Knowledge base

The knowledge base is everything you've taught Huey about your store. It's the difference between Huey saying "check our policies page" and "yes, you can return that within 30 days as long as the tags are on — here's the link to start: ..."

Manage at Librarian.

Three content types

AI documents

These are the most flexible. Each document has:

• Title — Huey uses this to pick which docs to read
• Content — markdown, no length limit (be reasonable, ~2-5 paragraphs each)
• Scope — workspace-wide, or scoped to specific products / paths
• Channels — which widget channels can use this doc (filter by storefront)

When to use AI documents over help articles

Path-scoped documents

Got a doc that only applies on /products/snowboards/*? Set its scope to path and Huey will only read it when the conversation started on a matching page.

Useful for:

• Product-line-specific sizing charts
• Seasonal promo rules
• Wholesale-only pricing notes

Help articles

Help articles are structured for visitors to read directly. They show up in the chat widget under "Browse help" and as standalone pages at your-store.com/help/<article-slug> (when configured).

Each article has:

• Title + slug
• Category (creates the navigation tree)
• Body (rich text)
• Tags

When a visitor asks something Huey thinks is covered by an article, Huey will reference the article inline with a clickable link.

Learned products

Humagician auto-imports your Shopify catalog and keeps a learned-product summary per product. This lets Huey answer "do you have anything blue under $400?" without doing a live API call every time.

The import refreshes daily. Force a sync at Librarian → Learned Products → Sync now.

You can edit individual product entries to add merchandising notes:

• "This board is best for intermediate riders, slightly soft flex"
• "Pair with the Pro Bindings — they're sized right"

Importing from another tool

If you've been using a help-desk's KB (Gorgias, Intercom, etc.), you can import the docs directly:

Librarian → Knowledge Base → Import → choose source → preview → confirm.

Auto-suggestion (Librarian agent)

The Librarian agent reads recent conversations on a schedule and proposes new KB entries when it sees gaps:

• "Three visitors this week asked about international shipping — should we add an article?"
• "This product description is missing fit notes — want me to draft one from the chat history?"

Suggestions land at Librarian → Suggestions. Approve, edit, or reject.

Configure how often Librarian runs in Librarian → Settings.

What Huey actually reads

For every conversation, Huey:

1. Pulls all workspace-scoped AI documents
2. Adds any path-scoped docs that match the current visitor's URL
3. Adds product-scoped docs for any product mentioned or visible
4. References help articles when the question matches one
5. Cross-references learned products for anything price/inventory-related

If your KB is sprawling, Huey picks what's most relevant rather than dumping everything in. You can see what got included on each conversation in the Insights → AI Usage drill-down.

Tips for a great KB

• Specific beats long. Three short documents per topic > one giant document.
• Update in chunks. Edit a few docs whenever you change a policy. Don't wait for a quarterly review.
• Use the same words shoppers use. If your customers say "the green one", your KB should mention "green" — not just "emerald".
• Test by asking Huey. Open the operator chat and ask Huey "how would you answer X?" to see what it pulls.