Class AgentAbstract

constructor

• new Agent(input): Agent

• Parameters

  • input: AgentInput

  Returns Agent

llmChain

outputParser

allowedTools

• get allowedTools(): undefined | string[]

• Returns undefined | string[]

inputKeys

• get inputKeys(): string[]

• Returns string[]

constructScratchPad

• constructScratchPad(steps): Promise<string | BaseMessage[]>

• Construct a scratchpad to let the agent continue its thought process

  Parameters

  • steps: AgentStep[]

  Returns Promise<string | BaseMessage[]>

finishToolName

• finishToolName(): string

• Name of tool to use to terminate the chain.

  Returns string

AbstractllmPrefix

• llmPrefix(): string

• Prefix to append the LLM call with.

  Returns string

AbstractobservationPrefix

• observationPrefix(): string

• Prefix to append the observation with.

  Returns string

plan

• plan(steps, inputs, callbackManager?): Promise<AgentAction | AgentFinish>

• 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: CallbackManager

    Callback manager to use for this call.

  Returns Promise<AgentAction | AgentFinish>

  Action specifying what tool to use.

returnStoppedResponse

• returnStoppedResponse(earlyStoppingMethod, steps, inputs, callbackManager?): Promise<AgentFinish>

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

  Parameters

  • earlyStoppingMethod: StoppingMethod
  • steps: AgentStep[]
  • inputs: ChainValues
  • OptionalcallbackManager: CallbackManager

  Returns Promise<AgentFinish>

StaticcreatePrompt

• createPrompt(_tools, _fields?): BasePromptTemplate<any, BasePromptValueInterface, any>

• Create a prompt for this class

  Parameters

  • _tools: StructuredToolInterface<ToolInputSchemaBase, any, any>[]

    List of tools the agent will have access to, used to format the prompt.
  • Optional_fields: Record<string, any>

    Additional fields used to format the prompt.

  Returns BasePromptTemplate<any, BasePromptValueInterface, any>

  A PromptTemplate assembled from the given tools and fields.

Staticdeserialize

• deserialize(data): Promise<Agent>

• Load an agent from a json-like object describing it.

  Parameters

  • data: SerializedZeroShotAgent & {
        llm?: BaseLanguageModelInterface<any, BaseLanguageModelCallOptions>;
        tools?: ToolInterface<StringInputToolSchema, any, any>[];
    }

  Returns Promise<Agent>

StaticfromLLMAndTools

• fromLLMAndTools(_llm, _tools, _args?): Agent

• Construct an agent from an LLM and a list of tools

  Parameters

  • _llm: BaseLanguageModelInterface<any, BaseLanguageModelCallOptions>
  • _tools: StructuredToolInterface<ToolInputSchemaBase, any, any>[]
  • Optional_args: AgentArgs

  Returns Agent

StaticgetDefaultOutputParser

• getDefaultOutputParser(_fields?): AgentActionOutputParser

• Get the default output parser for this agent.

  Parameters

  • Optional_fields: OutputParserArgs

  Returns AgentActionOutputParser

StaticvalidateTools

• validateTools(_tools): void

• Validate that appropriate tools are passed in

  Parameters

  • _tools: StructuredToolInterface<ToolInputSchemaBase, any, any>[]

  Returns void