Knowledge

Page AI Agent → Train → Content → Sources in the dashboard (app.thehumind.com).

This is where you feed your AI agent. Everything you add becomes context the AI can pull from when answering visitors. Four content types, organizable into nested folders.

The four knowledge types

Type	Visibility	Best for
Article	Public	Long-form content published in your help center. Cited as a source in the AI's answers.
Snippet	Private (AI only)	FAQs, hints to the AI, sensitive info. Never shown to visitors.
Web page	Public	Sync an external URL (e.g. your existing help center). Auto re-synced daily.
Document	Public	PDF file (manual, brochure, terms).

On the Sources page, click "Add content" in the top right. A dialog presents two groups: Create new knowledge (Article, Snippet) and Import or sync (Web page, Document).

Articles: public, cited as a source

An article is long-form content the AI can both use to answer and cite as a source to the user. In the chat, the AI will display a link to the article when relying on it; the user can click through to read the full version.

Fields:

• Title (required)
• Excerpt (optional): short summary. AI auto-generation available.
• Content (required): rich-text editor (headings, lists, bold, links, images, dividers).

Statuses:

• Draft: saved but not exposed to the AI. Only the title is required.
• Published: the AI can use and cite it. Title and content required.

Worth knowing:

• You can regenerate the excerpt with AI or reformat the HTML with AI from the article's side panel.
• The editor shows a view counter and last-edit timestamp.

When to use an article

When the information is useful to the AI and worth showing publicly to the end user: policies, guides, deep-dive articles, detailed FAQs you also want accessible from your help center.

Snippets: private, hints to the AI

A snippet is a title + content pair that acts as an internal hint to the AI. Never cited, never shown to the visitor. (In French the UI calls them Fragments.)

Fields:

• Title (required): used to identify the snippet in your list, and gives the AI a hint about its context.
• Content (required): the actual information or directive.

Worth knowing:

• Always active (no draft/published states).
• Ideal for: short FAQs ("What is your return policy?"), internal nuances ("Our brand is positioned as premium, never suggest low-cost alternatives"), confidential information the AI must know but never reveal as a citation.
• Scope it (optional): by default a snippet is general, retrieved across your store. You can instead attach it to specific products or whole collections so it applies only there. See Shared snippets.

When to use a snippet

When you want the AI to have the information but not show it as an official source. If you want it public, use an article instead.

Snippets vs. Instruction rules

This is the most important distinction in the AI Agent section, and the easiest to confuse:

	Snippets (this page)	Instruction rules (Instructions page)
Where	Knowledge → Sources	Instructions → Personalized instructions
Activation	Retrieved when relevant to the question	Always active in every reply
Typical use	FAQ, contextual info	Non-negotiable behavior rules
Example	"Return policy: 30 days, free returns over €50"	"Never name competitors directly"

In short: a snippet is information the AI can pull when needed. An instruction rule is a directive the AI must follow on every answer, regardless of the question. Don't put a FAQ in instructions: you'll saturate the AI's context with content that's only relevant in 1% of conversations. And don't put behavior rules in snippets: they may not be retrieved when they should.

Web pages: external sync

Import the content of a public web page. Humind crawls the URL you provide and all the nested pages under the same domain.

Fields:

• URL(s): one or more URLs. "Add Another URL" button to add more.
• Active period (optional): start date and/or end date.

In-form guidance:

"For the best results, enter the top-level domain instead of a specific page (e.g. https://help.yourcompany.com)."

Auto re-sync:

• Every web page is re-crawled automatically every day at midnight (server time).
• The "Last sync" field in the web page's side panel shows the date of the most recent crawl.
• A "Force Sync" button is available to trigger an immediate re-sync.

Web page statuses:

• queued: waiting to be processed
• processing: being crawled
• published: synced and available to the AI
• failed: sync failed (the tooltip shows the error detail)

Documents: PDF

Import one or more PDF files (drag-and-drop zone or file picker).

PDF only

Although the upload zone appears to accept other formats, only PDF files are actually processed. Other formats are rejected with an error.

Worth knowing:

• Multi-upload supported.
• Optional active period (active_from / active_until), same as web pages.
• On partial upload (some PDFs accepted, others rejected), a notification details successes and failures.

Folders and subfolders

You can organize knowledge in a folder tree, with no depth limit.

Create a folder:

• "New folder" button at the top of the list, or via the context menu.
• Fields: name (required) and icon (emoji, optional).
• If you click "New folder" while inside a folder, the new one is created as a subfolder of that one.

Move an item:

• Drag and drop onto any folder in the list, or onto any breadcrumb segment (including the root).
• Multi-select: tick the header checkbox or individual rows, then drag-drop in bulk.

Rename or delete a folder:

• Right-click or three-dot menu on the folder row.
• Options: Rename and Delete. To move a folder, use drag-and-drop.

Search, filters, scheduling

• Search: bar at the top of the list. Searches titles, returns up to 10 results.
• Tag filters: funnel icon. Tags are auto-extracted by the AI and listed in the filter menu.
• Active scheduling: every knowledge item (article, snippet, web page, document) can have a start date and/or end date. The AI uses the item only within that window. The list shows a "Scheduled" badge (orange) before the start date and "Expired" after the end date.

Shopify auto-sync

If your account is connected to Shopify (integration installed via the Shopify App Store), Humind automatically syncs two content types into your knowledge:

• Shopify Pages (the static pages you create in Shopify: "About", "Contact", "FAQ", etc.)
• Shopify Blog articles (every article from every blog you have)

How it works:

1. Right after the Shopify app installation, a first sync runs automatically.
2. A root folder "Shopify" is created in your knowledge, with two subfolders "Pages" and "Blog".
3. All pages and articles are imported as web pages (webpage type) into those folders.
4. An hourly sync then keeps everything up to date, no webhooks to configure, all handled by Humind.
5. You can move synced items to other folders; they won't be moved back on subsequent syncs.
6. You can rename or delete the "Shopify" folder; the next sync will recreate it if needed.

Manual Shopify sync

At the top of the Sources page, a "Sync Shopify" button appears when your Shopify integration is active. It triggers an immediate sync without waiting for the next hourly window. If a sync is already running, you'll see the message "A Shopify sync is already in progress.".

Shopify product pages

Shopify product pages are not synced as knowledge: your products are managed separately under AI Agent → Train → Content → Catalog. To add product-specific knowledge (care instructions, sizing tips...), use Product snippets.

Public vs. private: recap

Type	Visible to AI	Cited in answers	Publicly visible
Published article	✅	✅	✅ (help center)
Draft article	❌	❌	❌
Snippet	✅	❌	❌
Web page	✅	✅ (link to source)	✅ (its own URL)
PDF document	✅	✅	✅