Back to Admin

Knowledge Base

Managing the support knowledge base — creating entries, AI enrichment, and reviewing AI-suggested content.

Overview

The system has two separate knowledge bases: Support KB (problem/solution pairs for the Support Chatbot and email support replies) and Sales KB (product and service FAQs for the Sales Chatbot and email sales replies). Both are managed from the admin panel and searched by their respective chatbots.

The Support KB is the foundation of the Support Chatbot. Entries include category, subcategory, problem, solution, and keywords. A well-maintained KB improves answers and reduces tickets. The Review Queue holds pending entries from self-learning and tech submissions for admin approval.

How the Knowledge Base Works

  1. A customer asks the chatbot a question.
  2. The system searches the KB using text matching and keyword relevance.
  3. If matching articles are found, they're included in the AI's context.
  4. The AI generates a response based on the KB content.
  5. If no KB match is found and web search is enabled, the AI may search the web and suggest the result for admin review.

Knowledge Base List

Navigate to Knowledge Base in the sidebar to see all entries. Each row shows the problem text, category badge, and subcategory. Click the expand arrow to see the full solution text, keywords, and action buttons.

Searching & Filtering

ControlTypeDescription
SearchText inputSearch by problem or solution text. Type and results update automatically (300ms delay).
CategoryDropdownAll Categories (default), Alarm, CCTV, Access Control, Intercom, General.

Adding a New Entry

Click Add New Entry to open the entry dialog.

Entry Fields

FieldRequiredDescription
ProblemYesDescribe the problem or question in clear, natural language. This is what the chatbot tries to match when a customer asks a question. Example: "Alarm panel showing low battery warning and beeping intermittently."
SolutionYesThe solution or answer. Be detailed and include step-by-step instructions where applicable. Example: "The low battery warning means the backup battery needs replacement. 1) Open the alarm panel enclosure. 2) Disconnect the old battery. 3) Connect a new 12V 7Ah sealed lead-acid battery. 4) Close the panel and enter your master code to clear the warning. If the beeping continues after 24 hours, contact support for a technician visit."
CategoryNoThe system category. Options: Alarm (default), CCTV, Access Control, Intercom, General. This helps the AI find the right article when the customer's question is about a specific system.
SubcategoryNoA more specific category within the main category. Example: "battery," "programming," "false alarms."
KeywordsNoTags that help the search engine find this entry. Add keywords one at a time — type a keyword and press Enter. Example: "low battery," "beeping," "DSC," "backup battery."

AI Assist

After entering the problem and solution, click AI Assist to have the AI automatically suggest a category and keywords based on your content. The AI analyzes the problem and solution text and fills in the category, subcategory, and keywords fields. You can modify the suggestions before saving.

Editing an Entry

Click the Edit button on any expanded entry to open the edit dialog. All fields are editable. Click Save Entry to update.

Deleting an Entry

Click the Delete button on any expanded entry. A confirmation dialog appears — click Delete to confirm. This action cannot be undone.

AI Enrichment

AI enrichment improves KB entries by generating better categories, keywords, and embeddings (vector representations used for semantic search). You can enrich entries in two ways:

  • Single entry — Click the Enrich button on an expanded entry to enrich just that entry.
  • Bulk enrichment — Click Enrich All at the top of the KB list to enrich all entries at once.

Review Queue

The Review Queue (Knowledge Base → Review Queue) shows AI-suggested KB entries that need admin approval before being published. These entries come from:

  • Self-learning — When the support chatbot searches the web for an answer, it saves the result as a pending entry for review.
  • Technician submissions — When technicians submit KB entries from the Tech Portal.

Queue Entry Fields

FieldDescription
Original QueryThe customer's original question that triggered the self-learning. Read-only.
Suggested ProblemThe AI's suggested problem description. Editable — you can refine it before approving.
Suggested SolutionThe AI's suggested solution. Editable — review and improve the answer for accuracy.
CategoryThe category for the new entry. Dropdown: Alarm, CCTV, Access Control, Intercom.
SourceIf the entry came from a web search, a link to the source URL is shown.

Approving & Rejecting

  • Approve — Click to approve the entry. You can edit the problem, solution, and category before approving. The entry is added to the published KB and becomes available to the chatbot. Both problem and solution are required to approve.
  • Reject — Click to reject the entry. It is removed from the queue and not published.

How Self-Learning Works

When the support chatbot encounters a question it can't answer from the KB, and web search is enabled, it searches the web for an answer. If it finds useful information:

  1. The AI uses the web results to answer the customer's question.
  2. The question and answer are saved as a "pending" KB entry.
  3. The entry appears in the Review Queue for admin review.
  4. An admin reviews, edits if needed, and approves or rejects the entry.
  5. Approved entries become part of the KB and are used for future conversations.

Self-learning is controlled by two settings in Settings → Support → AI Behavior: Web Search Enabled and Self-Learning Enabled.

Converting Tickets to KB

When you resolve a support ticket, you can convert it to a KB entry by clicking Add to Knowledge Base on the ticket detail page. This creates a new KB entry using the ticket subject as the problem and the resolution text as the solution. You can then edit the entry in the KB to refine the wording and add keywords.