Function generateText

Parameters

• args: {
      prompt: string;
      promptDir?: string;
      skills?:
          | Skill[]
          | (
              (
                  variables?: Record<string, string | number | boolean>,
              ) => Skill[] | Promise<Skill[]>
          );
      variables?: Record<string, string | number | boolean>;
  } & Partial<Omit<CallSettings, "maxOutputTokens">> & {
      activeTools?: (keyof Tools)[];
      maxSteps?: number;
      onStepFinish?: GenerateTextOnStepFinishCallback<Tools>;
      output?: Output;
      prepareStep?: PrepareStepFunction<Tools>;
      stopWhen?: StopCondition<Tools> | StopCondition<Tools>[];
      toolChoice?: ToolChoice<Tools>;
      tools?: Tools;
  }

  Generation arguments.

  • prompt: string

    Prompt file name.

  • OptionalpromptDir?: string

  • Optionalskills?:
        | Skill[]
        | (
            (
                variables?: Record<string, string | number | boolean>,
            ) => Skill[] | Promise<Skill[]>
        )

    Skill packages to provide to the LLM. Injects {{ _system_skills }} and adds the load_skill tool. Can be a static array or a function that receives the resolved variables and returns skills.

  • Optionalvariables?: Record<string, string | number | boolean>

    Variables to interpolate.

  • prompt

    Prompt file name.

  • variables

    Variables to interpolate.

  • tools

    Tools the model can call (optional).

  • toolChoice

    Tool selection strategy (optional).

  • OptionalactiveTools?: (keyof Tools)[]

    Limit which tools are active without changing types

  • OptionalmaxSteps?: number

    Maximum number of automatic tool execution rounds (multi-step)

  • OptionalonStepFinish?: GenerateTextOnStepFinishCallback<Tools>

    Callback after each step completes

  • Optionaloutput?: Output

    Structured output specification (e.g., Output.object({ schema }))

  • OptionalprepareStep?: PrepareStepFunction<Tools>

    Customize each step before execution

  • OptionalstopWhen?: StopCondition<Tools> | StopCondition<Tools>[]

    Custom stop conditions for multi-step execution

  • OptionaltoolChoice?: ToolChoice<Tools>

    Tool choice strategy: 'auto', 'none', 'required', or specific tool

  • Optionaltools?: Tools

    Tools the model can call