gbox-sdk
    Preparing search index...

    Interface ActionAI

    interface ActionAI {
        background?: string;
        includeScreenshot?: boolean;
        instruction: string;
        outputFormat?: "base64" | "storageKey";
        screenshotDelay?:
            | `${number}h`
            | `${number}m`
            | `${number}s`
            | `${number}ms`;
        settings?: Settings;
    }

    Hierarchy (View Summary)

    Index

    Properties

    background?: string

    The background of the UI action to perform. The purpose of background is to let the action executor to understand the context of why the instruction is given including important previous actions and observations

    includeScreenshot?: boolean

    Whether to include screenshots in the action response. If false, the screenshot object will still be returned but with empty URIs. Default is false.

    instruction: string

    Direct instruction of the UI action to perform (e.g., 'click the login button', 'input username in the email field', 'scroll down', 'swipe left')

    outputFormat?: "base64" | "storageKey"

    Type of the URI. default is base64.

    screenshotDelay?: `${number}h` | `${number}m` | `${number}s` | `${number}ms`

    Delay after performing the action, before taking the final screenshot.

    Execution flow:

    1. Take screenshot before action
    2. Perform the action
    3. Wait for screenshotDelay (this parameter)
    4. Take screenshot after action

    Example: '500ms' means wait 500ms after the action before capturing the final screenshot.

    Supported time units: ms (milliseconds), s (seconds), m (minutes), h (hours) Example formats: "500ms", "30s", "5m", "1h" Default: 500ms Maximum allowed: 30s

    settings?: Settings

    AI action settings