- New scrape_content() method in enrichment_service: extracts meta
description, H1/H2 headings, paragraphs, images (filtered for size),
social links, service items, and detected languages using BeautifulSoup
- Scans 6 pages per prospect: /, /about, /a-propos, /services,
/nos-services, /contact
- Results stored as JSON in prospect.scraped_content_json
- New endpoints: POST /content-scrape/{id} and /content-scrape/batch
- Added to full_enrichment pipeline (Step 5, before security audit)
- CONTENT_SCRAPE job type for scan-jobs tracking
- "Content Scrape" batch button on scan-jobs page
- Add beautifulsoup4 to requirements.txt
Tested on batirenovation-strasbourg.fr: extracted 30 headings,
21 paragraphs, 13 images.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Detect API-level errors (quota exceeded, invalid URL) in response JSON
and store in scan_error instead of silently writing zeros
- Show scan error message on the performance card when present
- Show "No performance data — configure PAGESPEED_API_KEY" when all
scores are 0 and no error was recorded
- Add accessibility and best practices scores to performance card
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix contact_type column: Enum(ContactType) → String(20) to match the
migration (fixes "type contacttype does not exist" on insert)
- Rewrite scrape_contacts with structured-first approach:
Phase 1: tel:/mailto: href extraction (high confidence)
Phase 2: regex fallback with SVG/script stripping, international phone
pattern (requires + prefix, min 10 digits)
Phase 3: address extraction from Schema.org JSON-LD, <address> tags,
and European street address regex (FR/DE/EN street keywords)
- URL-decode email values, strip tags to plain text for cross-element
address matching
- Add /mentions-legales to scanned paths
Tested on batirenovation-strasbourg.fr: finds 3 contacts (email, phone,
address) vs 120+ false positives and a crash before.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Reverts the expandable sub-row design back to a clean one-row-per-member
table. All per-store management now happens inside the edit modal.
Table: simple 4-column layout (Member | Stores & Roles | Status | Actions)
with view + edit buttons. Store badges show orange for pending stores.
Edit modal enhanced with per-store cards showing:
- Store name, code, and status badge (Active/Pending)
- Role dropdown + Update button (for active stores)
- Resend invitation button (for pending stores)
- Remove from store button
- "Remove from all stores" link at bottom
Removed: expandedMembers, flattenedRows, toggleMemberExpand,
resendStoreInvitation, resendInvitation (member-level).
Added: resendForStore, removeFromStore (work inside edit modal).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix IPv6 host parsing with _strip_port() utility
- Remove dangerous StorePlatform→Store.subdomain silent fallback
- Close storefront gate bypass when frontend_type is None
- Add custom subdomain management UI and API for stores
- Add domain health diagnostic tool
- Convert db.add() in loops to db.add_all() (24 PERF-006 fixes)
- Add tests for all new functionality (18 subdomain service tests)
- Add .github templates for validator compliance
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move db.commit() from services to API endpoints and Celery tasks.
Services now use db.flush() only; endpoints own the transaction.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Migrates scanning pipeline from marketing-.lu-domains app into Orion module.
Supports digital (domain scan) and offline (manual capture) lead channels
with enrichment, scoring, campaign management, and interaction tracking.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
