All Products
Search

This Product

    :Implement conversational search based on an enterprise knowledge base in the console

    Document Center

    :Implement conversational search based on an enterprise knowledge base in the console

    Last Updated:Jun 21, 2026

    This topic describes how to upload data to the OpenSearch LLM-Based Conversational Search Edition console to implement conversational search based on an enterprise knowledge base.

    Create an OpenSearch LLM-Based Conversational Search Edition instance

    For more information, see Create an OpenSearch LLM-Based Conversational Search Edition instance.

    Configure an enterprise knowledge base

    To implement conversational search based on an enterprise knowledge base, you must upload enterprise-specific knowledge first. Find the instance that you want to manage and click Manage in the Actions column to view the instance details. In the left-side pane of the instance details page, choose Configuration Center > Data Configuration and configure an enterprise knowledge base. You can import files, web page URLs, or website URLs based on your business requirements. For more information, see Import data.

    • Import a file

      Click Import File. In the Import File panel, click Upload Local File, select a local file, and then click Upload File.

      You can upload unstructured files such as doc, docx, pdf, html, txt, and ppt, as well as structured data files like JSON and Excel (UTF-8 encoded).

    • Import web pages

      Click Web Page URL Import. In the Web Page URL Import panel, click the Web Page Import tab. Enter the URLs of the web pages that you want to upload, with each URL in a separate row, and click Import.

    • Import a website

      1. Click Web Page URL Import. In the Web Page URL Import panel, click the Website Import tab. On the Website Import tab, click Create Task. In the Create Task dialog box, enter the website URL and specify the category of the website content, and click OK. You can configure the URL Filtering, XPath Selector, and CSS Selector parameters based on your business requirements.

      2. After the operation is complete, you can view the number of crawled web pages. Then, click OK to import the website.

      Note

      The following sample website URL is used for testing: https://www.alibabacloud.com/help/zh/open-search/product-overview.

    After you import data, you can view the number of imported documents in the Data Query section on the Data Configuration page. By now, a private enterprise knowledge base is created.

    The Total Documents count (for example, 23) in the upper-left corner of the page indicates the number of imported documents.

    Test the conversational search effects

    In the left-side pane of the instance details page, click Q&A Test to test the conversational search effects of the enterprise knowledge base. Enter an enterprise-specific question in the dialog box. The model matches relevant information in the knowledge base and generates an answer accordingly.

    The test interface is divided into a conversation area on the left and a parameter configuration panel on the right. The conversation area displays user questions, model-generated answers, and orange reference links. The right-side panel contains settings for model selection (for example, opensearch-qwen), and toggles for multi-turn conversation and streaming output. It also includes tabs for Prompt parameters (such as attitude, rule, noanswer, language, and role), document retrieval parameters, and query rewriting parameters.

    Note

    You can also specify parameters for specific scenarios and desired effects. For more information, see Q&A Test.

    Summary

    The preceding information describes how to implement conversational search based on an enterprise knowledge base in the console. Subsequently, you can integrate relevant OpenSearch API operations with your business to implement conversational search based on the enterprise knowledge base. You can create different types of knowledge bases for a variety of scenarios, such as smart documents, e-commerce shopping guides, and educational Q&A.