The power of the BeDesk AI Agent lies in its knowledge. A well-curated knowledge base is the key to providing fast, accurate, and helpful answers to your customers. The "Knowledge" section within the AI Agent settings is your central hub for building and managing the information your agent uses.
This article will guide you through the different types of knowledge sources you can add and how to manage them effectively.
Select which provider should be used for ingesting content from Admin > Settings > AI & Agents page.
There are four primary sources you can use to train your AI Agent:
Websites: Sync content directly from any publicly accessible website, like your main company site or a documentation portal.
Articles: Instantly leverage all the help center articles you've already written within BeDesk.
Documents: Upload private files (like PDFs) containing internal information you want the agent to know without making it public.
Snippets: Create custom question-and-answer pairs to handle frequently asked questions with precision.
This feature allows you to scrape and ingest content from any public website. It's perfect for syncing your blog, documentation, or product feature pages.
From the AI Agent > Knowledge page, click the + Add website button in the top right.
The "Add website" dialog will appear. Fill in the required information.
Enter the URL of the site you want to sync: Provide the starting URL for the website. Only publicly accessible URLs are supported.
Scan Options: Choose how deeply the agent should scan the site.
Scan only child pages of specified url: This will only scan pages that are "under" the URL you entered. For example, https://yourdocs.com/features/ would scan .../features/setup but not yourdocs.com/about.
Scan all pages of specified domain: This will attempt to find and scan every page on the entire domain.
Scan only specified url: This will only ingest the content from the single URL you entered.
Advanced Options (Optional): These options give you fine-grained control over what content is extracted.
Content CSS selector: Specify the main content area of your pages (e.g., main-content, article). This helps the AI focus on the important text and ignore irrelevant parts.
CSS selectors to exclude: Add a list of selectors to ignore, such as footer, .sidebar, .nav. This prevents the agent from learning from irrelevant text on the page. AI agent will already exclude some elements like navigation, scripts, footer, inline images etc.
Click Sync.
Website content will be ingested in the background using CRON jobs and requires Automated Tasks (CRON Jobs) to be setup.
After you add a website, it will appear in the Websites index page with a "Scan pending" status. The initial scan may take some time depending on the size of the site.
Once the scan is complete, the status will update. You can then manage the website and its content.
By clicking the preview icon (eye) next to any synced page, you can see a clean "Content preview" of the exact text the AI has extracted and learned from.
To manage a website, click the "Manage" dropdown on the main Knowledge page or the three-dot menu on the Websites index page. You have the following options:
Re-sync: Re-scans the entire website to update the AI's knowledge with the latest content.
View pages: Takes you to the list of all individual pages that have been synced from that website.
Delete website: Permanently removes the website and all of its synced content from the AI's knowledge base.
This is the easiest way to empower your AI Agent. It can use the help center articles you've already written in BeDesk.
To manage which articles the AI can access:
Click the options button in knowledge page articles section to enable to disable or articles for AI agent.
Click Manage articles button from knowledge page or View all articles from help center manager page in admin area to view all available articles and select which ones specifically should be available for AI agent.
Use this option to upload documents containing information that isn't available on a public website. This is ideal for internal price lists, specifications, or support policies that you want the AI to access securely.
From the AI Agent > Knowledge page, click the + Upload documents button.
The "Upload documents" dialog will open.
You can either click Select files to browse your computer or simply drag and drop files into the designated area.
Before uploading, please note the following:
Supported Formats: Only text-based formats are supported, including: PDF, Office, PowerPoint, Excel, CSV, JSON, and XML.
File Size: The maximum size for a single file is 40MB.
You can specify the language of the document from the dropdown menu. This helps the AI process the text more accurately.
Click Upload to begin processing your file(s).
After a document is successfully imported, you can manage it using the Manage button next to its name.
View content: See a preview of the text that was extracted from the document.
Delete: Permanently remove the document from the AI's knowledge base.
Snippets are small, custom pieces of text that you can add directly to the AI's knowledge base. They offer the highest level of control, allowing you to inject specific facts or pre-defined answers that the AI will prioritize.
While they are perfect for creating traditional Question-Answer pairs to handle FAQs, their use extends further. You can use snippets for specific facts, standardized responses, defining terms or correcting information.
The management options are similar to Articles. Use the more options button to enable or disable all snippets for AI agent, or enable only specific snippets for the agent from manage snippets page.
If queues are enabled, articles, documents and snippets will be ingested in the background, so you don't need to keep knowledge page open.