- Breaking change
- API
- Inventory
Multi-location inventory, allocation, and reservations#
Inventory has its own API. Stock is tracked per location, demand is routed across locations, and checkout holds real claims instead of decrementing a counter.
- Inventory API: inventory items and per-location levels, adjustments, safety stock, policies, transfers, counts, and merchant-owned reservations. Read levels to inspect stock; create a reservation when your integration needs to hold it.
- Locations API: create and manage locations, publish their geography, and enable them for inventory. Identity, geography, and the inventory block are versioned independently so concurrent writes cannot silently overwrite each other.
- Allocation policies: rank locations into priority groups to decide which one serves demand. Single-location merchants can skip policies entirely and route to a fixed location.
- New webhook events use the
inventory.prefix, alongsideorder.inventory_exception.createdandorder.inventory_exception.resolved. See the event catalog.
Breaking: product variants no longer carry their own stock count. quantity_available and enforce_inventory_limit have been removed from ProductVariant. A variant now declares inventory_tracking (not_tracked or tracked) and points inventory_item_id at an inventory item whose quantities live on per-location levels. Tracking is never inferred from a product being physical, so a variant you do not explicitly mark tracked sells without a stock check.
Idempotency keys are required on quantity-changing commands. Adjustments and reservation transitions reject a request with no Idempotency-Key. Flint will not generate one, because a generated key cannot protect the case that matters: the request succeeded and you never received the response. Reusing a key with identical input returns the original result and its original level projections.