AI desk note
Groq narrative overlay — levels from Parkhu gates
Guide
Shows: optional Groq stance, brief, and per-symbol actions. Entry / stop / T1 / hold are stamped from the deterministic pack.
Decide: use as process commentary. Claude still owns capital and final sizing. If skipped, trust regime + ideas alone.
How stocks get filtered
Each step removes names that fail a hard rule
Guide
Shows: top → bottom how many symbols survive each hard filter, with keep-% vs the prior step. Tap ? on a step for what it means and why it applies.
Decide: early collapse → weak breadth, fewer new ideas. Late drops (earnings) → process working; don’t loosen filters.
Survivors
Final gate passers — why selected or not
Guide
Shows: top 50 names left after all hard filters, with status (idea / watchlist / rejected) and why they were not picked when rejected.
Decide: rejected for R:R or score is process working; portfolio/sector caps mean the book is full — not a loose-gate signal.
Regime
Indices, VIX, flows, global cues
Guide
Shows: Nifty/BankNifty, India VIX, FII/DII, Asia/Europe/US, crude, USDINR.
Decide: bearish + high VIX / FII selling → cut new risk. Bullish + supportive flows → act on high-score ideas. Mixed → wait.
Ideas
Cleared gates — Claude shortlist (no sizing)
Guide
Shows: score, band, entry / stop / T1, R:R. No qty or capital.
Decide: prefer higher score + R:R ≥ 2 + uncrowded sector. Watch = monitor only. Zero ideas is valid.
Open ledger
Live suggestions — MFE / MAE / P&L%
Guide
Shows: suggestion ledger as of the selected date (from trades CSV) — entry vs last, P&L%, MFE, MAE. Empty before the first tracked idea is normal.
Decide: needs action → exit/reduce before new risk. Deep MAE → respect stops. Strong MFE, faded P&L → trail / partial.
Sectors
Name counts only — not capital %
Guide
Shows: idea + open names per risk sector.
Decide: avoid stacking into a sector that already dominates the open book.