1Z0-184-25 Question 147
Single answerYou are developing a Python-based Retrieval-Augmented Generation (RAG) application on Oracle Cloud Infrastructure. Your goal is to store and retrieve embeddings for a large corpus of documents while maintaining low-latency lookups and automatic scaling. How should you store your vector embeddings to optimize performance and simplify management in this scenario?
- A
Store the embeddings as JSON files in Oracle Cloud Infrastructure Object Storage, accessing them via presigned URLs on every query
- B
Use Oracle Database 23c on Autonomous Database, enabling built-in vector search capabilities and scaling automatically based on demand
- C
Deploy a local SQLite database within each compute instance to store embeddings, relying on block storage volume replication for durability
- D
Embed the vector data in custom log messages, relying on Oracle Cloud Infrastructure Logging for retrieval
Show answer and explanation
Correct answer: B
Explanation
To build a robust RAG solution on Oracle Cloud Infrastructure, storing embeddings in a system with native vector query capabilities and automatic scaling is crucial. Oracle Database 23c running on Autonomous Database provides high-performance vector queries, simplified management (including security, patching, and scaling), and straightforward integration with Python. This design aligns with Oracle� recommended best practices for developing scalable AI-driven applications in OCI. Refer to Oracle� Database and Autonomous Database documentation for more details on vector search features and implementation guidelines.
- A. Incorrect.
Option 1 is incorrect because storing embeddings as JSON files in Object Storage can introduce significant latency, especially when querying large volumes of documents. While Object Storage is cost-effective for archived data, it� not generally suited for low-latency, real-time embedding lookups in a RAG workflow.
- B. Correct.
Option 2 is correct. Oracle Database 23c includes built-in vector search capabilities, and using it on Autonomous Database provides autoscaling, managed security, and integrated support for advanced data types. This combination ensures both high performance and minimal administrative overhead when managing vector data in a Python-based RAG application.
- C. Incorrect.
Option 3 is incorrect because deploying a local SQLite database on each compute instance leads to data duplication, higher overhead for synchronization, and limited horizontal scalability. This approach is neither cost-effective nor practical for handling large-scale RAG workloads.
- D. Incorrect.
Option 4 is incorrect because Oracle Cloud Infrastructure Logging is not designed to function as a high-performance key-value or vector store. Logging is intended for audit trails and diagnostic data, not for real-time embedding retrieval in RAG applications.