Store knowledge is currently in beta.
Overview
Store knowledge is a small library of plain-language documents about your store. Layers reads these documents when it decides which tags, collections, and metafields to include in a product’s index. It also reads them when it looks into a product flagged from the How AI sees this product panel. Use it to tell Layers things a catalog alone can’t express: what a house-specific tag means, why a metafield matters to shoppers, or how to treat a naming pattern you use across products. Documents shape how Layers makes future indexing decisions. They don’t re-rank existing search results on their own.Before you start
- You need permission to view the Configure → Knowledge page. Editing, creating, and archiving documents require the knowledge-edit permission. See Manage roles and permissions.
- Knowledge documents guide AI judgement. They don’t override deterministic ranking, request transforms, or merchandising rules.
- Layers removes archived documents from the index it searches, so they stop influencing new decisions.
Open the Knowledge page
- In the Layers dashboard, open Configure.
- Select Knowledge in the sidebar under the search and discovery group.
- The page lists your active documents as cards, with any pending documents shown first.
Create a document
- Select New document in the top right.
- Fill in the fields:
- Title — A short name for the document. Up to 255 characters.
- Scope — A label that groups related documents, such as
general,apparel, ormetafields. Up to 100 characters. - Confidence (0-1) — Optional. How much weight Layers should give this document when it conflicts with another. Leave blank if you’re not sure.
- Tags — Optional comma-separated labels for filtering the list.
- Body — The document itself, in plain language. Up to 50,000 characters.
- Select Create.
Edit a document
- Find the card for the document you want to change.
- Select the pencil icon on the card.
- Update any field and select Save changes.
Archive a document
- Find the card for the document you want to remove.
- Select the trash icon on the card.
- Confirm the archive prompt.
What to write
Documents work best when they tell Layers a rule, not a goal. Keep each document focused on one topic.House-specific tags
Metafield meaning
Naming patterns
Tips
- One idea per document. Short, focused documents give the AI a clearer signal than a single long document that mixes topics.
- Name the actor. Write “keep this value,” “drop this value,” “treat this collection as internal.” Instructions to a person read cleanly to the model too.
- Reach for knowledge when a rule is store-specific. For catalog-wide behavior the search pipeline should follow, use search instructions instead.
- Use flags to build knowledge. When you flag a value in How AI sees this product, Layers may add a document that captures the fix. Review those documents on this page and edit them if the wording is off.
Troubleshooting
My document doesn’t seem to be having any effect. Knowledge shapes new indexing decisions. Layers doesn’t automatically re-evaluate products that were already indexed before you added the document. Resync the affected products from the catalog list, or wait for the next scheduled resync. See Manually resync a single product. A value I told Layers to keep is still being dropped. Open the product in the catalog and use How AI sees this product to see the ignore reason. Flagging the exact value from that panel gives Layers a direct signal to reconcile with your document. I archived a document by mistake. Create a new document with the same body. There is no in-place restore.Next steps
- How AI sees this product — Inspect what Layers indexed for a product and flag values that look wrong
- Configure search instructions — Store-wide guidance that runs on every AI step in the search pipeline