​
ReasoningAgent

ReasoningAgent(
    name: str,
    llm_config: dict,
    grader_llm_config: dict | None = None,
    max_depth: int = 4,
    beam_size: int = 3,
    answer_approach: str = 'pool',
    verbose: bool = True,
    reason_config: dict = \{},
    **kwargs
)

(In preview) Assistant agent, designed to solve a task with LLM.

AssistantAgent is a subclass of ConversableAgent configured with a default system message. The default system message is designed to solve a task with LLM, including suggesting python code blocks and debugging. human_input_mode is default to “NEVER” and code_execution_config is default to False. This agent doesn’t execute code by default, and expects the user to execute the code.

Initialize a ReasoningAgent that uses tree-of-thought reasoning.

Parameters:
NameDescription
nameName of the agent

Type: str
llm_configConfiguration for the language model

Type: dict
grader_llm_configOptional separate configuration for the grader model.

If not provided, uses llm_config

Type: dict | None

Default: None
max_depthMaximum depth of the reasoning tree

Type: int

Default: 4
beam_sizeDEPRECATED.

Number of parallel reasoning paths to maintain

Type: int

Default: 3
answer_approachDEPRECATED.

Either “pool” or “best” - how to generate final answer

Type: str

Default: ‘pool’
verboseWhether to show intermediate steps

Type: bool

Default: True
reason_configConfiguration for the reasoning method.

Supported parameters: method (str): The search strategy to use.

Options: - “beam_search” (default): Uses beam search with parallel paths - “mcts”: Uses Monte Carlo Tree Search for exploration - “lats”: Uses Language Agent Tree Search with per-step rewards - “dfs”: Uses depth-first search (equivalent to beam_search with beam_size=1) Common parameters: max_depth (int): Maximum depth of reasoning tree (default: 3) forest_size (int): Number of independent trees to maintain (default: 1) rating_scale (int): Scale for grading responses, e.g.

1-10 (default: 10) Beam Search specific: beam_size (int): Number of parallel paths to maintain (default: 3) answer_approach (str): How to select final answer, “pool” or “best” (default: “pool”) MCTS/LATS specific: nsim (int): Number of simulations to run (default: 3) exploration_constant (float): UCT exploration parameter (default: 1.41) Example configs: \\{"method": "beam_search", "beam_size": 5, "max_depth": 4} \\{"method": "mcts", "nsim": 10, "exploration_constant": 2.0} \\{"method": "lats", "nsim": 5, "forest_size": 3}

Type: dict

Default: {}
**kwargs

​
Instance Attributes

​
method



​
Instance Methods

​
generate_forest_response

generate_forest_response(
    self,
    messages: list[dict],
    sender: autogen.Agent,
    config: dict | None = None
) -> tuple[bool, str]

Generate a response using tree-of-thought reasoning.

Parameters:
NameDescription
messagesInput messages to respond to

Type: list[dict]
senderAgent sending the messages

Type: autogen.Agent
configOptional configuration

Type: dict | None

Default: None
Returns:
TypeDescription
tuple[bool, str]Tuple[bool, str]: Success flag and generated response

​
rate_node

rate_node(
    self,
    node: autogen.ThinkNode,
    ground_truth: str = None,
    is_outcome: bool = False
) -> float

Rate the quality of a reasoning path using the grader agent.

Parameters:
NameDescription
nodeNode containing the reasoning trajectory to evaluate

Type: autogen.ThinkNode
ground_truthOptional ground truth to provide to the grader

Type: str

Default: None
is_outcomeindicates whether the rating is for an outcome (final answer) or a process (thinking trajectory).

Type: bool

Default: False
Returns:
TypeDescription
floatfloat: Normalized score between 0 and 1 indicating trajectory quality

OpenAIWrapperSenderRequiredError