andersswanson.dev

The Oracle AI Optimizer and Toolkit is a free and open-source tool designed to make it easier for developers and data engineers to build, benchmark, and optimize AI workflows running on Oracle Database.

It provides a modular framework for experimenting with model selection, prompt engineering, agents (tool calling), retrieval-augmented generation (RAG), and hybrid query optimization.

In this article, we’ll set up a local, containerized test environment for the AI Optimizer and Toolkit – enabling you to develop and test AI with Oracle Database without writing a single line of SQL.

Prerequisites:

• Docker-compatible container runtime
• docker-compose utility for managing Oracle Database Free and AI Optimizer containers.

Download and build the AI Optimizer container

First, clone or download the source code from GitHub. Then, Build the optimizer image from the root of the cloned repository:

git clone https://github.com/oracle/ai-optimizer.git
cd ai-optimizer
docker build -f src/Dockerfile -t ai-optimizer-aio .

git clone https://github.com/oracle/ai-optimizer.git
cd ai-optimizer
docker build -f src/Dockerfile -t ai-optimizer-aio .

Configure a minimal docker-compose.yml script for Oracle Database Free and the optimizer image:

services:
  ai-optimizer:
    image: ai-optimizer-aio:${TAG:-latest}
    container_name: ai-optimizer
    ports:
      - 8501:8501

  oraclefree:
    image: gvenzl/oracle-free:23.26.0-slim-faststart
    container_name: oraclefree
    ports:
      - 1521:1521
    environment:
      - ORACLE_PASSWORD=Welcome12345
    volumes:
      - ./oracle:/container-entrypoint-initdb.d
    healthcheck:
      test: ["CMD-SHELL", "lsnrctl status | grep READY"]
      interval: 15s
      timeout: 10s
      retries: 5
      start_period: 30s

services:
  ai-optimizer:
    image: ai-optimizer-aio:${TAG:-latest}
    container_name: ai-optimizer
    ports:
      - 8501:8501

  oraclefree:
    image: gvenzl/oracle-free:23.26.0-slim-faststart
    container_name: oraclefree
    ports:
      - 1521:1521
    environment:
      - ORACLE_PASSWORD=Welcome12345
    volumes:
      - ./oracle:/container-entrypoint-initdb.d
    healthcheck:
      test: ["CMD-SHELL", "lsnrctl status | grep READY"]
      interval: 15s
      timeout: 10s
      retries: 5
      start_period: 30s

Create an oracle directory and place the following grant_permissions.sql script inside it:

-- Set as appropriate for your database.
alter session set container = freepdb1;

create user testuser identified by testpwd;
grant create session to testuser;
grant unlimited tablespace to testuser;
grant connect, resource to testuser;

-- Set as appropriate for your database.
alter session set container = freepdb1;

create user testuser identified by testpwd;
grant create session to testuser;
grant unlimited tablespace to testuser;
grant connect, resource to testuser;

Then, start the containers with docker-compose:

docker-compose up -d

docker-compose up -d

Once the containers start, navigate to the AI Optimizer and Toolkit Streamlit UI in your browser at http://localhost:8501/config:

Configure the AI Optimizer with Oracle Database Free

We’ll now configure the optimizer container so it’s connected to the Oracle Database Free container.

From the optimizer config window, select the “Databases” tab and enter the following information:

• Database User: testuser
• Database Password: testpwd
• Database Connect String: oraclefree:1521/freepdb1

Then, click “Save Database”. You should see the message “Current Status: Connected” displayed below the database configuration.

The database container is now ready to use for AI development! Check out the next article, Connect LLMs and use your data, for an example of LLM configuration and RAG with private data.

References

• AI Optimizer homepage
• AI Optimizer on Oracle LiveLabs
• Source code!