agents.experiments.supervisor.base_supervisor¶

Base supervisor implementation for multi-agent systems.

This module provides the core supervisor classes that can manage multiple agents, handle tool synchronization, and support dynamic agent creation.

Classes¶

`AgentMetadata`	Metadata for a registered agent.
`BaseSupervisor`	Base supervisor agent for managing multiple agents.
`DynamicSupervisor`	Extended supervisor with dynamic agent creation capabilities.
`SupervisorState`	State model for supervisor agents.

Module Contents¶

class agents.experiments.supervisor.base_supervisor.AgentMetadata(/, **data)¶

Bases: pydantic.BaseModel

Metadata for a registered agent.

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.

Parameters:: data (Any)

class agents.experiments.supervisor.base_supervisor.BaseSupervisor(name, engine, agents=None, **kwargs)¶

Bases: haive.agents.react.agent.ReactAgent

Base supervisor agent for managing multiple agents.

This supervisor can register agents, delegate tasks, and coordinate multi-agent workflows.

Initialize the supervisor.

Parameters:

name (str) – Supervisor name
engine (haive.core.engine.aug_llm.AugLLMConfig) – LLM configuration
agents (dict[str, Any] | None) – Optional initial agents to register
**kwargs – Additional arguments passed to ReactAgent

delegate_task(agent_name, task)¶

Delegate a task to a specific agent.

Parameters:

agent_name (str) – Name of the agent to delegate to
task (str) – Task description

Returns:

Result from the agent

Return type:

str

get_agent(name)¶

Get a registered agent by name.

Parameters:: name (str) – Agent identifier
Returns:: The agent instance if found, None otherwise
Return type:: Any | None

get_execution_status()¶

Get current execution status.

Returns:: Status information including active agent and recent history
Return type:: dict[str, Any]

list_agents()¶

List all registered agents.

Returns:: Dictionary of agent metadata
Return type:: dict[str, AgentMetadata]

register_agent(name, description, agent)¶

Parameters:

name (str) – Agent identifier
description (str) – Description of agent capabilities
agent (Any) – The agent instance

Return type:

None

class agents.experiments.supervisor.base_supervisor.DynamicSupervisor(*args, **kwargs)¶

Bases: BaseSupervisor

Extended supervisor with dynamic agent creation capabilities.

This supervisor can create new agents on demand based on requirements.

Initialize dynamic supervisor.

create_agent(name, description, agent_type='simple')¶

Create a new agent dynamically.

Parameters:

name (str) – Agent name
description (str) – Agent description
agent_type (str) – Type of agent to create

Returns:

True if agent was created successfully

Return type:

bool

enable_agent_creation(enable=True)¶

Enable or disable dynamic agent creation.

Parameters:: enable (bool) – Whether to enable agent creation
Return type:: None

get_capabilities()¶

Get supervisor capabilities.

Returns:: Dictionary describing supervisor capabilities
Return type:: dict[str, Any]

class agents.experiments.supervisor.base_supervisor.SupervisorState(/, **data)¶

Bases: pydantic.BaseModel

State model for supervisor agents.

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.

Parameters:: data (Any)