Interface defining the structure of input data for creating an AgentExecutor. It extends ChainInputs and includes additional properties specific to agent execution.

interface AgentExecutorInput {
    agent: BaseSingleActionAgent | BaseMultiActionAgent | Runnable<ChainValues & {
        steps?: AgentStep[];
    }, AgentAction | AgentFinish | AgentAction[], RunnableConfig<Record<string, any>>>;
    callbackManager?: CallbackManager;
    callbacks?: Callbacks;
    earlyStoppingMethod?: StoppingMethod;
    handleParsingErrors?: string | boolean | ((e: OutputParserException | ToolInputParsingException) => string);
    handleToolRuntimeErrors?: ((e: Error) => string);
    maxIterations?: number;
    memory?: BaseMemory;
    metadata?: Record<string, unknown>;
    returnIntermediateSteps?: boolean;
    tags?: string[];
    tools: (StructuredToolInterface<ZodObjectAny> | StructuredToolInterface<ZodObjectAny>)[];
    verbose?: boolean;
}

Hierarchy (view full)

Properties

agent: BaseSingleActionAgent | BaseMultiActionAgent | Runnable<ChainValues & {
    steps?: AgentStep[];
}, AgentAction | AgentFinish | AgentAction[], RunnableConfig<Record<string, any>>>
callbackManager?: CallbackManager

Use callbacks instead

callbacks?: Callbacks
earlyStoppingMethod?: StoppingMethod
handleParsingErrors?: string | boolean | ((e: OutputParserException | ToolInputParsingException) => string)
handleToolRuntimeErrors?: ((e: Error) => string)
maxIterations?: number
memory?: BaseMemory
metadata?: Record<string, unknown>
returnIntermediateSteps?: boolean
tags?: string[]
tools: (StructuredToolInterface<ZodObjectAny> | StructuredToolInterface<ZodObjectAny>)[]
verbose?: boolean