If you’ve been working with Oracle AI Database 26ai, you already know it’s more than just a traditional database. It’s also a powerful vector database that unlocks new possibilities for organizations aiming to harness AI. You can now generate, store, and run similarity searches on vectors right alongside your business data.

Yet for organizations in highly regulated industries, these opportunities come with unique challenges. Regulatory requirements often mandate that all data processing—including vector generation—remain fully auditable and secure, which typically requires generation within the database itself to meet compliance requirements.  However, while this approach upholds regulation standards, it can also put a significant load on your database system, as generating embeddings is resource-intensive. And no one wants to impact critical database operations.

So, what should you do if your organization faces this challenge? Is it possible to “have your AI cake and eat it too?”

Ideally, you need a secure, offline environment where you can offload vector generation—without accessing the internet and without overloading your database.

And that’s exactly why we’re excited to introduce the Oracle Private AI Services Container.

What is Oracle Private AI Services Container?

With this new container, you can securely generate vector embeddings outside the database and still store them directly in Oracle AI Database 26ai. Everything works through a simple, OpenAI-compatible REST API, and—this is important—you can do it all without any internet access.

Use Your Choice of Embedding models

The container lets you use the same models you would use inside the database, making it seamless to use embeddings generated in the container alongside those generated in the database for similarity search operations. Out of the box, the container includes several text and image embedding models, such as:

  • all-mpnet-base-v2
  • all-MiniLM-L12-v2
  • multilingual-e5-base
  • multilingual-e5-large
  • clip-vit-base-patch32-txt
  • clip-vit-base-patch32-img

If you want even more flexibility, you can run any popular embedding model (e.g., Sentence Transformers) by converting them into the ONNX format using the Oracle ONNX Pipeline via the Oracle Machine Learning for Python Client 2.

Deploy and Scale Your Way

With the container, you don’t need special hardware. A Linux host running Oracle Linux 8, 9, or 10 is all it takes. Run it on your laptop, in your data center, or on cloud compute nodes. It’s lightweight and flexible.

Need scalability? No problem.

Spin up multiple containers on the same server using Docker or Podman. Add NGINX for load balancing. Or, if you’re operating at a larger scale, orchestrate everything with the Oracle Database Kubernetes Operator.

Get Started

The bottom line: The Oracle Private AI Services Container gives you the perfect balance—secure, offline vector generation without overloading your database while helping meet your organization’s compliance requirements.

You can download the Oracle Private AI Services Container directly from the Oracle Container Registry and get all of the setup details in the Oracle AI Vector Search User’s Guide.

Try it today!