Skip to main content

response_format

headless_lms_chatbot::chart_spec_generation

Function response_format 

Source 
fn response_format() -> LLMRequestResponseFormatParam
Expand description

The format the chart spec LLM is asked to answer in.

A plain JSON object rather than a named schema: Azure’s structured output accepts only a restricted subset of JSON Schema, which cannot describe a Vega-Lite specification. Handing the specification back as the whole answer means the model writes it in the shape it is written in everywhere else, and that the answer is guaranteed to parse.