1Z0-184-25 Question 108
Select 2You are designing a recommendation engine for your e-commerce platform, which requires advanced semantic search capabilities on product descriptions. You want to generate vector embeddings directly inside your Oracle Database environment using Oracle Cloud Infrastructure 2025 capabilities. Which two approaches correctly illustrate how to accomplish this?
- A
Invoke Oracle Machine Learning for Python (OML4Py) scripts stored in the database to compute embeddings for textual data, then store them in a dedicated vector data type column.
- B
Use a PL/SQL routine to call an external REST endpoint in Oracle Cloud Infrastructure Data Science that returns embeddings, storing the returned vectors in an Oracle Database vector data type.
- C
Create a standard Oracle Text index on the product descriptions, which automatically generates vector embeddings without any additional configuration.
- D
Export product descriptions to an external Python service for embedding calculations and import the results back into an Oracle Database RAW column.
Show answer and explanation
Correct answers: A, B
Explanation
In Oracle Database 23c and later, you can store and query vector embeddings natively using new vector data types. For in-database embedding generation, Oracle Machine Learning for Python (OML4Py) allows you to train or run inference directly in the database engine, avoiding data movement. Alternatively, you can call external OCI Data Science endpoints via PL/SQL to generate embeddings and store them in the database. Both methods enable seamless integration with Oracle Cloud Infrastructure while efficiently handling vector-based semantic queries. Refer to the Oracle Database and Oracle Machine Learning documentation for details on vector data types, indexing, and best practices for embedding-based search.
- A. Correct.
CORRECT. Oracle Machine Learning for Python allows in-database execution of Python-based machine learning tasks, including generating vector embeddings. Storing embeddings in a vector data type column ensures quick similarity queries. This is a widely recommended approach for keeping compute and data co-located.
- B. Correct.
CORRECT. Using PL/SQL to call an external REST endpoint in OCI Data Science is a valid approach if you want to leverage a powerful model housed in OCI. The returned vectors can be persisted in an Oracle Database vector data type, enabling efficient embedding-based search capabilities inside the database.
- C. Incorrect.
INCORRECT. While Oracle Text supports indexing text for full-text search, it does not automatically generate vector embeddings by default. Additional steps or integrations are required for vector-based semantics.
- D. Incorrect.
INCORRECT. Although exporting data to an external process and importing vectors back could work, it does not constitute generating embeddings 'inside' the Oracle Database. This approach presents additional overhead and complexity compared to in-database machine learning or direct REST calls from PL/SQL.