Pipeline de 11 estágios + AgentRuns

Estágios do Pipeline

1. 1
   RIGO · Scan

   Puxa mercados ativos da Gamma API (public, no key), normaliza, salva snapshots append-only immutable
2. 2
   RIGO · Filter

   Liquidez ≥$500, horas>48, não fechados, spread<20%, cache Redis TTL 5min
3. 3
   HOLT

   Pesquisa YES-biased, 5 fontes, Anthropic JSON mode, fallback heurístico ±5% cap (RESEARCH_CAP_PROBABILITY)
4. 4
   BRAM

   Pesquisa independent NO/contrarian, blindado (NÃO importa HOLT), anti-consensus pull 5%
5. 5
   ILSA

   Pool logit: 0.70·mercado (k=1.0 até calibrar) + 0.15·HOLT + 0.15·BRAM, momentum 0.25×move ±2pp, |q−p|≤5pp
6. 6
   KETT

   Kelly fracionado 0.25 × bancária (running_max_equity), PDF max 6% per position = $3.00 cap on $50 start
7. 7
   Risk Gate

   13 regras determinísticas EMERGENCY_STOP → EDGE → CONF → LIQ → TIME → STALE → POSITION → DUPLICATE → CORRELATION → EXPOSURE → DAILY LOSS → DRAWDOWN → ALL_CLEAR FIRE
8. 8
   RIGO Exec

   Ordem PAPER apenas (RuntimeError se trading_mode!=paper). Idempotência UNIQUE(pipeline_run_id, market_id) + SHA256 idempotency_key
9. 9
   TESS

   Mark-to-market Positions, 13 Survival metrics (ROI, WinRate, PF, DD, Runway), PortfolioSnapshot UNIQUE pipeline_run_id
10. 10
    Portfolio Mark (5min)

    Cada 300s, puxa preços atualizados → fallback DB MarketModel.source_id → mark_position(TESS) + PortfolioSnapshot
11. 11
    Settlements (60min)

    Cada 3600s, itera posições abertas → MarketModel resolved → settle_position(TESS) → Settlement UNIQUE position FK → MARKET_RESOLVED event