Getting Started with Knowledge Bases

Use a knowledge base to create a built-in RAG (Retrieval Augmented Generation) engine to enhance the capabilities of agentic AI.

 

Fulfill Your Business Goals

Goal Details How knowledge bases help

Give agentic AI access to corporate documents

To do their jobs effectively, AI agents need access to corporate documents. RAG engines are the most effective way to collect this information. However, creating your own RAG infrastructure is complex and difficult.

If you have a RAG engine already, the AI agents in Oracle Integration can work with it. However, if you don't have a RAG engine, not to worry. You can create one easily in Oracle Integration using a knowledge base.

A knowledge base offers a simplified experience with a RAG engine, so you can get started quickly without having to learn all the ins and outs of a RAG engine, including how to chunk the data.

Control the versions of documents that agentic AI uses

Business requirements change, and so do corporate documents. You need a way to keep your knowledge base up to date with the latest documentation so that your AI agents can work effectively.

A knowledge base helps you easily identify the versions of each uploaded documents, so you can update and delete documents as needed.

 

Knowledge Base Essentials

   

Definition

A knowledge base provides agentic AI with a built-in RAG (Retrieval-Augmented Generation) engine, which is often a document database. Create a knowledge base if your organization doesn't yet have a RAG framework and you need to provide agentic AI with a document database. You can create a knowledge base to work with agentic AI that you build in Oracle Integration or elsewhere.

A knowledge base offers an easy way to create a RAG framework without having to understand all of its complex details, such as embeddings and chunking. Knowledge bases have robust and configurable back-ends, along with default values that help you make decisions quickly and start uploading your documents through a drag-and-drop interface.

If your organization already has a RAG engine, such as a vector database into which you've ingested documents, agentic AI in Oracle Integration can connect to it and search against it. For example, you might have uploaded all your expense policies to the database. In such cases, you probably don't need to create a knowledge base.

Knowledge base

Usage

Working with a knowledge base involves two phases:

  • Ingestion: Add your documents to the vector database. Oracle Integration provides a predefined integration for ingestion, so you can easily drag and drop documents into your knowledge base. Most people convert PDFs to Markdown for this task. Oracle handles the chunking for you. After uploading the documents, run test queries to confirm success.

  • Retrieval: Ask an agent a question, and allow the agent to retrieve an answer from the knowledge base. You can even specify the number of results to receive. The output includes the search results, a matching score, the document version, and a link to the source document.

AI-powered capabilities

You can easily create tools for agentic AI using the recipes provided by Oracle Integration.

To explore all AI features, see AI Innovation and Oracle Integration in Using Integrations in Oracle Integration 3.

Availability

Knowledge bases are available in select Oracle Integration editions. See Oracle Integration Editions in Provisioning and Administering Oracle Integration 3.

Additionally, you must configure OpenSearch to use knowledge bases.