Question 3 — AZ-104 Case Study Set
Given that the system uses agents and embeddings, which two implementation decisions best align the application clients with Paideffort's architecture? (Select two answers)
Use the Foundry SDK with the Foundry project endpoint for project and agent operations.
Use the OpenAI SDK with the Azure OpenAI endpoint for embedding operations.
Use the OpenAI SDK with the Foundry project endpoint for all embedding and agent operations.
Use the Foundry SDK exclusively for embeddings because every model operation should use the project endpoint.
Correct answers
Use the Foundry SDK with the Foundry project endpoint for project and agent operations.
Use the OpenAI SDK with the Azure OpenAI endpoint for embedding operations.
Explanations
Correct Answer:
Use the Foundry SDK with the Foundry project endpoint for project and agent operations.
Use the OpenAI SDK with the Azure OpenAI endpoint for embedding operations.
Explanation:
Paideffort separates project and agent operations from embedding requests. The Foundry SDK is appropriate for Foundry project capabilities, while the OpenAI SDK can access embeddings through the Azure OpenAI /openai/v1 endpoint. Microsoft documents these endpoint responsibilities separately and notes that embedding requests should use the OpenAI endpoint rather than the Foundry project endpoint.
Wrong Answers:
Use the OpenAI SDK with the Foundry project endpoint for all embedding and agent operations.
This conflicts with Paideffort's endpoint separation. The project endpoint provides Foundry-specific capabilities, while the embedding request is intended for the Azure OpenAI endpoint.
Use the Foundry SDK exclusively for embeddings because every model operation should use the project endpoint.
Paideffort's embedding workload requires the OpenAI SDK path. Consolidating every model-related operation behind the project endpoint would not match the required endpoint behavior.