Free Databricks-Generative-AI-Engineer-Associate Actual Exam Questions - Question 7 Discussion

D misses the point since embedding all data in the prompt isn’t scalable or updatable easily.

B/C? I get why B is popular since it lets the agent pick the right tool on the fly, which fits well for mixed query types. But C’s approach of telling the LLM exactly how to categorize queries is simpler to implement without building a full agent system. It might be more lightweight and easier to debug, especially if you want strict control over how queries get routed. Still, B feels more flexible in the long run with multiple APIs and data sources.

It’s A because a vector store with RAG handles unstructured docs well, but you’d still need separate handling for API and tables, so this isn’t fully covering all needs like B does.

Not C, because relying on the LLM to output keywords and then parsing them adds unnecessary complexity. B's approach of integrating tools in the prompt for dynamic calling seems more straightforward and scalable.

Not A, since just ingesting PDFs won’t cover API or table queries well. B sounds better as it lets the agent decide dynamically which tool to use, matching the problem’s multi-tool nature.

Probably C since it clearly separates query types without overloading the prompt.

B, it’s cleaner to let the agent handle tool selection dynamically than hardcoding logic.

Probably B makes the most sense here too. Having a system prompt that lists all available tools and lets the agent call the right one feels like the cleanest way to handle different query types without overcomplicating the prompt or relying heavily on external parsing. It’s flexible and modular, so you can easily add or update tools without reworking the whole system. Also, it avoids the downsides of trying to cram all data into the prompt like in D, which isn’t scalable for event dates or standings that change often.

Maybe C is solid since it separates the query types clearly and lets the system handle each case specifically, avoiding overload in one step. It feels more modular compared to packing everything into one prompt like D.

Actually, C could work well too because it breaks down the problem clearly—having the LLM decide the query type first means the system can handle each type more efficiently with targeted logic. It’s kind of like delegating tasks upfront rather than mixing everything together or relying heavily on one architecture. Plus, it avoids overloading the prompt with all info at once like in D. So, C offers a neat way to separate concern and streamline query resolution without the complexity of multiple API calls bundled in an agent.

Guessing B makes the most sense here since it sounds like the agent can call different tools as needed, which fits the multiple query types. Seems cleaner than stuffing everything into the prompt.