1Z0-184-25 Question 66
Single answerYour company is building a recommendation engine on Oracle Cloud Infrastructure (OCI) to help customers find similar items based on text descriptions and embedded product metadata. You have decided to store embedding vectors in an OCI-managed database and perform similarity searches to identify items that are the most closely related to a given query vector. Which approach is the most appropriate to implement a scalable, low-latency similarity search solution using OCI services?
- A
Load all vectors into OCI Object Storage and iterate through each object to compute similarities in client-side code
- B
Use an external, unsupported open source search tool on a self-managed Compute instance to store and query embeddings
- C
Leverage Oracle Database 23c� vector search feature by keeping embeddings in a vector column, creating a specialized index, and using similarity queries
- D
Build a traditional B-tree index on the embedding columns in Autonomous Database and run typical range queries for finding overlapping data
Show answer and explanation
Correct answer: C
Explanation
Oracle Database 23c introduces native vector columns and specialized indexing to support similarity searches on embedded data, such as product descriptions, images, or other high-dimensional features. By indexing vector columns, the database can efficiently process nearest neighbor queries at scale. For more information, consult the Oracle Database 23c documentation on vector data types and similarity search best practices.
- A. Incorrect.
Incorrect. Storing raw vectors in Object Storage and looping through them in application code can be prohibitively slow and costly at scale. This approach does not leverage a native similarity search index and would require massive egress and computation on the client side.
- B. Incorrect.
Incorrect. While open source tools exist for similarity search, deploying and managing them outside of OCI� managed services increases operational overhead and may not integrate well with other OCI services. The question specifically asks for an OCI-based, scalable solution.
- C. Correct.
Correct. Oracle Database 23c supports vector search by storing embeddings in a dedicated vector column and using specialized indexing for efficient similarity queries. This approach is designed for low-latency, scalable search within OCI-managed databases.
- D. Incorrect.
Incorrect. A B-tree index is not suitable for high-dimensional vector similarity searches. This type of index is designed for range queries on scalar data, not for nearest neighbor computations on embeddings.