Skip to main content
This shield detect completion responses by the LLM that are from a semantic perspectice mean the model refuses to answer the user prompt.