agents.multi.experiments.proper_list_multi_agent¶

Proper list multi-agent that uses MultiAgentState and AgentNodeV3.

from typing import Any This implementation properly leverages the existing infrastructure: - MultiAgentState for proper state management - AgentNodeV3 for agent execution with state projection - create_agent_node_v3 for creating agent nodes - Proper engine syncing and recompilation tracking

Classes¶

`MetaListMultiAgent`	List multi-agent that uses MetaStateSchema for single agent embedding.
`ProperListMultiAgent`	List-based multi-agent that properly uses MultiAgentState and AgentNodeV3.

Functions¶

`meta_multi`(*agents[, name])	Create a meta multi-agent.
`sequential_multi`(*agents[, name])	Create a sequential multi-agent.

Module Contents¶

class agents.multi.experiments.proper_list_multi_agent.MetaListMultiAgent¶

Bases: haive.agents.base.agent.Agent, haive.core.common.mixins.recompile_mixin.RecompileMixin, collections.abc.Sequence[haive.agents.base.agent.Agent]

List multi-agent that uses MetaStateSchema for single agent embedding.

This is useful when you want to embed a sequence of agents as a single unit within another agent’s state using the MetaStateSchema pattern.

Examples

# Create a meta multi-agent
meta = MetaListMultiAgent("research_pipeline")
meta.append(PlannerAgent())
meta.append(ResearchAgent())
meta.append(WriterAgent())

# This can be embedded in another agent's state
parent_state = MetaStateSchema(agent=meta)

append(agent)¶

Add agent to the list.

Parameters:: agent (haive.agents.base.agent.Agent)
Return type:: MetaListMultiAgent

build_graph()¶

Build graph that executes agents sequentially.

Return type:: haive.core.graph.state_graph.base_graph2.BaseGraph

setup_agent()¶

Setup the meta multi-agent.

Return type:: None

class agents.multi.experiments.proper_list_multi_agent.ProperListMultiAgent¶

Bases: haive.agents.base.agent.Agent, haive.core.common.mixins.recompile_mixin.RecompileMixin, collections.abc.Sequence[haive.agents.base.agent.Agent]

List-based multi-agent that properly uses MultiAgentState and AgentNodeV3.

This implementation: - Uses MultiAgentState as the state schema - Uses AgentNodeV3 for proper agent execution - Handles state projection and hierarchical management - Supports recompilation tracking - Maintains the natural list interface

Examples

multi = ProperListMultiAgent("research_team")
multi.append(PlannerAgent("planner"))
multi.append(ResearchAgent("researcher"))
multi.append(WriterAgent("writer"))

# Agents are stored in MultiAgentState.agents
# Each agent gets its own state in MultiAgentState.agent_states
# Output tracked in MultiAgentState.agent_outputs

result = multi.invoke({"messages": [HumanMessage("Research AI")]})

append(agent)¶

Add agent to the list.

Parameters:: agent (haive.agents.base.agent.Agent)
Return type:: ProperListMultiAgent

build_graph()¶

Build graph using AgentNodeV3 for proper state handling.

Return type:: haive.core.graph.state_graph.base_graph2.BaseGraph

get_agent_by_name(name)¶

Get agent by name.

Parameters:: name (str)
Return type:: haive.agents.base.agent.Agent | None

get_agent_names()¶

Get ordered list of agent names.

Return type:: list[str]

get_execution_summary()¶

Get execution summary.

Return type:: dict[str, Any]

insert(index, agent)¶

Insert agent at position.

Parameters:

index (int)
agent (haive.agents.base.agent.Agent)

Return type:

ProperListMultiAgent

model_post_init(__context)¶

Initialize after model creation.

Return type:: None

pop(index=-1)¶

Remove and return agent.

Parameters:: index (int)
Return type:: haive.agents.base.agent.Agent

remove(agent)¶

Remove agent.

Parameters:: agent (haive.agents.base.agent.Agent | str)
Return type:: ProperListMultiAgent

setup_agent()¶

Setup the multi-agent system.

Return type:: None

then(agent)¶

Add next agent in sequence.

Parameters:: agent (haive.agents.base.agent.Agent)
Return type:: ProperListMultiAgent

when(condition, routes)¶

Add conditional routing for the last agent.

Parameters:

condition (collections.abc.Callable[[Any], str | bool])
routes (dict[str | bool, Union[str, haive.agents.base.agent.Agent, langgraph.graph.END]])

Return type:

ProperListMultiAgent

agents.multi.experiments.proper_list_multi_agent.meta_multi(*agents, name='meta_multi')¶

Create a meta multi-agent.

Parameters:

agents (haive.agents.base.agent.Agent)
name (str)

Return type:

MetaListMultiAgent

agents.multi.experiments.proper_list_multi_agent.sequential_multi(*agents, name='sequential_multi')¶

Create a sequential multi-agent.

Parameters:

agents (haive.agents.base.agent.Agent)
name (str)

Return type:

ProperListMultiAgent