fix
This commit is contained in:
@@ -27,6 +27,13 @@
|
||||
// description + feed attributes live on raw_products.mapped_data and are
|
||||
// exported/imported as-is (skip-processed must not strip them). For a
|
||||
// polluted local DB without a full wipe, use -mode backfill-categories.
|
||||
//
|
||||
// Remaining empty mapped categories (~products never in the dump): full or
|
||||
// categorize processing asks the platform LLM (OverloadedBot / OpenAI) to pick
|
||||
// a company taxonomy unique_id from Available Categories, then persists to
|
||||
// processed_products.category and empty mapped_data.category (source=llm).
|
||||
// Do not wipe the catalog — enqueue a process job for the uncategorized subset
|
||||
// (admin Products → Process, or API processing_type=categorize / full).
|
||||
// Fixture EANs are purged from non-A1 tenants automatically.
|
||||
// Processed attribute junk (poll clean, DB still has zavora/vzmetenje, …): use
|
||||
// -mode backfill-attributes — same SanitizeProductAttributes + category allowlist
|
||||
|
||||
Reference in New Issue
Block a user