RAG App

Overview

The RAG App Tool allows Agents in Dataworkz to access Retrieval-Augmented Generation (RAG) Applications. This enables the Agent to answer questions from unstructured sources such as PDFs, Word documents, or HTML files.

RAG Applications process a query and return relevant answers generated by a Large Language Model (LLM) using RAG techniques. For example:

• Query: “What products cannot be returned?”

• Answer: The tool retrieves the exact policy rules from a document (if available).

RAG Applications also support metadata filters, allowing you to partition the document corpus for more accurate retrieval.

Key Concepts

• RAG Application: A backend application that connects documents to a retrieval pipeline. Must be created before configuring a tool.

• RAG App Tool: A tool created inside an Agent to expose a specific RAG Application.

• LLM: The language model used to generate answers. Selected automatically once a RAG Application is chosen.

Note: Do not confuse RAG Application with RAG App Tool. A RAG Application must exist first. The RAG App Tool connects that application to an Agent.

Pre-requisite:

• A RAG Application must already exist in your account.

• Access to the Dataworkz Agent workspace.

Set up:

Creating a RAG App Tool

Navigate to Agents > Create Tool > RAG App Tool and complete the following sections:

1. About Tool

• Tool Name: Enter a descriptive name.

• Description: Provide a clear description of the tool’s purpose.

  • Example: “Use this tool to answer troubleshooting questions about network connectivity.”

2. Configuration

• RAG Application: Select the appropriate RAG Application from the dropdown.

• LLM: Automatically selected based on the RAG Application.

3. Input Parameters

• query (String): The default parameter that represents the user’s question.

  • The description can be updated to clarify usage.

• If metadata filters are defined in the RAG Application, they appear here as additional input parameters.

4. Output

• Default output: “The response for the question.”

• You can update the description for clarity.

5. Examples

• Add sample queries from your RAG Application.

  • Example: “List the steps to reset my account password.”

When to Use

Use the RAG App Tool when you want the Agent to:

• Retrieve answers from unstructured documents.

• Provide users with rules, policies, troubleshooting steps, or glossary definitions.

• Leverage document metadata for precise retrieval.

Validation

• Enter a query in the Test Tool panel.

• Select Test to check if the tool responds as expected.