Use the Relevance Generative Answering component

This is for:

Developer

In this article

Implement the RGA component
Use an expandable search box

The atomic-generated-answer component lets you use Coveo’s Relevance Generative Answering (RGA) feature to automatically generate answers to users' queries.

You must create your RGA model and associate it with a query pipeline before you can use the atomic-generated-answer component.

The generated answer panel includes the following features:

Example generated answer panel | Coveo RGA

The title of the RGA panel.

Clicking Dislike brings up a modal which asks the user to specify why they didn’t find the information helpful.

A Copy to clipboard button.

A toggle to hide or show the RGA panel.

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

Citations highlight the items that contain the text segments (chunks) that were used to generate the answer. Hover over a citation to view the chunk, or click a citation to open the item. An HTML item opens at the exact location of the chunk, which appears highlighted.

Notes

Set disableCitationAnchoring to true if you want items to open without pointing to the chunk location.
An item may appear more than once in the citations if different chunks from the same item were used to generate the answer.

Implement the RGA component

In the search page HTML, add the atomic-generated-answer component to the main layout section. You’ll probably want to place it before the status layout section, if present.

  <body>
    <atomic-search-interface id="search" search-hub="<YOUR_SEARCH_HUB>"> 
      <atomic-search-layout>
        <atomic-layout-section section="search">
          <atomic-search-box></atomic-search-box>
        </atomic-layout-section>
        <atomic-layout-section section="facets">
          <atomic-facet-manager>
            <atomic-facet
              field="objecttype"
              label="Type"
            ></atomic-facet>
            <atomic-facet
              field="filetype"
              label="File Type"
            ></atomic-facet>
          </atomic-facet-manager>
        </atomic-layout-section>
        <atomic-layout-section section="main">
          <atomic-generated-answer></atomic-generated-answer>
          <atomic-layout-section section="results">
            <atomic-result-list></atomic-result-list>
          </atomic-layout-section>
          <atomic-layout-section section="pagination">
            <atomic-load-more-results></atomic-load-more-results>
          </atomic-layout-section>
        </atomic-layout-section>
      </atomic-search-layout>
    </atomic-search-interface>
  </body>

For the Atomic RGA component to work, your search interface must use the pipeline with which the RGA model is associated.

You can do one of the following:

Set the search hub on the atomic-search-interface component, as shown in the example. Then you need to set up condition-based routing.

This is the recommended approach.
Omit the search-hub property and use search interface-enforced routing by setting the pipeline property.

This example produces the following search page:

Example search page with generated answer | Coveo RGA

If your search interface includes a sorting option, Relevance Generative Answering (RGA) works best when results are sorted by relevance, which is the default sorting option. If the results aren’t sorted by relevance, an answer may not be generated.

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

After you implement generative answering, your users may begin to use longer queries. Enable the textarea property on your atomic-search-box component so that it can expand to support multi-line queries.