agentsociety.configs.exp

Module Contents

Classes

WorkflowType	Defines the types of workflow steps in the simulation.
AgentFilterConfig	Configuration for filtering agents.
WorkflowStepConfig	Represents a step in the workflow process.
ExpConfig	Main configuration for the experiment.

Data

__all__

API

agentsociety.configs.exp.__all__

[‘WorkflowStepConfig’, ‘EnvironmentConfig’, ‘ExpConfig’, ‘WorkflowType’, ‘AgentFilterConfig’]

class agentsociety.configs.exp.WorkflowType

Bases: str, enum.Enum

Defines the types of workflow steps in the simulation.

• Description:

  • Enumerates different types of workflow steps that can be executed during simulation.

• Types:

  • STEP: Execute on a step-by-step unit.

  • RUN: Execute on a daily unit (day-based execution).

  • INTERVIEW: Sends an interview message to the specified agent.

  • SURVEY: Sends a questionnaire to the specified agent.

  • ENVIRONMENT_INTERVENE: Changes the environment variables (global prompt).

  • UPDATE_STATE_INTERVENE: Directly updates the state information of the specified agent.

  • MESSAGE_INTERVENE: Influences the agent’s behavior and state by sending a message.

  • NEXT_ROUND: Proceed to the next round of the simulation —— reset agents but keep the memory.

  • DELETE_AGENT: Delete the specified agents.

  • SAVE_CONTEXT: Save the context of the specified agents.

  • INTERVENE: Represents other intervention methods driven by code.

  • FUNCTION: Represents function-based intervention methods.

Initialization

Initialize self. See help(type(self)) for accurate signature.

STEP

‘step’

RUN

‘run’

INTERVIEW

‘interview’

SURVEY

‘survey’

UPDATE_STATE_INTERVENE

‘update_state’

MESSAGE_INTERVENE

‘message’

DELETE_AGENT

‘delete_agent’

ENVIRONMENT_INTERVENE

‘environment’

NEXT_ROUND

‘next_round’

SAVE_CONTEXT

‘save_context’

INTERVENE

‘other’

FUNCTION

‘function’

class agentsociety.configs.exp.AgentFilterConfig

Bases: pydantic.BaseModel

Configuration for filtering agents.

agent_class: Optional[Union[tuple[type[agentsociety.agent.Agent]], list[str]]]

None

The class of the agent to filter

filter_str: Optional[str]

None

The filter string of the agent to filter

validate_func()

class agentsociety.configs.exp.WorkflowStepConfig

Bases: pydantic.BaseModel

Represents a step in the workflow process.

model_config

‘ConfigDict(…)’

type: agentsociety.configs.exp.WorkflowType

‘Field(…)’

The type of the workflow step

func: Optional[Union[collections.abc.Callable, str]]

None

The function that extracts the metric - used for [FUNCTION] type

days: float

1

Duration (in days) for which this step lasts - used for [RUN] type

steps: int

1

Number of steps for which this step lasts - used for [STEP] type

ticks_per_step: int

300

Number of ticks per step - used for [RUN, STEP] type. For example, if it is 300, then the step will run 300 ticks in the environment.

target_agent: Optional[Union[list[int], agentsociety.configs.exp.AgentFilterConfig]]

None

List specifying the agents targeted by this step - used for [INTERVIEW, SURVEY, UPDATE_STATE_INTERVENE, MESSAGE_INTERVENE, DELETE_AGENT, SAVE_CONTEXT] type

interview_message: Optional[str]

None

Optional message used for interviews during this step - used for [INTERVIEW] type

survey: Optional[agentsociety.survey.Survey]

None

Optional survey instance associated with this step - used for [SURVEY] type

key: Optional[str]

None

Optional key identifier for the step - used for [ENVIRONMENT_INTERVENE, UPDATE_STATE_INTERVENE, SAVE_CONTEXT] type

save_as: Optional[str]

None

Optional key identifier for the step - used for [SAVE_CONTEXT] type

value: Optional[Any]

None

Optional value associated with the step - used for [ENVIRONMENT_INTERVENE, UPDATE_STATE_INTERVENE] type

intervene_message: Optional[str]

None

Optional message used for interventions - used for [MESSAGE_INTERVENE] type

description: Optional[str]

None

A descriptive text explaining the workflow step

serialize_func(func, info)

validate_func()

class agentsociety.configs.exp.ExpConfig

Bases: pydantic.BaseModel

Main configuration for the experiment.

model_config

‘ConfigDict(…)’

name: str

‘Field(…)’

Name of the experiment

id: uuid.UUID

‘Field(…)’

Unique identifier for the experiment

workflow: List[agentsociety.configs.exp.WorkflowStepConfig]

‘Field(…)’

List of workflow steps

environment: agentsociety.environment.EnvironmentConfig

None

Environment configuration

serialize_id(id, info)