Refactor vendor mappings to canonical PN catalog

bible-local/history.md

@@ -168,13 +168,12 @@ Reworked partnumber book storage so `qt_partnumber_book_items` is a deduplicated
 - `qt_partnumber_book_items` now stores one row per `partnumber` with fields:
   - `partnumber`
   - `lots_json` (`[{lot_name, qty}, ...]`)
-  - `is_primary_pn`
   - `description`
 - `qt_partnumber_books` now stores `partnumbers_json`, the sorted list of PN values included in that book.
 - QuoteForge read contract is now:
   - read active book from `qt_partnumber_books`
   - parse `partnumbers_json`
-  - load PN payloads via `SELECT partnumber, lots_json, is_primary_pn, description FROM qt_partnumber_book_items WHERE partnumber IN (...)`
+  - load PN payloads via `SELECT partnumber, lots_json, description FROM qt_partnumber_book_items WHERE partnumber IN (...)`
 - `PartnumberBookService.CreateSnapshot` now:
   - builds one logical item per PN,
   - serializes bundle composition into `lots_json`,
@@ -209,6 +208,106 @@ Reworked partnumber book storage so `qt_partnumber_book_items` is a deduplicated
 - Tests: `internal/services/partnumber_book_test.go`
 - Docs: `bible-local/vendor-mapping.md`
 
+## 2026-03-07: Remove `is_primary_pn` From Partnumber Books
+
+### Decision
+
+Removed `is_primary_pn` from PriceForge partnumber book storage and from the PriceForge → QuoteForge sync contract.
+
+### What changed
+
+- Added migration `031_drop_is_primary_pn.sql` dropping `is_primary_pn` from:
+  - `lot_partnumbers`
+  - `qt_partnumber_book_items`
+- Removed `IsPrimaryPN` from `internal/models/lot.go`.
+- `PartnumberBookService` no longer copies, compares, or upserts `is_primary_pn`.
+- `VendorMappingService` no longer writes `is_primary_pn` into `qt_partnumber_book_items`.
+- The sync contract for QuoteForge is now:
+  `SELECT partnumber, lots_json, description FROM qt_partnumber_book_items WHERE partnumber IN (...)`
+
+### Rationale
+
+- The flag is obsolete and does not participate in current business logic.
+- PN composition is now fully represented by `lots_json`; qty semantics belong there.
+- Keeping a dead compatibility field increases drift between documented and actual architecture.
+
+### Constraints
+
+- `lots_json` is the only quantity-bearing field in the PN book contract.
+- Any consumer still relying on `is_primary_pn` must be updated in the same change wave.
+
+### Files
+
+- Migration: `migrations/031_drop_is_primary_pn.sql`
+- Models: `internal/models/lot.go`
+- Services: `internal/services/partnumber_book.go`, `internal/services/vendor_mapping.go`
+- Docs: `bible-local/vendor-mapping.md`
+
+## 2026-03-07: Drop Legacy `lot_partnumbers` And Bundle Tables
+
+### Decision
+
+Removed `lot_partnumbers`, `qt_lot_bundles`, and `qt_lot_bundle_items` from active runtime architecture. The only canonical PN mapping store is `qt_partnumber_book_items`, one row per `partnumber`, with full composition in `lots_json`.
+
+### What changed
+
+- Added migration `032_drop_legacy_vendor_mapping_tables.sql` dropping:
+  - `qt_lot_bundle_items`
+  - `qt_lot_bundles`
+  - `lot_partnumbers`
+- `internal/models/models.go` no longer auto-migrates legacy mapping and bundle tables.
+- Active runtime paths now read canonical mappings from `qt_partnumber_book_items`.
+- Tests were updated to seed `qt_partnumber_book_items` instead of `lot_partnumbers` and bundle tables.
+- Active Bible docs now describe `qt_partnumber_book_items` as the only source of truth.
+
+### Rationale
+
+- QuoteForge and partnumber books already use `lots_json`.
+- Keeping legacy mapping tables in runtime created two conflicting contracts.
+- Multi-LOT PN composition belongs in `lots_json`, not in auxiliary bundle tables.
+
+### Constraints
+
+- `qt_partnumber_book_items` must remain deduplicated by `partnumber`.
+- Multi-LOT PN composition exists only in `lots_json`.
+- Vendor is metadata in `qt_vendor_partnumber_seen`, not part of the canonical mapping key.
+- Deprecated Go structs may remain temporarily for in-memory test compatibility, but runtime must not auto-create or depend on the dropped tables.
+
+### Files
+
+- Migration: `migrations/032_drop_legacy_vendor_mapping_tables.sql`
+- Models: `internal/models/models.go`, `internal/models/lot.go`
+- Services: `internal/services/vendor_mapping.go`, `internal/services/stock_import.go`, `internal/services/partnumber_book.go`
+- Warehouse: `internal/warehouse/snapshot.go`
+- Matcher: `internal/lotmatch/matcher.go`
+- Docs: `bible-local/BIBLE.md`, `bible-local/vendor-mapping.md`, `bible-local/architecture.md`, `bible-local/pricelist.md`, `bible-local/data-rules.md`
+
+## 2026-03-07: Vendor Mapping LOT Autocomplete Must Not Clip Long Names
+
+### Decision
+
+LOT autocomplete in the Vendor Mapping modal must use a custom popup, not native browser `datalist`, because operators work with long monospaced LOT names that must remain fully readable.
+
+### What changed
+
+- Vendor Mapping modal uses a custom positioned suggestion popup for LOT inputs.
+- Popup width may exceed the input width and is constrained by viewport, not by the text field width.
+- Native `datalist` is not used for this field anymore.
+
+### Rationale
+
+- Native browser suggestion UIs clip long LOT names unpredictably and differ across platforms.
+- Operators need to distinguish long hardware LOT names visually before selection; clipped suffixes are operationally unsafe.
+
+### Constraints
+
+- LOT suggestion popup must allow long names to be fully visible or significantly less clipped than the input width.
+- Popup must stay above the modal overlay and follow the active LOT input.
+
+### Files
+
+- UI: `web/templates/vendor_mappings.html`
+
 ---
 
 ## 2026-02-20: Pricelist Formation Hardening (Estimate/Warehouse/Meta)