agents.experiments.supervisor

Supervisor module for managing multi-agent systems.

This module provides a complete supervisor implementation that can manage multiple agents, handle tool synchronization, and support dynamic agent creation.

Key Components:

  • BaseSupervisor: Core supervisor using ReactAgent

  • DynamicSupervisor: Extended supervisor with agent creation

  • SupervisorState: State model with agent registry

  • DynamicSupervisorState: Extended state for dynamic capabilities

Example Usage:

Basic supervisor:

from haive.agents.experiments.supervisor import BaseSupervisor
from haive.agents.simple.agent import SimpleAgent

# Create supervisor
supervisor = BaseSupervisor(name="my_supervisor", engine=my_engine)

# Register agents
research_agent = SimpleAgent(name="researcher", engine=research_engine)
supervisor.register_agent("research", "Research specialist", research_agent)

# Use supervisor
result = supervisor.invoke("Research quantum computing trends")

Dynamic supervisor:

from haive.agents.experiments.supervisor import DynamicSupervisor

supervisor = DynamicSupervisor(name="dynamic_super", engine=my_engine)
supervisor.enable_agent_creation()

# Can create agents on the fly via tool calls
result = supervisor.invoke("Create a coding agent and write Python code")

Submodules