Class DebatePlanner

java.lang.Object

dev.langchain4j.agentic.patterns.debate.DebatePlanner

All Implemented Interfaces:

Planner

public class DebatePlanner
extends Object
implements Planner

A debate planner that orchestrates iterative adversarial rounds among debater agents.

Debaters generate independent positions in parallel, then enter critique rounds where they see each other's reasoning (via a "debateContext" scope key) and can revise their positions. Rounds continue until the ConvergenceStrategy reports convergence or maxRounds is reached, at which point the judge agent (the last registered subagent) renders a final verdict.

Field Summary

Fields

Modifier and Type	Field	Description
static final String	DEBATE_CONTEXT_KEY

Constructor Summary

Constructors

Constructor	Description
DebatePlanner()
DebatePlanner(int maxRounds)
DebatePlanner(int maxRounds, ConvergenceStrategy convergenceStrategy)
DebatePlanner(ConvergenceStrategy convergenceStrategy)

Method Summary

Modifier and Type	Method	Description
Map<String,Object>	executionState()	Returns the planner's current execution state as a map of serializable values.
Action	firstAction(PlanningContext planningContext)
void	init(InitPlanningContext initPlanningContext)
Action	nextAction(PlanningContext planningContext)
void	restoreExecutionState(Map<String,Object> state)	Restores the planner's execution state from a previously saved map.
AgenticSystemTopology	topology()

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface Planner

as, call, call, done, done, noOp, terminated

Field Details

DEBATE_CONTEXT_KEY

public static final String DEBATE_CONTEXT_KEY

See Also:

• Constant Field Values

Constructor Details

DebatePlanner

public DebatePlanner()

DebatePlanner

public DebatePlanner(int maxRounds)

DebatePlanner

public DebatePlanner(ConvergenceStrategy convergenceStrategy)

DebatePlanner

public DebatePlanner(int maxRounds,
 ConvergenceStrategy convergenceStrategy)

Method Details

init

public void init(InitPlanningContext initPlanningContext)

Specified by:

init in interface Planner

firstAction

public Action firstAction(PlanningContext planningContext)

Specified by:

firstAction in interface Planner

nextAction

public Action nextAction(PlanningContext planningContext)

Specified by:

nextAction in interface Planner

executionState

public Map<String,Object> executionState()

Description copied from interface: Planner

Returns the planner's current execution state as a map of serializable values. This state is persisted to the AgenticScope after each agent invocation, enabling the planner to resume from the correct position after a crash.

The returned state must be such that, when passed to Planner.restoreExecutionState(Map) and Planner.firstAction(PlanningContext) is called, the planner produces the correct resume action.

Stateless planners (e.g., parallel, conditional) can use the default empty implementation.

Specified by:

executionState in interface Planner

Returns:

a map of state entries to persist, or an empty map if no state needs saving

restoreExecutionState

public void restoreExecutionState(Map<String,Object> state)

Description copied from interface: Planner

Restores the planner's execution state from a previously saved map. Called by the execution loop before Planner.firstAction(PlanningContext) when recovering from a persisted scope.

Specified by:

restoreExecutionState in interface Planner

Parameters:

state - the previously saved execution state

topology

public AgenticSystemTopology topology()

Specified by:

topology in interface Planner