1Z0-184-25 Question 145
Single answerYou are building a Python-based Retrieval-Augmented Generation (RAG) application on Oracle Cloud Infrastructure (OCI). Your domain knowledge is stored in an Oracle Autonomous Database. The application must retrieve relevant data from the Autonomous Database at runtime, then feed that data as context to a Python-based Large Language Model (LLM) inference process. The solution should minimize cost, handle unpredictable spikes in usage, and protect database credentials. Which approach best satisfies these requirements?
- A
Deploy a persistent VM on OCI Compute, store the Autonomous Database credentials in a plain-text configuration file, and run your Python application continuously on the VM.
- B
Implement a serverless Oracle Function invoking a Python script that uses resource principals to authenticate with the Autonomous Database and retrieve context. The function then runs inference using an LLM library packaged with the function code.
- C
Use OCI Data Integration to synchronize data nightly from the Autonomous Database to Oracle Object Storage, and run your Python-based LLM inference on a local environment referencing the Object Storage data files.
- D
Create a custom Docker image with the Python LLM library and run it in OCI Container Engine for Kubernetes without configuring any secrets, leaving the database connection open at all times.
Show answer and explanation
Correct answer: B
Explanation
Serverless architectures, such as Oracle Functions with resource principals, allow you to securely invoke Autonomous Database queries on-demand and scale automatically to handle usage spikes. A RAG application typically requires dynamic retrieval of relevant content at inference time, making a stateless, serverless function ideal for cost efficiency, security, and elasticity. For more details, refer to Oracle Functions documentation and best practices around Resource Principals and Autonomous Database connectivity in the official OCI documentation.
- A. Incorrect.
Option 1 is incorrect. Running a dedicated VM 24/7 can be costly when user traffic is unpredictable, and storing credentials in plain text is a security risk.
- B. Correct.
Option 2 is correct. Oracle Functions offers a serverless model to scale automatically with usage. Resource principals eliminate the need for hard-coding or storing credentials, providing secure, on-demand access to the Autonomous Database when retrieving context. You can package Python libraries for LLM inference within the function deployment.
- C. Incorrect.
Option 3 is incorrect. Although data synchronization to Object Storage can work for some batch processes, it misses real-time retrieval for a RAG pipeline and still leaves you with potential local environment security issues and limited scalability.
- D. Incorrect.
Option 4 is incorrect. While a custom Docker image can be used for Kubernetes, neglecting proper secret management for Autonomous Database credentials and keeping the connection open at all times increases security and cost concerns.