Skip to main content
Knowledge construction is the first step in building the intelligent assistant’s brain. Here, you can convert existing enterprise documents, web links, or FAQ lists into knowledge units that AI can understand.

​
1. Generate Knowledge from Documents

Suitable for scenarios with existing business documents (such as product manuals, training materials). The system will automatically parse document content and split it into Q&A pairs.

​
Operation Steps

  1. Click the “Generate Knowledge from Documents” tab.
  2. Upload File:
    • Click the upload area to select a local file, or drag and drop the file directly into the dotted box.
    • Supported formats: .pdf, .docx, .doc, .md, .txt
    • File size limit: Max 100MB
  3. Set Attributes:
    • Knowledge Ownership: Defaults to “Personal Knowledge Base”, can be switched to “Enterprise Knowledge Base” for sharing with everyone.
    • Knowledge Group: Select the corresponding business group (such as “Default Group” or custom group) for subsequent classification management.
  4. Click the “Start Parsing and Generating” button.
Parsing large files may take a few minutes. You can check the task status (such as “Parsing Successful”, “Parsing Failed”) in the “Document Generation List” at the bottom of the page. The list also shows file name, ownership, group, and operator information.

​
2. Generate Knowledge from Web Pages

Suitable for scenarios where knowledge is scattered on official websites, help centers, or online documents. The system will automatically crawl the content of the specified URL and its subpages.

​
Operation Steps

  1. Click the “Generate Knowledge from Web Pages” tab.
  2. Configure Crawling Parameters:
    • Upload Link: Enter the target web page address. Supports entering multiple URLs at the same time, please separate with commas ,.
    • Limit Quantity: Set the maximum number of pages to crawl (default 10).
    • Max Depth: Set the crawling level depth (default 1, i.e., only crawl the current page; set to 2 to crawl the current page and its directly linked subpages).
    • Keywords: Enter retrieval keywords to help the system extract relevant content more accurately.
  3. Set Attributes: Configure “Knowledge Ownership” and “Knowledge Group” as well.
  4. Click “Start Parsing and Generating”.
In the “Web Page Generation List”, you can monitor the crawling progress. If a web page fails to parse, you can click the Retry Icon (circular arrow) in the list to crawl again.

​
3. Direct Knowledge Entry

Suitable for supplementing scattered knowledge points or quickly entering specific Q&A pairs. Supports three entry modes.

​
Mode 1: Enter Single QA Knowledge

The most commonly used mode for adding standard “Question-Answer” pairs.
  1. Select the “Enter Single QA Knowledge” radio box.
  2. Fill in Content:
    • Question: Enter the question the user might ask (required).
    • Answer: Enter standard reply content (required, supports line breaks).
    • Insert Attachment: If necessary, click the button to upload relevant images or documents as supplements.
  3. After configuring ownership and group, click “Confirm Entry”.

​
Mode 2: Enter Single Paragraph Knowledge

Suitable for non-Q&A format plain text knowledge (such as company profile, policy terms).
  1. Select “Enter Single Paragraph Knowledge”.
  2. Paste or enter the complete text paragraph in the input box. If necessary, click the button to upload relevant images or documents as supplements.
  3. The system will automatically retrieve relevant information from this paragraph to answer user questions later.

​
Mode 3: Batch Entry Knowledge

Suitable for existing organized Excel Q&A tables.
  1. Select “Batch Entry Knowledge”.
  2. Download Template: Click to download the Excel template provided by the system.
  3. Edit Table: Fill in questions and answers according to the template format (one pair per row).
  4. Upload Table: Upload the edited Excel file.
All manually entered records can be managed in the “Manual Entry List” below, supporting search by name or batch deletion of erroneously entered items.