nat.plugins.langchain.agent.tool_calling_agent.register

Attributes

logger

Classes

TruncationRetryConfig	Configuration for retrying LLM calls that are truncated (finish_reason='length').
ToolCallAgentWorkflowConfig	A Tool Calling Agent requires an LLM which supports tool calling. A tool Calling Agent utilizes the tool

Functions

tool_calling_agent_workflow(config, builder)

Module Contents

logger

class TruncationRetryConfig(/, **data: Any)

Bases: pydantic.BaseModel

Configuration for retrying LLM calls that are truncated (finish_reason=’length’).

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

max_retries: int = None

token_increment: int | None = None

token_scaling: float | None = None

_check_scaling_strategy() → TruncationRetryConfig

build_scaling_fn() → collections.abc.Callable[[int], int]

Build a callable that computes the next max_tokens from the current value.

class ToolCallAgentWorkflowConfig

Bases: nat.data_models.agent.AgentBaseConfig

A Tool Calling Agent requires an LLM which supports tool calling. A tool Calling Agent utilizes the tool input parameters to select the optimal tool. Supports handling tool errors.

description: str = None

tool_names: list[nat.data_models.component_ref.FunctionRef | nat.data_models.component_ref.FunctionGroupRef] = None

handle_tool_errors: bool = None

max_iterations: int = None

max_history: int = None

truncation_retry: TruncationRetryConfig = None

max_empty_response_retries: int = None

system_prompt: str | None = None

additional_instructions: str | None = None

return_direct: list[nat.data_models.component_ref.FunctionRef] | None = None

async tool_calling_agent_workflow(

config: ToolCallAgentWorkflowConfig,

builder: nat.builder.builder.Builder,

)