Configure a search interface for Relevance Generative Answering (RGA)

This is for:

Developer System Administrator

To leverage Relevance Generative Answering (RGA) in your search interface, you must enable the RGA component for the search interface.

The RGA component adds the RGA question-answering experience to your search interface. For more information, see RGA component features.

Note

To generate answers you must associate an RGA model to the query pipeline that’s used by your search interface.
Relevance Generative Answering result
Important

If your search interface includes a sorting option, RGA works best when results are sorted by relevance, which is the default sorting option. Otherwise, an answer may not be generated.

Search page sorting option | Coveo

For more information on the reasons why an answer wouldn’t be generated, see When is an answer not generated?.

Enable the RGA component

You can enable the RGA component for the following Coveo-powered search interfaces:

Hosted search page

If your search interface was created using the hosted search page builder, enable the Relevance Generative Answering option in the builder to add the component to your search interface.

Notes

  • The RGA component appears at the top of the search results page.

  • RGA isn’t supported in hosted search pages created using Coveo JavaScript Search Framework.

Hosted Insight Panel

If your search interface was created using the Insight Panel builder, enable the Relevance Generative Answering option in the builder to add the component to your Hosted Insight Panel.

Note

The RGA component appears at the top of the search results page.

In-Product Experience (IPX)

To add the RGA component to a next-gen In-Product Experience (IPX) search interface, enable the Relevance Generative Answering option in the builder.

Notes

  • The RGA component appears at the top of the search results page.

  • RGA isn’t supported in legacy IPX search interfaces.

Coveo Atomic search page

If your interface uses the Coveo Atomic library, the interface must include the atomic-generated-answer component.

Tip

Coveo recommends that you use the textarea property of the atomic-search-box component to make the search box expand to support long queries.

Coveo Headless search page

If your interface uses the Coveo Headless library, the interface must use the GeneratedAnswer controller.

Coveo Quantic search page

If your Coveo for Salesforce interface uses the Quantic library, the interface must include the QuanticGeneratedAnswer component.

Tip

Coveo recommends that you use the textarea property of the QuanticSearchBox component to make the search box expand to support long queries.

RGA component features

The RGA component adds the Relevance Generative Answering (RGA) experience to your Coveo search interface.

RGA integrates generative AI question-answering with traditional search. A single search box supports both simple and complex natural language user queries, and provides both traditional search results and a generated answer.

When a user enters a query, RGA generates the answer in real time on the search results page. If the user then applies filters to narrow down the search results, the answer regenerates on-the-fly based on the selected filters.

The RGA component adds a dedicated area to your search interface that’s used to display the generated answer and other RGA-specific user features.

Notes
Relevance Generative Answering component

1

The answer that’s generated by RGA.

2

Users can click a thumbs-up or thumbs-down icon to provide feedback on the generated answer. A feedback modal appears to provide additional details about the feedback.

3

A copy button lets users copy the generated answer to their clipboard.

4

Users can choose to show or hide the RGA component in their search interface. RGA still generates an answer for a user query even when the component is hidden. In this case, showing the RGA component reveals the generated answer.

5

A disclaimer that advises the user to verify important information in the generated answer.

6

Citations highlight the items that contain the raw data that was used to generate the answer. Users can click a citation to open the item, or hover over a citation to view the specific item chunk that was used to generate the answer.

Relevance Generative Answering citation hover
Note

An item may appear more than once in the citations if different chunks from the same item were used to generate the answer.

7

A Show more/Show less feature that lets the user expand and collapse a generated answer.

Note

This feature is disabled by default. To enable it, use the collapsible property in the Coveo Atomic Library or Coveo Quantic Library.

When the Show more/Show less feature is enabled, the RGA component collapses the generated answer and displays Show more if the answer exceeds 250 pixels in height. This is done to limit the height of the RGA component in your search interface during answer generation. If the generated answer doesn’t exceed 250 pixels, the answer displays in full and Show more doesn’t appear.

Relevance Generative Answering show more

RGA answer feedback

RGA includes an answer feedback modal that opens when a user clicks the thumbs-up or thumbs-down icon in the RGA component.

Relevance Generative Answering feedback

The feedback modal is designed to collect additional details about the generated answer for evaluation and reporting purposes. Details from the feedback modal are logged as separate Relevance Generative Answering (RGA) UA events.

You use the feedback modal information to better evaluate the generated answers, such as when testing an RGA implementation. The preconfigured Generative Answering Performance dashboard report template contains a User Feedback section that provides a high-level look at the logged feedback details. For a more detailed analysis, you can use Snowflake to analyze the question-answer pairs for your RGA implementation.

The additional feedback allows Coveo to collect qualitative information for reporting purposes, which may inform future updates to the RGA models but doesn’t directly influence current generated answers. This feedback also helps Coveo have direct visibility on user feedback, automatically monitor RGA's answer quality over time, and derive new user experiences from the data.