Databricks Generative AI Engineer Associate Question 217
Select 4You are tasked with deploying an endpoint for a basic Retrieval-Augmented Generation (RAG) application using Databricks. Which steps should you follow in the correct order?
- A
Prepare a vector database to store and retrieve document embeddings.
- B
Deploy the model to a serving endpoint in Databricks.
- C
Create a pipeline to generate embeddings for documents.
- D
Set up a query interface that retrieves relevant documents and sends them to the model.
- E
Fine-tune the model on domain-specific data before deploying it.
Show answer and explanation
Correct answers: A, B, C, D
Explanation
Deploying an endpoint for a basic RAG application involves several critical steps: preparing a vector database to store document embeddings, generating those embeddings via a pipeline, deploying the model to a serving endpoint, and setting up a query interface to bridge the document retrieval and generation components. Fine-tuning the model is optional and is not required in a basic deployment scenario.
- A. Correct.
Correct. The first step in a basic RAG workflow is preparing a vector database to store and retrieve document embeddings. This provides the foundation for document retrieval.
- B. Correct.
Correct. Deploying the model to a serving endpoint is a critical step to enable the model to process input queries and generate responses.
- C. Correct.
Correct. Creating a pipeline to generate embeddings for the documents is essential for populating the vector database, as these embeddings are used for similarity searches.
- D. Correct.
Correct. Setting up a query interface allows the application to retrieve relevant documents from the vector database and pass them to the model for contextual generation.
- E. Incorrect.
Incorrect. While fine-tuning the model can improve performance, it is not a mandatory step for deploying a basic RAG application. Pretrained models can often be used directly for this purpose.