langchain.js
    Preparing search index...

    Class LLMSingleActionAgent

    Class representing a single action agent using a LLMChain in LangChain. Extends the BaseSingleActionAgent class and provides methods for planning agent actions based on LLMChain outputs.

    const customPromptTemplate = new CustomPromptTemplate({
      tools: [new Calculator()],
      inputVariables: ["input", "agent_scratchpad"],
    });
    const customOutputParser = new CustomOutputParser();
    const agent = new LLMSingleActionAgent({
      llmChain: new LLMChain({
        prompt: customPromptTemplate,
        llm: new ChatOpenAI({ model: "gpt-4o-mini", temperature: 0 }),
      }),
      outputParser: customOutputParser,
      stop: ["\nObservation"],
    });
    const executor = new AgentExecutor({
      agent,
      tools: [new Calculator()],
    });
    const result = await executor.invoke({
      input:
        "Who is Olivia Wilde's boyfriend? What is his current age raised to the 0.23 power?",
    });

    Hierarchy (View Summary)

    Index

    Constructors

    constructor

    Properties

    lc_namespace llmChain outputParser stop? ToolType

    Accessors

    allowedTools inputKeys returnValues

    Methods

    _agentActionType _agentType plan prepareForOutput returnStoppedResponse

    Constructors

    Properties

    lc_namespace: string[] = ...
    llmChain: LLMChain
    outputParser: AgentActionOutputParser
    stop?: string[]
    ToolType: StructuredToolInterface

    Accessors

    • get allowedTools(): undefined | string[]

      Returns undefined | string[]

    • get inputKeys(): string[]

      Returns string[]

    • get returnValues(): string[]

      Returns string[]

    Methods

    _agentActionType

    • _agentActionType(): string

      Return the string type key uniquely identifying multi or single action agents.

      Returns string

    _agentType

    • _agentType(): string

      Return the string type key uniquely identifying this class of agent.

      Returns string

    • Decide what to do given some input.

      Parameters

      • steps: AgentStep[]

        Steps the LLM has taken so far, along with observations from each.

      • inputs: ChainValues

        User inputs.

      • OptionalcallbackManager: any

        Callback manager.

      Returns Promise<any>

      Action specifying what tool to use.

    • Prepare the agent for output, if needed

      Parameters

      • _returnValues: AgentFinish
      • _steps: AgentStep[]

      Returns Promise<AgentFinish>

    • Return response when agent has been stopped due to max iterations

      Parameters

      • earlyStoppingMethod: StoppingMethod
      • _steps: AgentStep[]
      • _inputs: ChainValues
      • Optional_callbackManager: any

      Returns Promise<AgentFinish>

    Settings

    Member Visibility

    On This Page

    Constructors
    Properties
    Accessors
    Methods