bricks/docs/en/llmout.md

UserInputView

Control Functionality: Converts user-input data (including text, images, audio, video, etc.) into Markdown format for display. Audio and video media are presented using standalone player components.
Type: Container control
Parent Control: bricks.VBox

Initialization Parameters

• opts (Object) - Configuration options, inheriting general parameters from VBox:
  • data (Object) - Raw data object provided by the user, where keys correspond to field names and values to their respective content.
  • input_fields (Array) - Array of field definitions, each containing:
    • name (String) - Field name (e.g., video1, audio2, image3, plain text fields, etc.)
    • label (String, optional) - Display label used as a heading in Markdown output.

Note

: Blob objects in data are converted to previewable URLs via URL.createObjectURL().

Main Events

No explicit custom events. Content layout is automatically refreshed when show_input(data) is called.

---

LlmOut

Control Functionality: Dynamically constructs a response interface based on JSON data returned from a large language model (LLM), supporting rendering of various content types such as reasoning text, final responses, error messages, audio, video, and images.
Type: Container control
Parent Control: bricks.VBox

Initialization Parameters

• opts (Object) - Configuration options, inheriting general parameters from VBox. No special constructor parameters.

Internal state properties (managed by the update(data) method):

• reasoning_content (String) - Accumulated reasoning process text.
• content (String) - Accumulated final response content.
• error (String) - Error message text.
• images (Array) - List of image URLs.
• Media widget instances: rc_w, c_w, v_w, a_w, i_w, etc.

Main Events

update(data)

Trigger Condition: Manually invoked when new output data from the LLM is received.

The data parameter (Object) supports the following optional properties:

Property	Type	Description
reasoning_content	String	Text describing the reasoning process, typically rendered with Markdown
content	String	Final response content, supports Markdown formatting
error	String	Error message, displayed in red styling
audio	String	Audio resource, either a URL or Base64-encoded string; protocol prefix is handled automatically
video	String	Video resource, supports URL or Base64 encoding
image	String or Array[String]	Single image URL or an array of multiple image URLs

Behavior Details:

• Newly received content is automatically merged and appended to existing fields (append mode).
• When a media resource appears for the first time, the corresponding player component (e.g., AudioPlayer, VideoPlayer) is created; subsequent updates use its add_url() method.
• All content is added to the container in the following order: error → reasoning content → response content → video → audio → images.
• The method calls clear_widgets() to remove old components before re-rendering, ensuring visual consistency.

Registration Note: This control is registered via bricks.Factory.register('LlmOut', bricks.LlmOut), enabling configuration-driven instantiation.