XMLRenderers

Source

The input of a renderer is an XML string and a RenderContext. Two tags are rendered by default:

• root: the root tag
• p: the paragraph tag

For both tags, a viewer attribute can be used to specify whether the text inside the tag will be rendered or not according to RenderContext.

Here is how the viewer determines the visibility of the text:

[!WARNING] The input to the xml render should be valid xml strings. Currently, we automatically fix the following two issues:

• The xml_string is not a tree (e.g. <a></a><b></b>). We will automatically wrap the string with <root></root> if it is not wrapped.
• The xml_string is not escaped properly. We only escape &. Quotations, < and > are not escaped by us.

Currently, we support the following tags: <extra_info>Put your extra information here</extra_info> <clarification_hint>Put your clarification hint here, to avoid the LLM from confusing the output</clarification_hint> <strategy_hint>Put your strategy hint here, to guide the LLM to generate the output strategically in various social contexts</strategy_hint>