Microsoft Foundry Generative AI Chat App Practice Question 8: SDK Choice Embeddings
Microsoft Applied Skills: Develop a Generative AI Chat App Using the Microsoft Foundry SDK · Test 1 · question 8 of 10
A chat app already uses the Foundry SDK for project configuration and tracing. A new feature must generate text embeddings for a semantic search index, and the team wants the lowest possible call latency and the widest OpenAI API compatibility for that specific code path. Which SDK and endpoint should the embedding code use?
Show answer and explanation
Answer:
The SDK selection guidance is explicit: use the OpenAI SDK "when maximum OpenAI compatibility or lowest latency is required, when generating embeddings, or when using Models sold by Azure through Chat Completions." Its endpoint is https://<resource-name>.openai.azure.com/openai/v1. Mixing SDKs in one app is normal; the Foundry SDK stays in place for project-level work while this path takes the direct route.
- B is wrong because the Agent Framework is for "unified multi-agent orchestration for hosted agents and multi-agent systems." It adds an orchestration layer over the Responses API rather than reducing latency on an embedding call. - C is wrong because the Anthropic SDK exists for "Anthropic Claude models deployed in Foundry." It is a model-family choice, not a route to embeddings or lower latency. - D is wrong because the project client is documented for "Foundry-native operations where OpenAI has no equivalent": listing connections, project properties, tracing. Going through it adds a hop the embedding path does not need.
Reference: Get started with Microsoft Foundry SDKs and endpoints