Share via

Add a Dataverse knowledge source

Integrating Dataverse tables as your knowledge source allows you to ground your agent in the data contained in your tables. This process involves adding synonyms and glossary definitions of the tables and columns in your data. For more information, see Improve agent responses from Microsoft Dataverse.

To add Dataverse tables as a knowledge source, perform the following steps:

  1. Open the agent.

  2. Select Add knowledge from either the Overview or Knowledge pages.

  3. Select Dataverse (preview).

  4. Locate one or more of your Dataverse tables to add. Up to 15 Dataverse tables can be added per knowledge source. To narrow your selections, use the search field.

    Note

    Table recommendations are based on the name of your agent.

  5. Preview the tables to ensure the appropriate tables were added. The preview only displays 20 rows and a set of columns, however, all the rows and columns are included in the knowledge source.

  6. Review the knowledge name and description. The description should be as detailed as possible, especially if generative AI is enabled, as the description aids generative orchestration.

  7. Optionally, to help improve the quality of the answers, add synonyms, and glossary terms:

    1. Add synonyms for table columns that you selected. Select the Back button to accept changes.

    2. Add glossary terms to define ___domain-specific terminology and acronyms. Select the Back button to accept changes.

  8. Select Add to finish adding the knowledge source.

Synonyms and Glossary terms

Synonyms, glossary terms, and definitions for the synonyms and glossary entries aid in AI orchestration. They provide grounding data to improve generated responses. By providing extra information for the AI to understand and interpret the information in the tables, you increase the likelihood of the AI to recognize your users requests, and return responses based on the information provided to the AI.

For scenarios where your Dataverse table contains a column composed of numeric values, you need to provide a synonym for the AI to understand what's in the column. For example, your agent is providing travel assistance, and the Dataverse table contains a column named "cr_123_abc" that uses flight numbers to correspond to cities.

Screenshot of the Knowledge page, displaying a preview of the Dataverse table, highlighting the column's Synonyms and Descriptions.

Since the AI doesn't know how to qualify this information, it must be explicitly told how to interpret it. So, the maker adds a description for this column, such as the following example: "cr_123_abc represents the departure city for each flight represented by the flight code."

Sample glossary definitions

Glossary definitions are used to paraphrase the terminology in your Dataverse table, so your agent better understands user questions and can respond better.

The following table illustrates scenarios where adding definitions for glossary terms provides useful context for your agent.

Scenario Glossary term Sample description
Acronym VP "VP" refers to the Vice President value in the "JobTitle" column of the "Contact" table.
Custom ownership activity owner The "activity owner" is identified by the "PartyId" column in the "ActivityParty" table.
Custom field opportunity revenue "Opportunity revenue" refers to the "Custom Revenue" column in the "Opportunity" table.
Complex rules or filter overdue task "Overdue task" refers to the the "task" table, when the "state code" column has an open value, and the "scheduled end date" column has a value that is earlier than today.

Note

  • The descriptions in the table are examples. Test your descriptions to verify what descriptions return the best results.
  • It might take up to 15 minutes for updated glossary terms and definitions to become available.

Enable Search Support for Multiline Text and File Data Types in Dataverse Tables (Preview)

With Dataverse added as a knowledge source, you can apply unstructured reasoning to get higher-quality responses from Multiline Text(MemoType) and File(FileType) columns.

Prerequisites

Note

These steps are based on the prerequisites to perform a search on Dataverse data. Learn more in Configure Dataverse search for your environment.

  • Enable the Dataverse search capability.

  • You must have maker or admin access to modify views in the Power Apps maker portal.

Important

  • There are extra Dataverse capacity costs incurred with the creation of indexing for search. Learn more in How much will Dataverse search cost?.
  • Multiline and file type attachment support is a preview feature.

Configure the Dataverse table in Power Apps

For this preview feature, you need to explicitly include the table and Multiline Text and File columns as Searchable in the Quick Find View. For more information and detailed steps for the configuration, go to Select searchable fields and filters for each table.

  1. Sign into Power Apps and select the environment you want.

  2. Select Dataverse then Tables.

  3. Select the Dataverse table you added to the Copilot Studio agent.

  4. Turn on Searchable for each column you want to search over.

  5. In the Data experiences pane, select Views.

  6. From the list of views, select the Quick Find View type.

  7. Select a searchable column from the list to add to the view.

  8. Select Edit find tables columns to add searchable columns from the Find by options.

  9. Select Save and Publish to publish the changes to the view.

Known limitations

  • If you add a Dataverse knowledge source before configuring your Multiline Text and File columns in Power Apps, it may take up to two days for the system to backfill the request. To expedite, consider readding the Dataverse knowledge after configuring the field for search.
  • Tables, images, and text in non-organization-based languages are not supported in File attachments