AI-103: Developing AI Apps and Agents on AzureDemo

Groundedness and relevance are AI-assisted quality metrics in Microsoft Foundry — they require a model to judge outputs against retrieved context and the user query, so they belong to the 'AI quality (AI assisted) metrics' category (unlike 'AI quality (NLP) metrics,' which covers traditional statistical NLP metrics that don't include groundedness or relevance). Harmful content categories, such as hateful and unfair content, sexual content, violent content, and self-harm-related content, are evaluated through the 'risk and safety metrics' category. Together, these two categories provide the required groundedness, relevance, and harmful content scores for a RAG chat app evaluation.