Orders were being flushed but only committed every 10 orders via the
progress callback. With slow imports (~1.5 min/order), no commits
happened before the first 10 orders, causing data loss if the import
failed or was interrupted.
Changes:
- Add db.commit() after each successful order creation
- Add db.rollback() on order creation failure to prevent corrupt state
- Add db.commit() after order updates
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
func.case() is not valid SQLAlchemy - case() must be imported directly
from sqlalchemy. This was causing TypeError on /api/v1/admin/background-tasks/tasks/stats.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Historical import jobs were not appearing in the recent jobs list because
list_letzshop_jobs() only queried marketplace_import_jobs and letzshop_sync_logs.
Changes:
- Add LetzshopHistoricalImportJob query to list_letzshop_jobs()
- Add current_phase and error_message fields to LetzshopJobItem schema
- Fixed stuck job 8 (marked as failed)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replaces the "fail on missing product" behavior with graceful handling:
- Orders import even when products aren't found by GTIN
- Unmatched items link to a per-vendor placeholder product
- Exceptions tracked in order_item_exceptions table for QC resolution
- Order confirmation blocked until exceptions are resolved
- Auto-matching when products are imported via catalog sync
New files:
- OrderItemException model and migration
- OrderItemExceptionService with CRUD and resolution logic
- Admin and vendor API endpoints for exception management
- Domain exceptions for error handling
Modified:
- OrderItem: added needs_product_match flag and exception relationship
- OrderService: graceful handling with placeholder products
- MarketplaceProductService: auto-match on product import
- Letzshop confirm endpoints: blocking check for unresolved exceptions
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Restructured create_letzshop_order to follow the "validate first, then write"
architecture pattern:
Phase 1 (Read-only validation):
- Check if order already exists
- Parse all inventory units and collect GTINs
- Batch query all products by GTIN (single query instead of N queries)
- Validate all products exist - raise ValidationException BEFORE any writes
Phase 2 (Database writes):
- Only after all validation passes, create customer, order, and items
This ensures if validation fails, no database modifications happen, so the
endpoint/task simply doesn't commit - no rollback needed.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When create_letzshop_order raises an exception (e.g., product not found by GTIN)
after flushing the order to the database, the session is left in an inconsistent
state. Without rollback, subsequent database operations fail or hang, causing
the import to get stuck at "0 processed...".
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add the missing method to OrderService that was being called
from the admin orders API endpoint.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Update API endpoints and schemas to use the unified Order model:
- Update Letzshop order schemas with OrderItem support
- Update API responses to use new field names (external_*, status, etc.)
- Update confirm/reject endpoints to use OrderItem.external_item_id
- Update Letzshop order service for unified Order model queries
- Update documentation to reflect completed implementation
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add historical order import functionality
- Add order detail page route
- Update API endpoints for order confirmation flow
Note: These files need further updates to use the new unified order model.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add create_letzshop_order() for importing Letzshop orders
- Add find_or_create_customer() for marketplace imports (inactive by default)
- Add set_order_tracking() for shipping updates
- Add update_item_state() for marketplace confirmation flow
- Add get_order_by_external_shipment_id() for Letzshop lookups
- Update all methods to use snapshot fields instead of address FKs
- Products must exist by GTIN - raises error if not found
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Per Letzshop API, each inventory unit must be confirmed/declined individually.
This enables partial confirmation (some items confirmed, others declined).
Admin API endpoints:
- POST /vendors/{id}/orders/{id}/confirm - confirm all items
- POST /vendors/{id}/orders/{id}/reject - decline all items
- POST /vendors/{id}/orders/{id}/items/{id}/confirm - confirm single item
- POST /vendors/{id}/orders/{id}/items/{id}/decline - decline single item
Order detail modal now shows:
- Product name, EAN, SKU, MPN, price per item
- Per-item state badge (unconfirmed/confirmed/declined)
- Per-item confirm/decline buttons for pending items
- Bulk confirm/decline all buttons
Order status logic:
- If all items declined -> order is "declined"
- If any item confirmed -> order is "confirmed"
- Partial confirmation supported
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Fix historical import to also update sync_status when it's out of
sync with letzshop_state (e.g., confirmed orders showing as pending)
- Fix frontend to read stats from response.statistics nested object
- Orders with matching letzshop_state but wrong sync_status now get fixed
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add admin API endpoints for order management
- Add orders page with vendor selector and filtering
- Add order schemas for admin operations
- Support order status tracking and management
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add admin API endpoints for inventory management
- Add inventory page with vendor selector and filtering
- Add admin schemas for cross-vendor inventory operations
- Support digital products with unlimited inventory
- Add integration tests for admin inventory API
- Add inventory management guide documentation
Mirrors vendor inventory functionality with admin-level access.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add historical order import with pagination support
- Add customer_locale, shipping_country_iso, billing_country_iso columns
- Add gtin/gtin_type columns to Product table for EAN matching
- Fix order stats to count all orders server-side (not just visible page)
- Add GraphQL introspection script with tracking workaround tests
- Enrich inventory units with EAN, MPN, SKU, product name
- Add LetzshopOrderStats schema for proper status counts
Migrations:
- a9a86cef6cca: Add locale and country fields to letzshop_orders
- cb88bc9b5f86: Add gtin columns to products table
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add new Marketplace section in admin sidebar with Letzshop sub-item
- Remove old Import and Letzshop Orders items from Product Catalog
- Create unified Letzshop management page with 3 tabs:
- Products tab: Import/Export functionality
- Orders tab: Order management with confirm/reject/tracking
- Settings tab: API credentials and CSV URLs
- Add unified jobs table showing imports, exports, and order syncs
- Implement vendor autocomplete using Tom Select library (CDN + fallback)
- Add /vendors/{vendor_id}/jobs API endpoint for unified job listing
- Move database queries to service layer (LetzshopOrderService)
- Add LetzshopJobItem and LetzshopJobsListResponse schemas
- Include Tom Select CSS/JS assets as local fallback
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove |safe from |tojson in HTML attributes (x-data) - quotes must
become " for browsers to parse correctly
- Update LANG-002 and LANG-003 architecture rules to document correct
|tojson usage patterns:
- HTML attributes: |tojson (no |safe)
- Script blocks: |tojson|safe
- Fix validator to warn when |tojson|safe is used in x-data (breaks
HTML attribute parsing)
- Improve code quality across services, APIs, and tests
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Create letzshop_export_service.py with CSV generation logic
- Add admin endpoint: GET /api/v1/admin/vendors/{vendor}/export/letzshop
- Add vendor endpoint: GET /api/v1/vendor/letzshop/export
- Support language selection (en, fr, de) and include_inactive filter
- Generate Google Shopping compatible tab-delimited CSV format
- Add comprehensive integration tests for both endpoints
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Fix inventory_service.py: use product.vendor_sku instead of non-existent product.product_id
- Fix test: create MarketplaceProductTranslation for title (now in translations table)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add Optional type annotations to nullable parameters in service methods
to fix architecture validation warnings.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Sorting:
- Add id DESC as tiebreaker when created_at timestamps are equal
- Ensures consistent newest-first ordering on imports and marketplace pages
Language bug fix:
- Add language field to startImport() payload in marketplace.js
- Add language column to MarketplaceImportJob database model
- Store and return language in API responses
- Add cache-busting version parameter to script tag
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Bug: Language selector worked on UI but import always used 'en'
Root causes:
1. Frontend: startImport() was not including language in API payload
2. Backend: language was not stored in import job database record
3. Backend: language was not returned in API response models
Fixes:
- Add language to payload in marketplace.js startImport()
- Add language column to MarketplaceImportJob model
- Store language when creating import job in service
- Include language in both response model converters
- Add database migration for language column
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Import Error Tracking:
- Add MarketplaceImportError model to store detailed error information
- Store row number, identifier, error type, message, and row data for each error
- Add API endpoint GET /admin/marketplace-import-jobs/{job_id}/errors
- Add UI to view and browse import errors in job details modal
- Support pagination and error type filtering
Translation Tabs:
- Replace flat translation list with tabbed interface on product detail page
- Add language tabs with full language names
- Add copy-to-clipboard functionality for translation content
- Improved UX with better visual separation of translations
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Service layer:
- Remove db.commit() calls from credentials_service.py (SVC-006)
- Move transaction control to API endpoint level
- Rename client.py -> client_service.py (NAM-002)
- Rename credentials.py -> credentials_service.py (NAM-002)
JavaScript:
- Use centralized logger in admin letzshop.js (JS-001)
- Replace console.log/error with LogConfig logger
Frontend templates:
- Use page_header_flex macro for page header (FE-007)
- Use error_state macro for error display (FE-003)
- Use table_wrapper macro for vendors table (FE-005)
- Use modal macro for configuration and orders modals (FE-004)
All 31 Letzshop tests pass. Architecture validation: 0 errors, 0 warnings.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Fix collect_tests to use JSON report parsing (was returning 0 tests)
- Add Test Collection panel to testing dashboard showing total tests,
unit/integration/performance breakdown, and file count
- Reorganize sidebar: create Platform Health section with Testing Hub,
Code Quality, and Background Tasks
- Keep Developer Tools for Components and Icons only
- Platform Monitoring now contains Import Jobs and Application Logs
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Adds a unified view of all background tasks (imports and test runs)
under Platform Monitoring. Includes real-time status polling,
statistics overview, and task history.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Improve the testing dashboard to run pytest in the background:
- Add background task execution using FastAPI's BackgroundTasks
- Create test_runner_tasks.py following existing background task pattern
- API now returns immediately after starting the test run
- Frontend polls for status every 2 seconds until completion
- Show running indicator with elapsed time counter
- Resume polling if user navigates away and returns while tests running
- Tests continue running even if user closes the page
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add pytest-json-report to requirements-test.txt for JSON test output
- Improve error handling in test runner service to catch JSONDecodeError
and fallback to parsing stdout when JSON report is unavailable
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add a new Testing Dashboard page that replaces the old Testing Hub with
pytest integration:
- Database models for test runs, results, and collections (TestRun,
TestResult, TestCollection)
- Test runner service that executes pytest with JSON reporting and
stores results in the database
- REST API endpoints for running tests, viewing history, and statistics
- Dashboard UI showing pass rates, trends, tests by category, and top
failing tests
- Alembic migration for the new test_* tables
The dashboard allows admins to:
- Run pytest directly from the UI
- View test run history with pass/fail statistics
- See trend data across recent runs
- Identify frequently failing tests
- Collect test information without running
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add admin marketplace products page to browse imported products
- Add admin vendor products page to manage vendor catalog
- Add product detail pages for both marketplace and vendor products
- Implement copy-to-vendor API to copy marketplace products to vendor catalogs
- Add vendor product service with CRUD operations
- Update sidebar navigation with new product management links
- Add integration and unit tests for new endpoints and services
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add language parameter to import endpoints and background tasks
- Extract translation fields (title, description, short_description)
- Create/update MarketplaceProductTranslation records during import
- Add MarketplaceProductTranslationSchema for API responses
- Map product_type column to product_type_raw to avoid enum conflict
- Parse prices to numeric format (price_numeric, sale_price_numeric)
- Update marketplace product service for translation-based lookups
- Update CSV export to retrieve titles from translations
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- test_create_import_job_database_error: Monkeypatch db.flush instead
of db.commit since service uses flush() per architecture rules
- test_create_product_already_exists: Store product_id before exception
and rollback session to clear PendingRollbackError state
- test_delete_product_success: Add db.flush() to delete_product service
method to ensure deletion is applied before verification
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove unused register_user() method and helper methods
- Remove legacy UserRegister schema (customer registration uses CustomerService)
- Remove wrapper methods that just delegated to auth_manager
- Simplify auth_service to focus on login and vendor access control
- Clean up tests to match simplified service
The only registration path is now /api/v1/shop/auth/register for customers,
which uses CustomerService.register_customer().
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Change register_user() to use db.flush() instead of db.commit()
and remove db.rollback() to follow the established architecture:
- Services use flush() for database operations
- Endpoints handle transaction commit
- Exception handlers manage rollback
This resolves SVC-006 architecture violation.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Services now use db.flush() instead of db.commit() for database operations
- API endpoints handle transaction commit after service calls
- Remove db.rollback() from services (let exception handlers manage this)
- Ensures consistent transaction boundaries at API layer
This pattern gives API endpoints full control over when to commit,
allowing for better error handling and potential multi-operation transactions.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Move database queries to service layer and use proper Pydantic responses:
Schema changes (models/schema/marketplace_import_job.py):
- Add AdminMarketplaceImportJobResponse with extra fields (id, error_details,
created_by_name)
- Add AdminMarketplaceImportJobListResponse with items/total/page/limit format
Service changes (app/services/marketplace_import_job_service.py):
- Add convert_to_admin_response_model() for admin-specific response
- Add get_all_import_jobs_paginated() for admin listing
- Add get_import_job_by_id_admin() without access control
API changes (app/api/v1/admin/marketplace.py):
- Use service methods instead of direct DB queries
- Use proper Pydantic response models instead of dicts
- All business logic now in service layer
Cleanup:
- Remove duplicate methods from admin_service.py
Architecture validation now passes with 0 violations.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Services:
- admin_service: Add source_url to import job response, fix vendor_name
to use vendor.name relationship instead of vendor_name field
- marketplace_product_service: Include reserved_quantity and
available_quantity in InventoryLocationResponse
- vendor_service: Fix _get_product_by_id_or_raise to use database ID
(int) instead of marketplace_product_id (str)
Schema:
- Add InventorySummaryResponse model for marketplace product service
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Update architecture rules to be stricter (API-003 now blocks ALL exception
raising in endpoints, not just HTTPException)
- Update get_current_vendor_api dependency to guarantee token_vendor_id presence
- Remove redundant _get_vendor_from_token helpers from all vendor API files
- Move vendor access validation to service layer methods
- Add Pydantic response models for media, notification, and payment endpoints
- Add get_active_vendor_by_code service method for public vendor lookup
- Add get_import_job_for_vendor service method with vendor validation
- Update validation script to detect exception raising patterns in endpoints
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Vendors can now override company contact information for specific branding.
Fields are nullable - if null, value is inherited from parent company.
Database changes:
- Add vendor.contact_email, contact_phone, website, business_address, tax_number
- All nullable (null = inherit from company)
- Alembic migration: 28d44d503cac
Model changes:
- Add effective_* properties for resolved values
- Add get_contact_info_with_inheritance() helper
Schema changes:
- VendorCreate: Optional contact fields for override at creation
- VendorUpdate: Contact fields + reset_contact_to_company flag
- VendorDetailResponse: Resolved values + *_inherited flags
API changes:
- GET/PUT vendor endpoints return resolved contact info
- PUT accepts contact overrides (empty string = reset to inherit)
- _build_vendor_detail_response helper for consistent responses
Service changes:
- admin_service.update_vendor handles reset_contact_to_company flag
- Empty strings converted to None for inheritance
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
vendor_service.py:
- Remove 6 db.commit() calls from service methods
- Use db.flush() in create methods to get IDs without committing
- Remove db.rollback() calls (endpoint handles transaction)
- Methods affected: create_vendor, toggle_verification, set_verification,
toggle_status, set_status, add_product_to_catalog
vendors.py (endpoints):
- Add db.commit() after set_verification() call
- Add db.commit() after set_status() call
This follows the industry pattern: one request = one transaction,
with commit controlled at the API endpoint level.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add new methods to vendor_service.py:
- get_vendor_by_id(): Fetch vendor by ID with proper exception
- get_vendor_by_identifier(): Fetch by ID or code
- toggle_verification/set_verification(): Manage vendor verification
- toggle_status/set_status(): Manage vendor active status
Refactor vendors.py API endpoints:
- Remove _get_vendor_by_identifier helper with direct DB queries
- All endpoints now use vendor_service methods
- Remove direct db.commit() calls from endpoints
This fixes API-002 violations and improves architecture compliance.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
SQLAlchemy Error Fix:
- Add .unique() when using joinedload(Company.vendors)
- When eagerly loading collection relationships with joinedload, SQLAlchemy can return duplicate rows
- The unique() method deduplicates results and is required for joined collection loads
Error was:
InvalidRequestError: The unique() method must be invoked on this Result, as it contains results that include joined eager loads against collections
This is a standard SQLAlchemy pattern for handling one-to-many relationships with eager loading.
- Add database migration to make vendor.owner_user_id nullable
- Update Vendor model to support company-based ownership (DEPRECATED vendor.owner_user_id)
- Implement company_service with singleton pattern (consistent with vendor_service)
- Create Company model with proper relationships to vendors and users
- Add company exception classes for proper error handling
- Refactor companies API to use singleton service pattern
Architecture Change:
- OLD: Each vendor has its own owner (vendor.owner_user_id)
- NEW: Vendors belong to a company, company has one owner (company.owner_user_id)
- This allows one company owner to manage multiple vendor brands
Technical Details:
- Company service uses singleton pattern (not factory)
- Company service accepts db: Session as parameter (follows SVC-003)
- Uses AuthManager for password hashing (consistent with admin_service)
- Added _generate_temp_password() helper method
Problem:
- GET /api/v1/admin/code-quality/stats returned 500 error
- Pydantic validation error: "last_scan Field required"
- When no scan data exists, service returned dict without last_scan field
- DashboardStatsResponse model required last_scan field
Solution:
1. Added last_scan: None to empty state return dictionary
2. Made last_scan field explicitly optional with default value (= None)
3. Ensures field is always present in response
Changes:
- app/services/code_quality_service.py: Added "last_scan": None
- app/api/v1/admin/code_quality.py: Changed to last_scan: str | None = None
This fixes the 500 error when accessing /admin/code-quality page
with no architecture scan data in the database.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
