AgentDecision class

The action decision produced by the LLM Core after reasoning.

The LLM may decide to:

Constructors

AgentDecision({required AgentDecisionType type, String? toolName, Map<String, dynamic>? toolInput, String? thought, String? replanReason, double confidence = 1.0, String? rawResponse})
const
AgentDecision.complete(String reason, {double confidence = 1.0})
Convenience factory for a goal-complete signal.
factory
AgentDecision.replan(String reason)
Convenience factory for a re-plan request.
factory
AgentDecision.think(String thought, {double confidence = 0.8})
Convenience factory for a pure-thinking step.
factory
AgentDecision.useTool(String tool, Map<String, dynamic> input, {String? thought, double confidence = 0.9})
Convenience factory for a tool-use decision.
factory

Properties

confidence → double
How confident the LLM is in this decision.
final
hashCode → int
The hash code for this object.
no setterinherited
rawResponse → String?
Raw LLM response string before parsing.
final
replanReason → String?
Explanation for why a re-plan is needed.
final
runtimeType → Type
A representation of the runtime type of the object.
no setterinherited
thought → String?
The reasoning text the LLM produced.
final
toolInput → Map<String, dynamic>?
Input arguments for the tool.
final
toolName → String?
When type == AgentDecisionType.useTool, the tool to invoke.
final
type → AgentDecisionType
The kind of action the agent decided to take.
final

Methods

noSuchMethod(Invocation invocation) → dynamic
Invoked when a nonexistent method or property is accessed.
inherited
toString() → String
A string representation of this object.
override

Operators

operator ==(Object other) → bool
The equality operator.
inherited