A Generative AI Engineer has been asked to build an LLM-based question-answering application. The application should take into account new documents that are frequently published. The engineer wants…

Question

Options

• AFor the application a prompt, a retriever, and an LLM are required. The retriever output is inserted
• BThe LLM needs to be frequently with the new documents in order to provide most up-to-date
• CFor the question-answering application, prompt engineering and an LLM are required to generate
• DFor the application a prompt, an agent and a fine-tuned LLM are required. The agent is used by

• A
  82% (23)
• B
  7% (2)
• C
  4% (1)
• D
  7% (2)

Explanation

Option A describes a standard RAG chain: a prompt template + a retriever (vector store) + an LLM. This is the lowest-cost, lowest-effort architecture for a QA app over frequently updated documents because: (1) new documents are simply indexed into the vector store - no model retraining required, (2) it uses an off-the-shelf retriever rather than a costly agent, (3) it requires minimal custom development. Option B (frequent fine-tuning) is expensive and operationally complex. Option C (prompt engineering alone without retrieval) can't handle new documents reliably. Option D (agent + fine-tuned LLM) is the most expensive and complex option, the opposite of the stated goals.

No community discussion yet for this question.