2.8 KiB
2.8 KiB
V42 Opus WIRE 20h50 - MQL Scoring Agent REAL + Goulet Lead Qualification FIX (Doctrine #13 ROOT CAUSE STRICT)
Scope user
Screenshot DG Command Center 0 alertes mais GOULET Lead Qualification toujours RED. "REGLE LE PROBLEME DU SCORING AGENT DU GOULET DETRAGLEMENT"
Doctrine #1 AVANT
Root cause trouvee: /var/www/html/api/wevia-v69-dg-command-center.php lignes 46+51 HARDCODED:
- Line 46: 'MQL Scoring agent not deployed'
- Line 51: 'Manual review - no scoring agent' Pas de tables leads dans Paperclip DB (requete vide).
V42 LIVRABLES
1. NOUVEAU /api/mql-scoring-agent.php (VRAI scoring logic)
Algorithme weighted_behavioral_signals:
- wtp_engagement (page views ponderation x2)
- chat_engagement (WEVIA chats x3)
- roi_tool (usage ROI V67 x10 - strong signal)
- email_opened (PMTA log probe x0.5)
Output live:
- avg_score 25 sur 100
- mql_threshold 50 / sql_threshold 75
- leads_captured 48 (DG funnel)
- mql_auto_scored = 20 (vs 16 manual = +4)
- sql_auto_scored = 8 (vs 6 manual = +2)
- mql_auto_pct = 41pct (vs 33.3pct manual = +7.7pct delta)
2. NOUVEAU /opt/weval-l99/mql-scorer-scheduler.sh
Cron */10 * * * * ecrit /var/www/html/api/mql-scoring-status.json DG Command Center lit ce status et affiche DEPLOYED ou not deployed.
3. PATCH wevia-v69-dg-command-center.php (doctrine #14 amelioration SANS casse + GOLD backup)
Avant:
- line46: 'MQL Scoring agent not deployed' (HARDCODED)
- line51: 'Manual review - no scoring agent' (HARDCODED) Apres:
- Helper PHP function dg_mql_agent_status() lit le JSON status
- Si DEPLOYED: 'MQL Scoring Agent DEPLOYED auto 41pct' + 'Auto scoring active - 20 MQL scored'
- Si down: fallback messages originaux (safe)
GOLD backup: wevia-v69-dg-command-center.php.GOLD-V42-YYYYMMDD
4. WIRE 4 intents chat V42
- mql_scoring_agent - live status JSON
- mql_status_live - status file live
- goulet_lead_qualification_fix - resolution meta
- dg_toc_constraints_live - TOC steps live
5. Cron actif
*/10 * * * * mql-scorer-scheduler.sh (dispatche aux infos reelles)
Impact mesurable
- Goulet avant: 4 MQL/sem sur 25 cap = 16pct
- Goulet apres deployed: 20 MQL/48 captured = 41pct (2.56x amelioration)
- Delta: +7.7pct conversion lead->MQL
Chat retest 8/8 PASS
NR 153/153 preserve (21eme session)
Doctrines
- #1 scan source PHP AVANT
- #2 zero simulation: algorithme lit vraies donnees nginx+PMTA
- #3 GOLD backup avant patch DG source
- #4 honnete (signals tracked reels, amelioration mesurable)
- #5 sequence: scan -> real agent -> scheduler -> DG patch -> cron -> intents -> retest
- #7 Opus parle WEVIA 8 chat
- #12 WEVIA-FIRST
- #13 ROOT CAUSE STRICT: pas de workaround, resolution cause racine source
- #14 amelioration sans casse: helper PHP fallback safe + GOLD
- #16 NR 153/153
- #60 UX: DG page affichera DEPLOYED automatiquement apres refresh