Skip to main content
Retrieval-Augmented Generation (RAG) for Textual Columns makes unstructured long texts, such as product reviews or support tickets, available for AI-based summarization and generation. RAG indexes the content of specific columns so that Wisdom can answer qualitative questions about the data rather than just returning quantitative counts.

​
Configuration Steps

To configure RAG for a table, follow these steps:
  1. Navigate to Domains > Select your preferred domain > Data Sources > Tables and locate the table for which you want to configure RAG columns.
  2. Click the three vertical dots icon, then click RAG Settings.
Snippet showing the RAG Settings option selected in a table menu.
  1. In the RAG Configuration for Textual Columns window that opens, complete the following:
    • In the Text content columns for search field, select the columns that contain rich text content that will be searched over.
    • In the Filter columns for queries field, select columns with structured data (such as IDs or dates) to narrow search results.
    • In the Primary key field, choose one or more columns that uniquely identify each row in the table.
    • In the Description of search columns text box, write a brief description of what the combination of the search columns represents. This helps the AI better understand the context of your data.
    • (Optional) Click Advanced Options to specify which columns to include for citation titles and URLs in the AI’s answers.
  2. Click Save to apply your settings. A notification will appear confirming that the configuration has been updated and that the data sync has been triggered.
RAG Configuration Columns 01 Pn

​
Use Case Example: Analyzing Product Reviews

Imagine a table containing Amazon product reviews with the following columns: Review ID, Product Name, Date, and Feedback (the open-ended field that contains long user-generated texts). To enable thematic analysis, you would configure the table as follows:
Configuration FieldSelected ColumnPurpose
Text content for searchFeedbackContains the long texts that the AI will summarize.
Filters for queriesProduct Name, DateAllows Wisdom to narrow the search to specific items or timeframes.
Primary KeyReview IDUniquely identifies each row to aid in accurate data retrieval.
This setup would enable WisdomAI to answer these kinds of questions:
  • “What are the positive themes for feedback on iPhone?”
  • “What are people complaining about for the iPhone?”
  • “What themes are emerging from feedback about the iPhone?”
Filter LimitationsIf you do not include a column, such as Date, in the Filters for queries field, WisdomAI cannot effectively use it to narrow down results. For example, if Date is not selected as a filter, WisdomAI cannot answer: “What is the positive feedback for iPhone in 2026?”

Advanced Data Modeling

Define relationships and context in your data to enable more powerful analysis.

Provide Feedback to the System

Learn how to rate answers and provide corrections to improve the system’s accuracy.

Understand Domains

Manage and customize your data domains to refine context and improve query results.