ART Framework API Docs
    Preparing search index...

    Interface ExecutionOutput

    Represents the structured output from an LLM execution call during the PES Agent's execution phase. This is parsed from the raw LLM response.

    ExecutionOutput

    interface ExecutionOutput {
        content?: string;
        nextStepDecision?: "continue" | "wait" | "complete_item" | "update_plan";
        thoughts?: string;
        toolCalls?: ParsedToolCall[];
        updatedPlan?: { intent?: string; plan?: string; todoList?: TodoItem[] };
    }
    Index

    Properties

    content? nextStepDecision? thoughts? toolCalls? updatedPlan?

    Properties

    Optionalcontent

    content?: string

    The main response content from the LLM. This is the primary textual output from the execution step.

    nextStepDecision?: "continue" | "wait" | "complete_item" | "update_plan"

    The agent's decision on how to proceed after this execution.

    • 'continue': Proceed to the next iteration or step.
    • 'wait': Pause and wait for external input (rare in execution phase).
    • 'complete_item': Mark the current item as complete and move to the next.
    • 'update_plan': Modify the execution plan (intent, todo list, etc.).
    thoughts?: string

    The agent's thoughts or reasoning for this execution step. This may include decision-making logic or reflections.

    toolCalls?: ParsedToolCall[]

    Any tool calls the LLM decided to make during execution. These are parsed and will be executed by the ToolSystem.

    updatedPlan?: { intent?: string; plan?: string; todoList?: TodoItem[] }

    Updates to the plan if the agent decided to modify it. This can include changes to intent, plan description, or the todo list.

    Settings

    Member Visibility

    On This Page

    Properties