AI Chatbot With Knowledge-base

What It Is? #

The AI chatbot with knowledge-base is Helpmate’s core feature: a site-specific assistant that answers customer questions 24/7 using content you train it on. It reduces support load, improves satisfaction, and keeps answers consistent.

Where to Train: Knowledge Base #

All training is done in Helpmate AI → Knowledge Base. The page has tabs for different source types. You must have at least one general source (e.g. from the setup wizard’s quick-train or the Edit Website Overview / general context) before the other tabs are used. Training uses your data source credits (shown as “Trained Sources” on the page); sync with the refresh icon if numbers look stale.

Training Sources #

WP Posts #

Train the chatbot on WordPress posts and pages. The table lists posts/pages with title, type, author, date, status, and featured image. Select items and use Train (or the bulk train action) to add them to the knowledge base. Trained items appear in the Trained Data Sources section. Use See Trained Data to review what was stored, or Delete to remove a source.

Products (WooCommerce) #

If WooCommerce is installed, a Products tab appears. Use it to train the chatbot on product titles, descriptions, and metadata so it can answer product-related questions. Select products and train them the same way as posts; trained products show under Trained Data Sources.

Text #

Paste or type any text (policies, promotions, custom info). Save it as a source so the chatbot can use it to answer queries. Good for one-off or timely content that isn’t a full page or post.

URL #

Paste a public URL and save. The plugin fetches the page content and trains on it. Use this for a specific page on your site or an external page (e.g. a supplier’s guide) you want the chatbot to know about.

Q&A (FAQs) #

Add question-and-answer pairs for precise control over common questions.

Question: The exact or typical question (e.g. “How long does order processing take?”).
Answer: The reply you want the chatbot to give (e.g. “Orders are typically processed within 24 hours.”).
Show in Chat: When on, this Q&A can be shown in the chat when relevant.
Show as Quick Option: When on, this Q&A can appear as a suggested quick-reply or option so users can tap it without typing.

Save each pair; they appear in the trained list and can be edited or deleted.

File #

Upload a file; the plugin extracts text and trains on it. Supported formats:

PDF, DOC, DOCX, XLS, XLSX — uploaded to the server for text extraction (image-based PDFs are supported where the backend can process them).
CSV, TXT, JSON — read as text in the browser and sent for training.

Give the source a Title (e.g. “FAQ Document”, “Product Catalog”). Either paste text in the content field or choose a file. File size limits: 300 KB (free plan), 500 KB (Pro). After upload, the extracted or pasted content is stored as a trained source. Pro users with an OpenAI API key set in Manage API can also use Edit Website Overview to edit or delete the general website context.

Managing Trained Data #

View: In each tab, use See Trained Data (or equivalent) next to an item to open the stored content and confirm it’s correct.
Delete: Use Delete next to a trained item to remove it from the knowledge base. The chatbot will no longer use that content.
Bulk actions: Where available, select multiple items and use bulk delete or bulk train from the table actions.

Training is tied to your API key. If you change your API key, trained data is not carried over; you’ll need to retrain. Prefer upgrading your key instead of changing it if you want to keep the same knowledge base.

Tips for Good Answers #

Train top FAQs first (Q&A tab), then core policies (pages or text), then key products and high-traffic pages.
Keep one source of truth per topic to avoid conflicting answers.
Use clear titles for text and file sources so you can find and update them later.
Test the chatbot (e.g. from Test Chatbot in the admin) with real customer-style questions and add or adjust training where answers are wrong or missing.