agentsociety.agent.block

Module Contents

Classes

BlockParams
BlockOutput
Block	A foundational component similar to a layer in PyTorch, used for building complex systems.

Data

TRIGGER_INTERVAL
__all__

API

agentsociety.agent.block.TRIGGER_INTERVAL

1

agentsociety.agent.block.__all__

[‘Block’, ‘BlockParams’, ‘BlockOutput’]

class agentsociety.agent.block.BlockParams

Bases: pydantic.BaseModel

block_memory: Optional[dict[str, Any]]

None

class agentsociety.agent.block.BlockOutput

Bases: pydantic.BaseModel

class agentsociety.agent.block.Block(toolbox: agentsociety.agent.toolbox.AgentToolbox, agent_memory: Optional[agentsociety.memory.Memory] = None, block_params: Optional[Any] = None)

A foundational component similar to a layer in PyTorch, used for building complex systems.

Initialization

• Description:

  • Initializes a new instance of the Block class with optional LLM, Memory, Simulator, and Trigger components.

• Args:

  • name (str): The name of the block.

  • llm (Optional[LLM], optional): An instance of LLM. Defaults to None.

  • environment (Optional[Environment], optional): An instance of Environment. Defaults to None.

  • memory (Optional[Memory], optional): An instance of Memory. Defaults to None.

ParamsType

None

Context

None

OutputType

None

NeedAgent: bool

False

name: str = <Multiline-String>

description: str = <Multiline-String>

actions: dict[str, str]

None

classmethod default_params() → ParamsType

classmethod default_context() → Context

classmethod __init_subclass__(**kwargs)

set_agent(agent: Any)

property agent: Any

property toolbox: agentsociety.agent.toolbox.AgentToolbox

property llm: agentsociety.llm.LLM

property memory: agentsociety.memory.Memory

property agent_memory: agentsociety.memory.Memory

property block_memory: agentsociety.memory.KVMemory

property environment: agentsociety.environment.Environment

async before_forward()

Before forward - prepare context

async after_forward()

After forward - update/clean context

abstract async forward(agent_context: agentsociety.agent.context.DotDict)

• Description:

  • Each block performs a specific reasoning task. This method should be overridden by subclasses.

  • Subclasses can define their own parameters as needed.

• Args:

  • *args: Variable length argument list.

  • **kwargs: Arbitrary keyword arguments.

• Raises:

  • NotImplementedError: Subclasses must implement this method.