AI Optimizer

The Oracle AI Optimizer and Toolkit (the AI Optimizer) consists of an API Server (ai-optimizer-server) and an optional web-based GUI (ai-optimizer-client) component. Both the API Server and GUI can be run on bare-metal or inside containers.

Architecture Overview Architecture Overview

The following additional components, not delivered with the AI Optimizer, are also required. These can be run On-Premises or in the Cloud:

The AI Optimizer is specifically designed to run in container orchestration systems, such as Kubernetes. For more information on deploying the AI Optimizer in Kubernetes, using a Helm Chart, please review the Advanced - Microservices documentation.

API Server (AI Optimizer Server)

The workhorse of the AI Optimizer is the API Server, referred to as the AI Optimizer Server. By default, the AI Optimizer Server will start on port 8000

All the docs

The AI Optimizer Server API documentation can be accessed at http://<IP Address>:<Port>/v1/docs# of a running instance.

API Server API Server

Powered by FastAPI and Uvicorn, the AI Optimizer Server acts as an intermediary between the clients, AI Models, and the Oracle Database.

Client (AI Optimizer Client)

The provided web-based GUI client is built with Streamlit and interacts with the API Server via REST calls.

GUI GUI

BYO Client

You can develop and replace the provided client with any REST capable client.

Database

Oracle Database 23ai, including Oracle Database 23ai Free provides a persistent data store for the AI Optimizer.

Database Database

Reduced capabilities

The AI Optimizer can be used to interact with language models without having the database configured, but additional functionality such as RAG, will not be available without the database.

The 23ai database provides:

  • the Vector Store for split and embedded documents used for Retrieval Augmented Generation (RAG).
  • storage for the Testbed Q&A Test Sets and Evaluations
  • storage of AI Optimizer settings and configuration

Document Source

Access to document sources for the purpose of embedding and populating the Vector Storage is provided. Documents can be accessed from your local storage, OCI Object Storage, or from a web page.

AI Models

The AI Optimizer provides the ability to connect to any language or embedding model to be used for completions and creating vectors. Adding, Deleting, and Modifying access to AI Models is quick and easy.