"Jubiler Agat" now matches "agat-jubiler.pl" by checking individual
words in any order, not just concatenated substring.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Evaluate top 3 Brave results instead of just taking the first one.
Add domain name matching signal (+2 pts when domain contains company name).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Show Brave search description under company name and scraped page
text snippet (first 500 chars) as expandable row below each
candidate, helping admin verify if the URL matches the company.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Added norda-biznes.info, bizraport.pl, aplikuj.pl, lexspace.pl,
drewnianeabc.pl, f-trust.pl, itspace.llc to directory blacklist
- Delay first poll by 3s so thread has time to populate total
- Better completion messages (show count, handle 0 remaining)
- Increase poll interval to 3s
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Bulk query now excludes companies that already have pending/accepted
candidates, so only truly new companies are processed via Brave API.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Added imsig.pl, monitorfirm.pb.pl, zwiazekpracodawcow.pl,
transfermarkt.pl, mapcarta.com and other directories/portals
that returned false positives in first production run.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Brave free tier rate limits aggressively (429 after ~1 req/s).
Added retry logic (3 attempts: 3s, 6s, 9s waits) and increased
inter-company delay from 2s to 5s. Error candidates are now
cleaned up before retry to allow re-discovery.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace single latest_result field with cumulative log array and
offset-based polling to prevent missed entries and race conditions.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Automated discovery using Brave Search API to find company websites,
scrape verification data (NIP/REGON/KRS/email/phone), and present
candidates with match badges in the data quality dashboard.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Treat & as word connector (P&P, S&K stay as single tokens)
- Add prefix matching with legal suffix stripping (Sp. z o.o., S.A.)
- Add reverse prefix for brand vs legal name (Pixlab Softwarehouse ↔ Pixlab Sp. z o.o.)
- Compound names like TERMO-BUD still correctly rejected (no space separator)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
TERMO vs TERMO-BUD was incorrectly accepted (score 1.0) because
denominator only counted company words. Now uses max(company, google)
so extra words in either name lower the score.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Replace substring matching with word-boundary tokenized matching
- Short names (1-2 words): require ALL significant words to match
- Longer names (3+): require at least 50% word overlap
- Pick best-scoring result instead of first match
- Add company_name validation to competitor_monitoring_service
- Show Google profile name in dashboard hints for admin verification
- Display mismatch warning when Google name differs from company name
Prevents cases like "IT Space" matching "Body Space" (score 0.50 < 1.00 threshold).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add clickable field coverage bars to filter companies missing specific data
- Add quick-action buttons (Registry/SEO/GBP) per company in dashboard table
- Add stale data detection (>6 months) with yellow badges
- Implement weighted priority score (contacts 34%, audits 17%)
- Add data hints in admin company detail showing where to find missing data
- Add "Available data" section showing Google Business data ready to apply
- Add POST /api/company/<id>/apply-hint endpoint for one-click data fill
- Extend website content updater with phone/email extraction (AI + regex)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Extract 12-field completeness scoring to utils/data_quality.py service
- Auto-update data_quality_score and data_quality label on company data changes
- Add /admin/data-quality dashboard with field coverage stats, quality distribution, and sortable company table
- Add bulk enrichment with background processing, step selection, and progress tracking
- Flow GBP phone/website to Company record when company fields are empty
- Display Google opening hours on public company profile
- Add BulkEnrichmentJob model and migration 075
- Refactor arm_company.py to support selective steps and progress callbacks
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Previously search_place() blindly returned the first result, which
could be a completely unrelated business. Now validates that at least
one significant word from the company name appears in the Google
result before accepting it. Prevents wrong GBP profiles being linked
to companies (e.g. Rozsadni Bracia getting Zielony Zolwik's profile).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Change section margins from --spacing-xl (2rem) to --spacing-md (1rem)
and reduce inner padding for a more compact layout.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove the dark/light background toggle from logo selection modal,
company detail pages (admin and public), and the toggle-logo-bg API
endpoint. The feature didn't meet UX requirements.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add explicit white background to gallery cards so only the inner
image container changes to dark, keeping card borders, text and
badges on their original light backgrounds.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When choosing a logo, admins can now switch between light and dark
backgrounds to see which works better for transparent logos. The
selected background preference is automatically saved when confirming.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The toggle was only on the admin panel - added it next to
"Pobierz logo" on the public company detail page too.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds a per-company setting to display logos on dark background,
useful for logos with white text or light-colored elements.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
img.get('class') can return an empty list [], causing IndexError
when accessing [0]. Added `or ['']` fallback.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The script was calling /firma?nip=X (wrong endpoint) instead of using
fetch_ceidg_by_nip() which does two-phase /firmy?nip=X then /firma/{id}.
Now uses the same service and field mapping as the admin panel button.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The script was calling auditor.audit_company() but not auditor.save_audit_result(),
causing profiles to be found but never persisted to the database.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- GBP: access .completeness_score attribute + call save_audit()
- Social: count saved DB records instead of parsing audit result dict
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- KRS: search_by_nip() returns dict, not just KRS number
- SEO: SEOAuditor(database_url) + audit_company(company_dict)
- Social: SocialMediaAuditor() + audit_company(company_dict)
- GBP: GBPAuditService(db) + audit_company(company_id)
- Support multiple company IDs in one invocation
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Allows running the same enrichment workflow as the "Uzbrój firmę" button
directly from the command line, without needing browser/admin login.
Usage: python3 scripts/arm_company.py <company_id> [--force]
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Two buttons side by side: "Uzbrój firmę" runs only missing steps,
"Zaktualizuj dane" forces re-run of all steps regardless of status.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The logo path was hardcoded to .webp even when the actual file was .svg,
causing broken image display for SVG logos like Orlex Design.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Logo check only looked for .webp files, missing SVG logos like
Orlex Design. Now checks both extensions.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When "Uzbrój firmę" runs automatically, logo step now picks the
recommended candidate instead of showing the gallery (which would
be overwritten by page reload). Manual "Szukaj logo" still shows
the gallery for admin selection.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move logo selection to step 5 (last), reorder: registry, SEO, social,
GBP, logo. After clicking "Szukaj logo", admin sees thumbnail gallery
of candidates and can pick the best one or cancel.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Each workflow action card now shows a numbered badge (1-5) indicating
the recommended execution order for administrators.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CEIDG enrichment now creates individual CompanyPKD records from the PKD
list, matching the pattern used by KRS enrichment for consistency.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
KRS API returns two date formats for za_okres field:
"01.01.2011 - 31.12.2011" (older) and "OD 01.01.2013 DO 31.12.2013" (newer).
Use regex to extract all date patterns instead of splitting on ' - '.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Map remaining KRS API fields: registration date, company agreement date,
duration, capital currency, OPP status, representation rules, shareholders
(wspólnicy), and financial reports (sprawozdania finansowe).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
KRS API returns kontakt_krs with www and email fields. These were not
being mapped to Company.website and Company.email. Now enrichment
extracts these and normalizes the URL (adds https:// prefix, lowercase).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The KRS enrichment function was only mapping forma_prawna, legal_name
and capital. Now maps REGON, address (ulica + nr_domu), city, postal,
primary PKD code on Company model, and NIP if missing. Fixed field
name mismatches: nr_domu (not numer), opis (not nazwa), glowna (not
glowny). City names are title-cased from KRS uppercase format.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Individual action buttons now reload the page after 1.5s to reflect
updated data and status indicators. The master "Uzbrój firmę" workflow
skips per-step reload and only reloads once after all steps complete.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New admin page at /admin/companies/{id}/detail showing company data,
completeness score, and action buttons for registry data (CEIDG/KRS),
logo fetch, SEO audit, social media audit, and GBP audit.
Includes "Uzbrój firmę" master button for sequential execution.
Company list now links to detail page with "NOWA" badge for recent entries.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
PDF files: "Otwórz" (inline in browser) + "Pobierz" (download)
DOCX files: "Pobierz" only (browsers can't display DOCX inline)
Removes LibreOffice on-the-fly conversion - simpler and more reliable.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Use /usr/bin/libreoffice (Gunicorn has limited PATH)
- Set HOME=/tmp (LibreOffice needs writable home dir)
- Handle FileNotFoundError with user-friendly message
- Improve error flash messages in Polish
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
DOCX/DOC documents are now converted to PDF using LibreOffice headless
when the user clicks "Otwórz". The converted PDF is cached next to the
original file so subsequent views are instant.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
PDF documents now have an "Otwórz" button that opens
the file in the browser's built-in PDF viewer.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
DocumentUploadService.get_file_path() resolves paths using uploaded_at,
so import scripts must store files in directories matching that date.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add document management routes (upload, download, soft-delete) to board blueprint,
link BoardDocument to BoardMeeting via meeting_id FK, add documents section to
meeting view template, and include import scripts for meeting 2/2026 data and PDFs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
