Progress Agentic RAG for OpenEdge Demo

In this demo, we showcase how Progress Agentic RAG and the OpenEdge MCP Server work together to retrieve and enrich information for MountainSports, a sample customer whose unstructured data has already been preloaded into Agentic RAG prior to the demonstration. We begin by querying this existing unstructured content and correlating the retrieved results with structured records stored in the sports2020 demo database, creating a unified and context-rich view of MountainSports’ sporting-goods catalog. When certain details are missing from both the structured and unstructured sources, the workflow automatically performs targeted internet searches to supplement those gaps with authoritative external information. This end-to-end sequence illustrates how OpenEdge and Agentic RAG deliver accurate, context-aware insights by seamlessly combining preexisting unstructured data, structured database records, and real-time internet intelligence.

Visit the Progress Agentic RAG for OpenEdge page for additional information.

Transcript:
Agentic RAG allows us to query unstructured data efficiently. For this demo, the MountainSports data is already preloaded. Let’s ask: ‘I would like to present a birthday gift to my nephew with a nice jacket. Show me what you have.’ Agentic RAG searches the dataset, finds the most relevant information, and returns the results.

In the Sports2020 database, certain tables are exposed as REST APIs, providing details on golf balls and golf clubs. Next, we’ll query this data through Agentic RAG.

We use Agentic RAG workflows to manage queries. First, it rephrases the query based on context. Then it retrieves information in layers: starting with the Knowledge Box, falling back to the OpenEdge MCP if needed, and finally, if necessary, searching the internet with Perplexity. The collected data is summarized and delivered to the application. The workflow is flexible and adapts to different needs.

Using our app, we query: ‘Show all golf clubs and golf balls with their price, brand, stock, and availability.’ The query is rephrased and sent to Agentic RAG. If the Knowledge Box does not have the information, it falls back to the OpenEdge database. Here we see the relevant structured data successfully retrieved.

Next, we ask: ‘Do we have any offers on different kinds of golf balls? Provide all details and their ratings.’ Agentic RAG retrieves the relevant unstructured data, giving us a list of golf balls with ratings and citations of where the information came from.

“Finally, we ask: ‘Are there any subscription services for MountainSports enthusiasts?’ This information isn’t in the Knowledge Box or database, so the workflow falls back to Perplexity, retrieving the answer from the internet.