1Z0-184-25 Question 143
Single answerYou are creating a retrieval-augmented generation (RAG) application in Python on Oracle Cloud Infrastructure (OCI). The application needs to store and retrieve vector embeddings for a large corpus of text so that a Large Language Model (LLM) can provide context-aware responses. Which approach is recommended to efficiently store these embeddings at scale and minimize retrieval latency in OCI?
- A
Store the vector embeddings as BLOB objects in OCI Object Storage and periodically download them to Python for inference.
- B
Leverage Oracle Database 23c� built-in vector search feature to store and query vector embeddings.
- C
Use the ephemeral root file system on an OCI Compute instance to store all embeddings in local text files.
- D
Deploy an in-memory open-source key-value store on a separate Compute instance to hold all vector embeddings.
Show answer and explanation
Correct answer: B
Explanation
Oracle Database 23c natively supports vector search, making it a recommended choice for retrieval-augmented generation workloads when large-scale embeddings need to be stored, indexed, and queried efficiently. Storing them in built-in specialized data structures ensures performance benefits, reduces operational overhead, and leverages Oracle Cloud Infrastructure� fully integrated capabilities. For more information, refer to Oracle documentation on Database 23c and vector search best practices.
- A. Incorrect.
Storing the embeddings as BLOBs in OCI Object Storage is convenient for long-term storage but not ideal for fast, repeated vector searches. Periodic downloads for inference increase latency and overhead.
- B. Correct.
Oracle Database 23c introduces native vector search functionality, integrating seamlessly with OCI. Storing embeddings in a database that supports vector operations significantly reduces retrieval latency and management complexity.
- C. Incorrect.
An ephemeral file system on a Compute instance is not suitable for scaled production workloads. It lacks durability and indexing capabilities for fast, vector-based lookups, risking data loss on instance termination.
- D. Incorrect.
While using an external in-memory key-value store can offer speed, it can become complex to maintain high availability, scaling, and data durability. It also doesn�t leverage Oracle� native vector search features.