Databricks Generative AI Engineer Associate Question 222
Select 4You are tasked with deploying an endpoint for a basic Retrieval-Augmented Generation (RAG) application in Databricks. Which steps should you sequence to successfully deploy the endpoint?
- A
Prepare and pre-process the knowledge base or documents to be used for retrieval.
- B
Deploy a vector database and index the knowledge base for similarity search.
- C
Fine-tune a large language model (LLM) specifically for the RAG task.
- D
Define and deploy a serving endpoint in Databricks to handle RAG queries.
- E
Test the endpoint by issuing queries and validating the responses.
Show answer and explanation
Correct answers: A, B, D, E
Explanation
For deploying a basic RAG application endpoint, the steps include preparing the knowledge base, indexing it using a vector database, defining and deploying the serving endpoint in Databricks, and finally testing the endpoint. Fine-tuning an LLM is optional and not required for a basic deployment, as pre-trained models can often suffice for generating responses when combined with retrieved knowledge.
- A. Correct.
Correct. Preparing and pre-processing the knowledge base ensures the data is in a suitable format for retrieval and indexing.
- B. Correct.
Correct. A vector database is essential for indexing the knowledge base and performing similarity searches in a RAG application.
- C. Incorrect.
Incorrect. Fine-tuning an LLM is not mandatory for deploying a basic RAG application, as pre-trained models can often be used.
- D. Correct.
Correct. Defining and deploying a serving endpoint in Databricks is a necessary step to handle incoming RAG application queries.
- E. Correct.
Correct. Testing the endpoint ensures that it functions as expected and provides accurate responses.