should be executed and the results

fed back into the model for a final response.

generate_text

Generate a text response from the AI model.

Note

Implementations should handle function calling recursively - if the returned

ModelMessage contains function_calls, they should be executed and the results

fed back into the model for a final response.

async generate_text(input: UserMessage | ModelMessage | SystemMessage | FunctionMessage, *, system: SystemMessage | None = None, memory: Memory | None = None, functions: dict[str, microsoft.teams.ai.function.Function[pydantic.main.BaseModel]] | None = None, on_chunk: Callable[[str], Awaitable[None]] | None = None) -> ModelMessage

Parameters

Name Description
input
Required

The input message to process (user, model, function, or system message)
system
Required

Optional system message to guide model behavior
memory
Required

Optional memory storage for conversation history
functions
Required

Optional dictionary of available functions the model can call
on_chunk
Required

Optional callback for streaming text chunks as they arrive

Keyword-Only Parameters

Name Description
system
Default value: None
memory
Default value: None
functions
Default value: None
on_chunk
Default value: None

Returns

Type Description

ModelMessage containing the generated response, potentially with function calls