Reference

Class AgentNetwork

Variable

Variable	Data Type	Default	Description
id	UUID4	uuid.uuid4()	Stores auto-generated ID as identifier. Not editable.
name	str	None	Stores a name of the network.
members	List[Member]	list()	Stores a list of Member objects.
formation	Formation	None	Stores Formation enum.
should_reform	bool	False	Whether to reform the network during the activation.
network_tasks	List[Task]	list()	A list of Task objects unassigned to any network members.
prompt_file	str	None	Absolute file path to the prompt file w/ JSON formatted prompt.
process	TaskHandlingProcess	TaskHandlingProcess.SEQUENTIAL	Enum of the task handling process.
consent_trigger	Callable[..., Any]	None	A trigger event (func) for consentual processing.
pre_launch_callbacks	List[Callable[..., Any]]	list()	Stores callbacks to run before the network launch.
post_launch_callbacks	List[Callable[..., Any]]	list()	Stores callbacks to run after the network launch.
step_callbacks	Callable[..., Any]	None	Stores callbacks to run at every step of each member agent takes during the activation.
cache	bool	True	Whether to store cache.
execution_logs	List[Dict[str, Any]]	list()	Stores a list of execution logs of all the tasks in the network.

Class Methods

Method	Params	Returns	Description
launch	kwargs_pre: Optional[Dict[str, str]] = None kwargs_post: Optional[Dict[str, Any]] = None start_index: int = None	Tuple[TaskOutput, TaskGraph]:	Core method to launch the network and execute tasks

Properties

Property	Returns	Description
key	str	Unique identifier.
managers	List[InstanceOf[Member]]	A list of manager members.
manager_tasks	List[InstanceOf[Task]]	A list of tasks handled by managers.
tasks	List[InstanceOf[Task]]	All the tasks in the network.
unassigned_member_tasks	List[InstanceOf[Task]]	Unassigned member-level tasks.

---

Class Member

Variable

Variable	Data Type	Default	Description
agent	InstanceOf[Agent]	None	Agent as a member
is_manager	bool	False	Whether the member is a manager.
can_share_knowledge	bool	True	Whether the member can share its knowledge among the other network members.
can_share_memory	bool	True	Whether the member can share its memories among the other network members.
tasks	List[InstanceOf[Task]]	list()	Assinged tasks.

Properties

Property	Returns	Description
is_idling	bool	Whether it has unprocessed assgined task/s

---

Class Agent

Variables

Variable	Data Type	Default	Description
id	UUID4	uuid.uuid4()	Stores auto-generated ID as identifier. Not editable.
role	str	None	Stores a role of the agent.
goal	str	None	Stores a goal of the agent.
backstory	str	None	Stores backstory of the agent. Utilized as system prompt.
self_learn	bool	False	Whether to self-learn
tools	List[InstanceOf[ToolSet] | Type[Tool]]	None	Stores tools to be used when executing a task.
knowledge_sources	List[BaseKnowledgeSource | Any]	None	Stores knowledge sources in text, file path, or url.
embedder_config	Dict[str, Any]	None	Stores embedding configuration for storing knowledge sources.
with_memory	bool	False	Whether to store tasks and results in memory.
memory_config	Dict[str, Any]	None	Stores configuration of the memory.
short_term_memory	InstanceOf[ShortTermMemory]	None	Stores ShortTermMemory object.
long_term_memory	InstanceOf[LongTermMemory]	None	Stores LongTermMemory object.
user_memory	InstanceOf[UserMemory]	None	Stores UserMemory object.
use_developer_prompt	bool	True	Whether to use the system (developer) prompt when calling the model.
developer_promt_template	str	None	File path to the prompt template.
user_promt_template	str	None	File path to the prompt template.
networks	List[Any]	list()	Stores a list of agent networks that the agent belongs to.
allow_delegation	bool	False	Whether the agent can delegate assinged tasks to another agent.
max_retry_limit	int	2	Maximum number of retries when the task execution failed.
maxit	int	25	Maximum number of total optimization loops during error handling.
callbacks	List[Callabale]	None	Stores a list of step callback functions.
llm	str | InstanceOf[LLM] | Dict[str, Any]	None	Stores the main LLM.
func_calling_llm	str | InstanceOf[LLM] | Dict[str, Any]	None	Stores the function calling LLM.
respect_context_window	bool	True	Whether to follow the main model's maximum context window size.
max_execution_time	int	None	Stores maximum execution time in seconds.
max_rpm	int	None	Stores maximum number of requests per minute.
llm_config	Dict[str, Any]	None	Stores configuration of LLM object.
config	Dict[str, Any]	None	Stores model config.

Class Methods

Method	Params	Returns	Description
update	**kwargs: Any	Self	Updates agents with given kwargs. Invalid keys will be ignored.
start	context: Any = None tool_res_as_final: bool = False	TaskOutput | None	Starts to operate the agent.
execute_task	task: [Task] context: Any = None	Tuple[str, str, Any, UsageMetrics]	Returns response from the model in plane text format.

Properties

Property	Returns	Description
key	str	Unique identifier of the agent using its ID and sanitized role.

ENUM Formation

class Formation(IntEnum):
    SOLO = 1
    SUPERVISING = 2
    SQUAD = 3
    RANDOM = 4
    HYBRID = 10

ENUM TaskHandlingProcess

class TaskHandlingProcess(IntEnum):
    HIERARCHY = 1
    SEQUENTIAL = 2
    CONSENSUAL = 3