1Z0-184-25 Question 123
Select 2You are building a Retrieval-Augmented Generation (RAG) application on Oracle Cloud Infrastructure (OCI) to provide real-time answers from a large set of internal documents. Your plan is to generate vector embeddings of the documents for similarity search and then pass relevant content to a generative Large Language Model. Which two strategies are recommended for ensuring secure data storage and efficient vector retrieval on OCI in this scenario?
- A
Use Oracle Cloud Infrastructure File Storage to store unencrypted embeddings as text files, then retrieve them using a basic full-text search approach.
- B
Use Oracle Autonomous JSON Database with built-in vector indexing and transparent data encryption to store embeddings, enabling rapid similarity searches on secure data.
- C
Use Oracle MySQL Database Service with traditional table partitioning for storing embeddings as strings, relying on standard B-tree indexes for retrieval.
- D
Leverage Oracle NoSQL Database, which supports vector data types and secure token-based access policies, to store embeddings and perform efficient vector searches.
Show answer and explanation
Correct answers: B, D
Explanation
A critical requirement for a Retrieval-Augmented Generation architecture is the ability to quickly retrieve documents or document fragments based on vector similarity while maintaining data security. Oracle Autonomous JSON Database and Oracle NoSQL Database both offer efficient vector indexing and built-in encryption or secure access controls, making them well-suited for RAG applications. Traditional file-based storage or standard relational indexing approaches cannot efficiently handle bulk vector queries. For additional information, refer to OCI documentation on Autonomous Database for JSON (including vector capabilities) and Oracle NoSQL Database best practices for vector searches.
- A. Incorrect.
Storing vector embeddings as unencrypted text files in OCI File Storage and relying on full-text search techniques does not provide efficient similarity search or sufficient data security. Full-text indices are not optimized for vector-based nearest-neighbor queries.
- B. Correct.
Oracle Autonomous JSON Database offers transparent data encryption, ensuring that sensitive embeddings remain secure at rest. With specialized vector indexing, it can perform efficient similarity comparisons, making it well-suited for RAG applications.
- C. Incorrect.
While Oracle MySQL Database Service can store embedding vectors as strings, basic indexing and table partitioning do not provide the specialized vector search capabilities needed for fast similarity scans in large-scale deployments.
- D. Correct.
Oracle NoSQL Database supports vector data types (as of recent releases) and token-based authentication, offering both efficient vector retrieval and robust security. This makes it a strong fit for storing embeddings in a RAG solution.