types

ChatRequest dataclass

A request for chat completion.

Attributes:

Name	Type	Description
messages	list[Message]	The list of messages in the conversation.
model	Optional[str]	The model to use for completion.
temperature	float	Sampling temperature for response generation.
max_tokens	Optional[int]	Maximum tokens to generate in the response.
tools	Optional[Sequence[ToolSpec]]	Tools available for the LLM to call.
tool_choice	Literal['auto', 'none', 'required']	Tool calling behavior.
response_format	Literal['text', 'json']	Desired response format.
stream	bool	Whether to stream the response.
task_tag	Optional[str]	Optional tag for tracking the request.

Source code in wintermute/ai/types.py

@dataclass(frozen=True)
class ChatRequest:
    """A request for chat completion.

    Attributes:
        messages (list[Message]): The list of messages in the conversation.
        model (Optional[str]): The model to use for completion.
        temperature (float): Sampling temperature for response generation.
        max_tokens (Optional[int]): Maximum tokens to generate in the response.
        tools (Optional[Sequence[ToolSpec]]): Tools available for the LLM to call.
        tool_choice (Literal["auto", "none", "required"]): Tool calling behavior.
        response_format (Literal["text", "json"]): Desired response format.
        stream (bool): Whether to stream the response.
        task_tag (Optional[str]): Optional tag for tracking the request.
    """

    messages: list[Message]
    model: Optional[str] = None
    temperature: float = 0.2
    max_tokens: Optional[int] = None
    tools: Optional[Sequence[ToolSpec]] = None
    tool_choice: Literal["auto", "none", "required"] = "auto"
    response_format: Literal["text", "json"] = "text"
    stream: bool = False
    task_tag: Optional[str] = None

ChatResponse dataclass

A response from a chat completion.

Attributes:

Name	Type	Description
content	str	The content of the response.
tool_calls	Optional[list[ToolCall]]	List of tool calls made by the LLM.
model	Optional[str]	The model used for the response.
provider	Optional[str]	The provider of the model.
prompt_tokens	Optional[int]	Number of tokens in the prompt.
completion_tokens	Optional[int]	Number of tokens in the completion.
latency_ms	Optional[int]	Latency of the request in milliseconds.

Source code in wintermute/ai/types.py

@dataclass(frozen=True)
class ChatResponse:
    """A response from a chat completion.

    Attributes:
        content (str): The content of the response.
        tool_calls (Optional[list[ToolCall]]): List of tool calls made by the LLM.
        model (Optional[str]): The model used for the response.
        provider (Optional[str]): The provider of the model.
        prompt_tokens (Optional[int]): Number of tokens in the prompt.
        completion_tokens (Optional[int]): Number of tokens in the completion.
        latency_ms (Optional[int]): Latency of the request in milliseconds.
    """

    content: str
    tool_calls: Optional[list[ToolCall]] = None
    model: Optional[str] = None
    provider: Optional[str] = None
    prompt_tokens: Optional[int] = None
    completion_tokens: Optional[int] = None
    latency_ms: Optional[int] = None

Message dataclass

A message in a chat conversation.

Attributes:

Name	Type	Description
role	Role	The role of the message sender.
content	str	The content of the message.
tool_name	Optional[str]	Name of the tool if role is "tool".
tool_call_id	Optional[str]	ID of the tool call if role is "tool".

Source code in wintermute/ai/types.py

@dataclass(frozen=True)
class Message:
    """A message in a chat conversation.

    Attributes:
        role (Role): The role of the message sender.
        content (str): The content of the message.
        tool_name (Optional[str]): Name of the tool if role is "tool".
        tool_call_id (Optional[str]): ID of the tool call if role is "tool".
    """

    role: Role
    content: str
    tool_name: Optional[str] = None
    tool_call_id: Optional[str] = None

ToolCall dataclass

A tool call made by the LLM during chat completion.

Attributes:

Name	Type	Description
id	str	Unique identifier for the tool call.
name	str	Name of the tool being called.
arguments	JSONObject	The arguments passed to the tool as a JSON object.

Source code in wintermute/ai/types.py

@dataclass(frozen=True)
class ToolCall:
    """A tool call made by the LLM during chat completion.

    Attributes:
        id (str): Unique identifier for the tool call.
        name (str): Name of the tool being called.
        arguments (JSONObject): The arguments passed to the tool as a JSON object.
    """

    id: str
    name: str
    arguments: JSONObject  # parsed JSON args

ToolSpec dataclass

Specification for a tool that can be called by the LLM.

Attributes:

Name	Type	Description
name	str	Name of the tool.
description	str	Description of the tool's purpose.
input_schema	JSONObject	JSON schema defining the tool's input.
output_schema	Optional[JSONObject]	Optional JSON schema for the tool's output.

Source code in wintermute/ai/types.py

@dataclass(frozen=True)
class ToolSpec:
    """Specification for a tool that can be called by the LLM.

    Attributes:
        name (str): Name of the tool.
        description (str): Description of the tool's purpose.
        input_schema (JSONObject): JSON schema defining the tool's input.
        output_schema (Optional[JSONObject]): Optional JSON schema for the tool's output.
    """

    name: str
    description: str
    input_schema: JSONObject
    # Optional: you can validate model's JSON outputs against this if you want
    output_schema: Optional[JSONObject] = None