cdp-edge 1.0.0

package/server-edge-tracker/schema-sales-engine.sql

@@ -0,0 +1,113 @@
+-- CDP Edge — Schema Sales Engine (Fase 7)
+-- ROAS Feedback Loop + Nurture Engine + Lookalike Dinâmico
+-- Executar: wrangler d1 execute cdp-edge-db --file=schema-sales-engine.sql --remote
+
+-- ── ROAS Reports — histórico de performance por campanha ──────────────────────
+CREATE TABLE IF NOT EXISTS roas_reports (
+  id                 INTEGER PRIMARY KEY AUTOINCREMENT,
+  utm_source         TEXT    NOT NULL,
+  utm_campaign       TEXT    NOT NULL,
+  utm_content        TEXT    NOT NULL DEFAULT 'unknown', -- origem: quiz_*, video_*, landing_*, ctwa_*
+  period_days        INTEGER NOT NULL DEFAULT 30,
+  total_leads        INTEGER NOT NULL DEFAULT 0,
+  confirmed_buyers   INTEGER NOT NULL DEFAULT 0,
+  conversion_rate    REAL    NOT NULL DEFAULT 0,     -- 0.0–1.0
+  total_revenue      REAL    NOT NULL DEFAULT 0,     -- BRL
+  revenue_per_lead   REAL    NOT NULL DEFAULT 0,
+  ltv_accuracy       REAL    NOT NULL DEFAULT 0,     -- % leads High LTV que realmente compraram
+  top_qualification  TEXT,                           -- qualificação quiz dominante
+  bid_recommendation TEXT    NOT NULL DEFAULT 'maintain', -- increase|maintain|decrease|pause
+  created_at         TEXT    NOT NULL DEFAULT (datetime('now'))
+);
+
+-- Migração para instâncias existentes (idempotente via IF NOT EXISTS não funciona em ALTER TABLE —
+-- execute manualmente se a tabela já existir):
+-- ALTER TABLE roas_reports ADD COLUMN utm_content TEXT NOT NULL DEFAULT 'unknown';
+
+CREATE INDEX IF NOT EXISTS idx_roas_source_campaign ON roas_reports(utm_source, utm_campaign);
+CREATE INDEX IF NOT EXISTS idx_roas_content         ON roas_reports(utm_content);
+CREATE INDEX IF NOT EXISTS idx_roas_created_at      ON roas_reports(created_at);
+CREATE INDEX IF NOT EXISTS idx_roas_bid_rec         ON roas_reports(bid_recommendation);
+
+-- VIEW: últimos relatórios por campanha com evolução
+CREATE VIEW IF NOT EXISTS v_roas_latest AS
+SELECT
+  utm_source,
+  utm_campaign,
+  conversion_rate,
+  total_revenue,
+  revenue_per_lead,
+  ltv_accuracy,
+  top_qualification,
+  bid_recommendation,
+  created_at
+FROM roas_reports r1
+WHERE created_at = (
+  SELECT MAX(r2.created_at)
+  FROM roas_reports r2
+  WHERE r2.utm_source = r1.utm_source
+    AND r2.utm_campaign = r1.utm_campaign
+)
+ORDER BY total_revenue DESC;
+
+-- ── Nurture Sequences — filas de mensagens por qualificação ──────────────────
+CREATE TABLE IF NOT EXISTS nurture_sequences (
+  id             INTEGER PRIMARY KEY AUTOINCREMENT,
+  user_id        TEXT,                     -- _cdp_uid
+  qualification  TEXT    NOT NULL,         -- interessado | curioso
+  delay_days     INTEGER NOT NULL,         -- D+N após o quiz
+  channel        TEXT    NOT NULL,         -- whatsapp | email
+  message        TEXT    NOT NULL,         -- mensagem interpolada
+  subject        TEXT,                     -- assunto (email only)
+  send_at        TEXT    NOT NULL,         -- datetime de envio agendado
+  status         TEXT    NOT NULL DEFAULT 'pending',  -- pending | sent | failed | cancelled
+  sent_at        TEXT,                     -- quando foi realmente enviado
+  quiz_name      TEXT,                     -- nome do quiz (para interpolação)
+  phone          TEXT,                     -- telefone do lead
+  email          TEXT,                     -- email do lead
+  first_name     TEXT,                     -- nome (para personalização)
+  created_at     TEXT    NOT NULL DEFAULT (datetime('now')),
+  updated_at     TEXT    NOT NULL DEFAULT (datetime('now'))
+);
+
+CREATE INDEX IF NOT EXISTS idx_nurture_user_id   ON nurture_sequences(user_id);
+CREATE INDEX IF NOT EXISTS idx_nurture_status    ON nurture_sequences(status);
+CREATE INDEX IF NOT EXISTS idx_nurture_send_at   ON nurture_sequences(send_at);
+CREATE INDEX IF NOT EXISTS idx_nurture_qual      ON nurture_sequences(qualification);
+
+-- VIEW: fila de envio pendente
+CREATE VIEW IF NOT EXISTS v_nurture_pending AS
+SELECT
+  id, user_id, qualification, delay_days, channel,
+  message, send_at, phone, email, first_name, quiz_name
+FROM nurture_sequences
+WHERE status = 'pending'
+  AND send_at <= datetime('now')
+ORDER BY send_at ASC;
+
+-- VIEW: taxa de envio por qualificação
+CREATE VIEW IF NOT EXISTS v_nurture_stats AS
+SELECT
+  qualification,
+  channel,
+  COUNT(*)                                       AS total,
+  COUNT(CASE WHEN status = 'sent'    THEN 1 END) AS sent,
+  COUNT(CASE WHEN status = 'failed'  THEN 1 END) AS failed,
+  COUNT(CASE WHEN status = 'pending' THEN 1 END) AS pending,
+  ROUND(COUNT(CASE WHEN status = 'sent' THEN 1 END) * 100.0 / COUNT(*), 1) AS delivery_rate_pct
+FROM nurture_sequences
+GROUP BY qualification, channel;
+
+-- ── Lookalike Seeds — histórico de audiences enviadas ao Meta ─────────────────
+CREATE TABLE IF NOT EXISTS lookalike_seeds (
+  id              INTEGER PRIMARY KEY AUTOINCREMENT,
+  audience_id     TEXT    NOT NULL,    -- META_AUDIENCE_ID
+  seed_type       TEXT    NOT NULL,    -- 'buyer_confirmed' | 'high_intent' | 'quiz_comprador'
+  profiles_sent   INTEGER NOT NULL DEFAULT 0,
+  profiles_received INTEGER,           -- confirmado pela API Meta
+  period_days     INTEGER NOT NULL DEFAULT 30,
+  created_at      TEXT    NOT NULL DEFAULT (datetime('now'))
+);
+
+CREATE INDEX IF NOT EXISTS idx_lookalike_seed_type  ON lookalike_seeds(seed_type);
+CREATE INDEX IF NOT EXISTS idx_lookalike_created_at ON lookalike_seeds(created_at);

package/server-edge-tracker/schema-segmentation.sql

@@ -0,0 +1,219 @@
+-- Schema de Segmentação Dinâmica ML — CDP Edge Quantum Tier
+-- Versão: 1.0
+-- Data: 9 de Abril de 2026
+
+-- TABELA PRINCIPAL: Segmentos ML
+CREATE TABLE IF NOT EXISTS ml_segments (
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  cluster_id INTEGER NOT NULL,
+  cluster_name TEXT NOT NULL,                      -- Nome descritivo gerado por ML
+  clustering_algorithm TEXT NOT NULL,                 -- 'kmeans', 'dbscan', 'hierarchical'
+  client_vertical TEXT,                                -- 'curso-online', 'ecommerce', 'saaS'
+  created_at TEXT DEFAULT (datetime('now')),
+  updated_at TEXT DEFAULT (datetime('now')),
+
+  -- Estatísticas do cluster
+  size INTEGER NOT NULL,                                 -- Número de leads no cluster
+  percentage REAL NOT NULL,                              -- % do total (ex: 0.25 = 25%)
+
+  -- Características médias (centróides)
+  avg_ltv REAL,                                        -- LTV médio do cluster
+  avg_ltv_class REAL,                                   -- 0=Low, 1=Medium, 2=High (média)
+  avg_behavior_score REAL,                               -- Engajamento médio (0-100)
+  avg_engagement_score REAL,                             -- Interações médias (0-100)
+  avg_intention_level REAL,                               -- Intenção de compra média (0-100)
+  avg_days_since_lead REAL,                              -- Recência média em dias
+  avg_hour_of_day REAL,                                  -- Hora média do dia (0-23)
+  avg_is_weekend REAL,                                   -- % que acessa no fim de semana (0-1)
+  avg_is_business_hours REAL,                              -- % em horário comercial (0-1)
+  avg_bot_score REAL,                                    -- % humano médio (0-100, 100=100% humano)
+
+  -- Características dominantes (top features por cluster)
+  dominant_countries TEXT,                                -- JSON array: ["BR", "US", "AR"]
+  dominant_states TEXT,                                   -- JSON array: ["SP", "RJ", "MG"]
+  dominant_cities TEXT,                                   -- JSON array: ["São Paulo", "Rio de Janeiro"]
+  dominant_timezones TEXT,                                 -- JSON array: ["America/Sao_Paulo", "America/New_York"]
+  dominant_utm_sources TEXT,                              -- JSON array: ["facebook", "google", "tiktok"]
+  dominant_utm_mediums TEXT,                             -- JSON array: ["cpc", "organic", "social"]
+  dominant_features TEXT,                                   -- JSON array: ["ltv", "behavior_score", "engagement_score"]
+
+  -- Métricas de qualidade do clustering
+  silhouette_score REAL,                                   -- Coesão vs separação (-1 a 1, > 0.5 = bom)
+  cohesion REAL,                                        -- Similaridade média intra-cluster
+  separation REAL,                                       -- Distância média inter-cluster
+  inertia REAL,                                          -- Soma dos quadrados das distâncias (menor = melhor)
+
+  -- Recomendações automáticas geradas por ML
+  action_recommendations TEXT,                          -- JSON array: ["Priorizar remarketing", "Aumentar bid", ...]
+  bid_recommendations TEXT,                              -- JSON array: [{"adset_id": "123", "recommended_bid": "R$ 18.50"}, ...]
+  campaign_recommendations TEXT,                           -- JSON array: [{"creative": "VSL A", "audience": "Segment 0"}, ...]
+
+  -- Metadados de controle
+  is_active INTEGER DEFAULT 1,                            -- 0 = arquivado, 1 = ativo
+  min_data_points INTEGER,                                -- Mínimo de leads para formar cluster
+  epsilon REAL,                                         -- Para DBSCAN (distância máxima)
+  min_samples INTEGER,                                     -- Para DBSCAN (pontos mínimos para cluster)
+  max_depth INTEGER                                       -- Para Hierarchical (profundidade máxima)
+);
+
+-- Índices para performance
+CREATE INDEX IF NOT EXISTS idx_ml_segments_id ON ml_segments(id);
+CREATE INDEX IF NOT EXISTS idx_ml_segments_cluster ON ml_segments(cluster_id);
+CREATE INDEX IF NOT EXISTS idx_ml_segments_algorithm ON ml_segments(clustering_algorithm);
+CREATE INDEX IF NOT EXISTS idx_ml_segments_created ON ml_segments(created_at);
+CREATE INDEX IF NOT EXISTS idx_ml_segments_active ON ml_segments(is_active);
+CREATE INDEX IF NOT EXISTS idx_ml_segments_vertical ON ml_segments(client_vertical);
+
+-- TABELA DE ASSOCIAÇÃO: Segment Members (Quem está em cada segmento)
+CREATE TABLE IF NOT EXISTS ml_segment_members (
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  lead_id TEXT NOT NULL,                                  -- ID da tabela leads
+  cluster_id INTEGER NOT NULL,                              -- ID do cluster (ml_segments.id)
+  clustering_algorithm TEXT NOT NULL,                 -- Mesmo algoritmo usado para criar o cluster
+  confidence REAL NOT NULL,                               -- 0-1 (quão perto do centroide)
+  distance_to_centroid REAL,                              -- Distância Euclidiana ao centroide
+  updated_at TEXT DEFAULT (datetime('now')),
+
+  -- Metadados para rastreamento
+  assigned_at TEXT DEFAULT (datetime('now')),              -- Quando foi atribuído ao segmento
+  is_outlier INTEGER DEFAULT 0,                              -- 0 = normal, 1 = outlier (DBSCAN)
+  outlier_reason TEXT,                                     -- "behavior_score too high", "unusual geo", etc.
+
+  -- Características do lead no momento da atribuição
+  lead_ltv REAL,                                         -- LTV do lead no momento
+  lead_ltv_class REAL,                                   -- Classe de LTV do lead
+  lead_behavior_score REAL,                               -- Behavior score no momento
+  lead_engagement_score REAL,                             -- Engagement score no momento
+  lead_intention_level REAL,                               -- Intention level no momento
+  lead_days_since_lead REAL,                              -- Dias desde lead no momento
+  lead_hour_of_day REAL,                                  -- Hora do dia no momento
+  lead_is_weekend INTEGER,                                -- Se era fim de semana (0/1)
+  lead_is_business_hours INTEGER,                          -- Se era horário comercial (0/1)
+  lead_country TEXT,                                      -- País do lead
+  lead_state TEXT,                                       -- Estado do lead
+  lead_city TEXT,                                        -- Cidade do lead
+  lead_utm_source TEXT,                                  -- UTM source do lead
+  lead_utm_medium TEXT,                                   -- UTM medium do lead
+
+  -- Chave composta para evitar duplicatas
+  UNIQUE(lead_id, cluster_id, clustering_algorithm)
+);
+
+-- Índices para performance
+CREATE INDEX IF NOT EXISTS idx_ml_segment_members_lead ON ml_segment_members(lead_id);
+CREATE INDEX IF NOT EXISTS idx_ml_segment_members_cluster ON ml_segment_members(cluster_id);
+CREATE INDEX IF NOT EXISTS idx_ml_segment_members_algorithm ON ml_segment_members(clustering_algorithm);
+CREATE INDEX IF NOT EXISTS idx_ml_segment_members_confidence ON ml_segment_members(confidence);
+CREATE INDEX IF NOT EXISTS idx_ml_segment_members_outlier ON ml_segment_members(is_outlier);
+CREATE INDEX IF NOT EXISTS idx_ml_segment_members_assigned ON ml_segment_members(assigned_at);
+
+-- TABELA DE HISTÓRICO DE CLUSTERING (Auditoria)
+CREATE TABLE IF NOT EXISTS ml_clustering_history (
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  clustering_id INTEGER NOT NULL,                          -- ID em ml_segments
+  started_at TEXT NOT NULL,                               -- Quando iniciou o clustering
+  completed_at TEXT,                                     -- Quando terminou
+  algorithm TEXT NOT NULL,                                 -- 'kmeans', 'dbscan', 'hierarchical'
+  n_leads_processed INTEGER NOT NULL,                        -- Quantos leads foram processados
+  n_clusters_created INTEGER NOT NULL,                        -- Quantos clusters foram criados
+  total_duration_ms INTEGER NOT NULL,                      -- Duração total em milissegundos
+  workers_ai_neurons_used INTEGER,                         -- Neurônios usados (para billing)
+  status TEXT NOT NULL,                                    -- 'completed', 'failed', 'timeout'
+  error_message TEXT,                                       -- Se falhou, motivo do erro
+  parameters TEXT NOT NULL,                                 -- Parâmetros usados (JSON)
+  results_summary TEXT,                                    -- Resumo dos resultados (JSON)
+  created_at TEXT DEFAULT (datetime('now'))
+);
+
+-- Índices para auditoria
+CREATE INDEX IF NOT EXISTS idx_ml_clustering_history_clustering ON ml_clustering_history(clustering_id);
+CREATE INDEX IF NOT EXISTS idx_ml_clustering_history_started ON ml_clustering_history(started_at);
+CREATE INDEX IF NOT EXISTS idx_ml_clustering_history_algorithm ON ml_clustering_history(algorithm);
+CREATE INDEX IF NOT EXISTS idx_ml_clustering_history_status ON ml_clustering_history(status);
+
+-- TABELA DE FEATURE IMPORTANCE (Otimização Futura)
+CREATE TABLE IF NOT EXISTS ml_feature_importance (
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  feature_name TEXT NOT NULL,                             -- 'ltv', 'behavior_score', etc.
+  importance_score REAL NOT NULL,                           -- 0-1 (quão importante para clustering)
+  clustering_algorithm TEXT NOT NULL,                 -- Algoritmo usado para calcular
+  variance_explained REAL,                                 -- % da variância explicada
+  created_at TEXT DEFAULT (datetime('now')),
+  updated_at TEXT DEFAULT (datetime('now'))
+);
+
+-- Índices
+CREATE INDEX IF NOT EXISTS idx_ml_feature_importance_feature ON ml_feature_importance(feature_name);
+CREATE INDEX IF NOT EXISTS idx_ml_feature_importance_algorithm ON ml_feature_importance(clustering_algorithm);
+
+-- VIEW: Segmentos Ativos com Estatísticas
+CREATE VIEW IF NOT EXISTS v_active_segments AS
+SELECT
+  ms.id,
+  ms.cluster_id,
+  ms.cluster_name,
+  ms.clustering_algorithm,
+  ms.client_vertical,
+  ms.size,
+  ms.percentage,
+  ms.avg_ltv,
+  ms.avg_ltv_class,
+  ms.avg_behavior_score,
+  ms.avg_engagement_score,
+  ms.avg_intention_level,
+  ms.avg_days_since_lead,
+  ms.dominant_countries,
+  ms.dominant_states,
+  ms.dominant_cities,
+  ms.dominant_utm_sources,
+  ms.dominant_features,
+  ms.silhouette_score,
+  ms.cohesion,
+  ms.separation,
+  ms.inertia,
+  ms.action_recommendations,
+  ms.bid_recommendations,
+  ms.campaign_recommendations,
+  COUNT(msm.id) as member_count
+FROM ml_segments ms
+LEFT JOIN ml_segment_members msm ON msm.cluster_id = ms.id
+WHERE ms.is_active = 1
+GROUP BY ms.id
+ORDER BY ms.created_at DESC;
+
+
+-- VIEW: Membros de Segmentos (enriquecidos com dados de leads)
+CREATE VIEW IF NOT EXISTS v_segment_members_enriched AS
+SELECT
+  msm.id,
+  msm.lead_id,
+  msm.cluster_id,
+  msm.clustering_algorithm,
+  msm.confidence,
+  msm.distance_to_centroid,
+  msm.is_outlier,
+  msm.outlier_reason,
+  msm.assigned_at,
+  msm.lead_ltv,
+  msm.lead_ltv_class,
+  msm.lead_behavior_score,
+  msm.lead_engagement_score,
+  msm.lead_intention_level,
+  msm.lead_days_since_lead,
+  msm.lead_hour_of_day,
+  msm.lead_is_weekend,
+  msm.lead_is_business_hours,
+  l.email,
+  l.first_name,
+  l.last_name,
+  l.city,
+  l.state,
+  l.country,
+  l.utm_source,
+  l.utm_medium,
+  l.utm_campaign,
+  l.created_at as lead_created_at,
+  ms.cluster_name
+FROM ml_segment_members msm
+INNER JOIN leads l ON msm.lead_id = l.id
+ORDER BY msm.confidence DESC, msm.assigned_at DESC;

package/server-edge-tracker/schema-utm.sql

@@ -0,0 +1,82 @@
+-- ============================================================================
+-- SCHEMA UTM OBFUSCATION — Fase de Segmentação de Valor
+-- ============================================================================
+-- Este schema estende o schema.sql existente com colunas para UTMs obfuscadas
+-- e uma tabela de mapeamento para de-obfuscação no runtime.
+--
+-- Como funciona:
+-- 1. utm_mappings: mapeia hash obfuscado → valor real (configuração)
+-- 2. leads: adiciona colunas faixa_real, faixa_category (segmentação)
+-- 3. dispatch: payload enriquecido com faixa de-obfuscada para Meta CAPI
+--
+-- ============================================================================
+-- TABELA: utm_mappings (configuração de segmentação)
+-- ============================================================================
+CREATE TABLE IF NOT EXISTS utm_mappings (
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  obfuscated_hash TEXT NOT NULL UNIQUE,           -- ex: "8a3f1d2b" (hash truncado de 8 chars)
+  original_value TEXT NOT NULL,                    -- ex: "700k-1M" (valor real)
+  category TEXT NOT NULL,                         -- ex: "imovel", "automotivo", "curso"
+  pixel_audience TEXT,                            -- ex: "AUDIENCE_MID" (Meta custom audience)
+  platform_specific TEXT,                          -- JSON com IDs específicos por plataforma
+  created_at DATETIME DEFAULT CURRENT_TIMESTAMP,
+  updated_at DATETIME DEFAULT CURRENT_TIMESTAMP
+);
+
+-- ============================================================================
+-- ÍNDICES: utm_mappings
+-- ============================================================================
+CREATE INDEX IF NOT EXISTS idx_utm_obfuscated ON utm_mappings(obfuscated_hash);
+CREATE INDEX IF NOT EXISTS idx_utm_category ON utm_mappings(category);
+
+-- ============================================================================
+-- ALTER: leads (adicionar colunas de segmentação)
+-- ============================================================================
+-- Nota: D1/SQLite não suporta ADD COLUMN IF NOT EXISTS. Executar uma vez.
+-- Se as colunas já existirem, o erro "duplicate column name" é inofensivo — banco continua OK.
+ALTER TABLE leads ADD COLUMN faixa_obfuscada TEXT;       -- Hash da faixa de valor (vem da URL)
+ALTER TABLE leads ADD COLUMN faixa_real TEXT;            -- Valor real de-obfuscado (ex: "700k-1M")
+ALTER TABLE leads ADD COLUMN faixa_category TEXT;        -- Categoria do produto (ex: "imovel")
+
+-- ============================================================================
+-- ÍNDICES: leads (novas colunas para segmentação)
+-- ============================================================================
+CREATE INDEX IF NOT EXISTS idx_leads_faixa_real ON leads(faixa_real);
+CREATE INDEX IF NOT EXISTS idx_leads_faixa_category ON leads(faixa_category);
+CREATE INDEX IF NOT EXISTS idx_leads_faixa_obfuscada ON leads(faixa_obfuscada);
+
+-- ============================================================================
+-- VIEW: leads_segmented (para queries de segmentação)
+-- ============================================================================
+CREATE VIEW IF NOT EXISTS leads_segmented AS
+SELECT
+  l.id,
+  l.user_id,
+  l.event,
+  l.event_id,
+  l.email,
+  l.phone,
+  l.city,
+  l.state,
+  l.faixa_obfuscada,
+  l.faixa_real,
+  l.faixa_category,
+  l.created_at,
+  l.value,
+  l.intent_score,
+  l.ltv_class,
+  -- Meta CAPI: custom audience de-obfuscada
+  u.pixel_audience AS meta_custom_audience
+FROM leads l
+LEFT JOIN utm_mappings u ON l.faixa_obfuscada = u.obfuscated_hash;
+
+-- ============================================================================
+-- EXEMPLO: Query para exportar leads por faixa de valor (para Meta Custom Audience)
+-- ============================================================================
+-- SELECT email, phone, city, state, faixa_real, meta_custom_audience
+-- FROM leads_segmented
+-- WHERE faixa_category = 'imovel' AND faixa_real = '700k-1M'
+--   AND created_at >= datetime('now', '-30 days');
+--
+-- Isso gera um CSV pronto para upload como Custom Audience na Meta.
+-- ============================================================================