- Fix get_files_by_status query: LATERAL join to get latest extraction
per file, returning vendor_name, vendor_match_confidence, total_amount
- Fix renderUnhandledFiles: use best_vendor_name, convert confidence
0-1 to %, use total_amount field
- Add rerun button for all files (not just failed) via rerunSingleFile()
- rerunSingleFile() calls /reprocess/{file_id} and reloads unhandled tab
- Fix retry_extraction endpoint to actually run extraction immediately
extract_vendor_suggestion now matches found CVR/domain/name against
vendors table for ALL code paths (not just fast path):
- CVR match → score 100
- Domain match → score 80
- Email-domain match → score 75
- Name substring → score 60
Frontend auto-links when match_score >= 80 (was only pdf_extraction source).
Shows score reason: CVR-match/domæne-match/navne-match in success toast.
Also: saves newly found CVR to extracted_vendor_cvr so fast path
works on subsequent calls for old emails too.
- extract_vendor_suggestion: fast path if extracted_vendor_cvr already set
- CVR lookup in vendors table → returns vendor_id if match found
- source='pdf_extraction' with confidence=0.95
- Frontend quickCreateVendor:
- If vendor_id returned → auto-link without modal (no manual input needed)
- New source label '📄 PDF-faktura' for pdf_extraction source
- Added execute_query_single import to emails router
Fixes 'list indices must be integers or slices, not str' error when
uploading a duplicate file. execute_query returns a list, so accessing
existing_file['file_id'] caused TypeError. Now uses execute_query_single
which returns a single dict row.
- Updated migrations.html to detect production environment and use podman/docker accordingly
- Added container runtime info to settings page
- Updated VERSION to 2.1.1
v1.3.148:
- Fixed JavaScript error: matrixHtml is not defined
- Removed duplicate table population code
- bodyRows.innerHTML already contains the matrix HTML
v1.3.147:
- Changed from /customers/{id}/invoices/sent (404) to /invoices/drafts
- Fetch all drafts, then filter by customer number
- Apply date filter ONLY on customer invoices (fixes bug)
- Proper order: fetch all -> filter customer -> filter date
v1.3.146:
- Changed to /customers/{customerNumber}/invoices/sent
- More direct approach - fetches only that customer's invoices
- Removed unnecessary customer filtering step
- Still applies 13-month date filter
v1.3.145:
- CRITICAL FIX: Date filter was running on all invoices
- Now correctly filters customer_invoices by date
- Prevents showing other customers' invoices
v1.3.143:
- Check drafts, booked, paid, unpaid endpoints
- Deduplicate invoices by invoice number
- Pagination support for each endpoint
- Filter by customer + 13 months date after fetching
v1.3.142:
- Changed from /invoices/sent to /invoices/drafts
- /drafts contains most active invoices (29 vs 0 in /sent)
- Still applies pagination and 13-month filter
v1.3.141:
- Removed unsupported customer filter from API params
- Fetch all invoices from /invoices/sent with pagination
- Filter by customer number in code
- Apply 13-month date filter after customer filter
v1.3.140:
- Use /invoices/sent endpoint with customer.customerNumber filter
- More efficient: only fetch invoices for specific customer
- Apply 13-month date filter after fetching
- Simplified endpoint logic (single endpoint vs 8 endpoints)
v1.3.139:
- Removed date filter from API params (causes 404 on some endpoints)
- Apply 13-month filter in code after fetching invoices
- Parse invoice dates and filter >= start_date
- More reliable filtering across all e-conomic endpoints
v1.3.138:
- Fixed ValueError: month must be in 1..12
- Use python-dateutil for proper month arithmetic
- Uses relativedelta for correct 13 month calculation
v1.3.136:
- Changed filter from 365 days to exactly 12 months
- Uses 1st day of the month 12 months ago
- Example: Jan 27, 2026 -> fetches from Feb 1, 2025
