Plan Your Deployment

To deploy this architecture, use the following high-level steps.

1. Prepare the vector store.

   Upload the inventory data by using Oracle Database 23ai. Make sure that there is a column that includes some description of the product. This is the column that will be vectorized. Using code on a compute instance (an Oracle Cloud Infrastructure Data Science notebook in this case) apply the Oracle Cloud Infrastructure Generative AI embeding model to the description column using a chunk size of 96 tokens to create the vector type column. Then, push the dataset into Oracle Database 23ai.

2. Create a compute instance and deploy functions.

   Create a compute instance to receive API calls from the front-end retail app and be able to make API calls to the various AI services. Deploy functions using Oracle Cloud Infrastructure Functions. This compute instance also serves as a central hub for any ad-hoc compute needs.

3. Create the front-end integration.

   Your compute instance is now ready to integrate with your application and to connect to the configure-price-quote (CPQ) application. This architecture uses Oracle APEX Application Development to create the web application, but you can also integrate this solution with existing chatbots or Oracle Digital Assistant.

This architecture provides several deployment options.

Select AI Deployment Option

Oracle Database 23ai provides Select AI, an NL2SQL tool built into the database.

The advantages of this deployment option are that the data remains in place and that it is well-suited for SQL development. This option, however, does not store unstructured data.

Description of the illustration select_ai_shopping_cart.png

select-ai-shopping-cart-oracle.zip

RAG Agent Deployment Option

Use the Oracle Cloud Infrastructure RAG Agent which combines the power of large language models (LLMs) and retrieval-augmented generation (RAG) with your enterprise data, letting users query diverse enterprise knowledge bases.

The service provides up-to-date information through a natural language interface and the ability to act directly on it.

This deployment option makes data sources available for the knowledge base such as pdf files, manuals, blogs, and so on. The LLM, however, is limited to the RAG Agent.

Description of the illustration ai_rag_agent_shopping_cart.png

ai-rag-agent-shopping-cart-oracle.zip

Custom Deployment Option

Use Oracle Cloud Infrastructure Data Science Quick Actions to manage every component of the data and LLM pipeline.

Data Science Quick Actions is a suite of actions that together can be used to deploy, evaluate, and fine tune foundation models in Oracle Cloud Infrastructure Data Science.

The advantages of this deployment option are the ability to fine-tune the models and to have full control of the data flow. Data Science expertise is recommended for this deployment.

Description of the illustration ai_custom_shopping_cart.png

ai-custom-shopping-cart-oracle.zip