PAPER TRADING MODE — no real orders placed. All trades are simulated.

Kite not connected — paste request_token in Kite Login box to enable trading

Market holiday — no scanning today

Parameter	Value
Pipeline
Scan Interval	2 min
Max Candidates	15
Min Widget Score	5.0
Data Source	yfinance (fallback — Kite not connected)
Ticker Mode	full
Risk
Capital	100,000 INR
Risk per Trade	1.0%
Max Order Qty	500
Max Position Value	200,000 INR
Max Drawdown	3.0%
Entry Validation
Entry Zone (Upper)	1.5%
Entry Zone (Lower)	0.3%
Exit Agent
Poll Interval	30s
T1 Partial Exit	50.0%
Breakeven Buffer	0.1%
SL Order Type	SL (limit at trigger - 1%)
Entry Fill Timeout	120s (entry buffer: CMP + 0.3%)
Trailing SL
Enabled	Yes
ATR Multiplier	1.5x (trail at high - N*ATR)
Circuit Detection
Status	Active (via WebSocket MODE_FULL)
Frozen Stocks	0
Auto-Calibration
Enabled	No
Min Trades	10 (before calibrating)
Disable Phase Below	20.0% win rate
Raise Threshold Below	35.0% win rate (+5.0)
Lower Threshold Above	60.0% win rate (-3.0)
Widget Score Raise	If 5-7 band < 30.0% → min=7
Status	Not yet calibrated
P&L Enforcement
Protect Gains Above	2,000 INR (+5.0 to min_score)
Losing Day Below	-2,000 INR (0.5x size)
Min Trades Before	3 (no enforcement in first 2 trades)
Human-in-the-Loop
Mode	manual
Auto Threshold	≥ 85.0 (semi_auto only)
Signal TTL	5 min (auto-expires if no action)
Max Up Move	+1.5% (chasing limit)
Max Down Move	-0.3% (failure limit, stricter)
On Approve	Re-fetch LTP → recalculate Entry/SL/T1/T2 → re-validate → execute
Pending Approvals	0
Rollout	Manual (wk1-2) → Semi-auto (wk3-4) → Auto (month2+)
WebSocket Exits
Status	Fallback to REST polling
Latency	~30s (REST poll)
Cooldown
Cooldown TTL	60 min
HOD Bypass	Enabled
Relative Strength
RS Enabled	No
RS Threshold	> 0.0
Nifty Symbol	^NSEI
Sector Penalty
Enabled	No
Min Advance %	10.0%
Penalty Range	-3.0 to -8.0
Active Phases	MORNING,MID_MORNING
Anti-Martingale
Enabled	Yes
Halve After	2 consecutive SL
Pause After	3 consecutive SL
Pause Duration	60 min
Size Multiplier	0.5x
Market Phases
MORNING (09:20-10:30)	score≥65.0 | max 3 pos
MID_MORNING (10:30-12:00)	score≥65.0 | max 3 pos
LUNCH (12:00-13:30)	score≥65.0 | max 1 pos (queue for approval)
AFTERNOON (13:30-14:30)	score≥65.0 | max 3 pos
EXIT_ONLY (14:30-15:10)	No entries — manage exits + force square-off
Scoring & Exit Models
BQS (Breakout Quality Score)	Disabled (legacy scoring) — Momentum + Room + Volume + Widget
ATE (Adaptive Trail Exit)	Disabled (fixed T1/T2) — 4-phase: Protect → Breakeven → Trail → Time
A/B Logging	Both BQS and legacy scores logged in parallel for comparison

Scout Agent v1.2

• Scrapes Chartink dashboard via Playwright (headless Chromium)
• Intercepts /widget/process API + reads Vue component state
• Extracts: symbol, CMP, volume, widget name, sector advances, industry counts
• Volume/CMP disambiguation: detects fake volume columns (decimal = price, not volume)

Filterer Agent v1.3

• Deduplicates stocks across widgets
• Multi-widget category scoring: Pattern=4, Momentum=3, Volume=3, Trend=2
• Partial credit: same-category widgets get +1 (not +0) — e.g., 12/26 + 9/21 = 3+1=4
• Pattern bypass: BO/VCP widgets pass at score 4+ (bypasses min 5)
• Sector bonus: +2.0 for stocks in sectors advancing > 10.0%
• Volume filter: skips illiquid stocks (volume < 10,000)
• Cooldown exclusion + HOD bypass
• Max candidates: 15

Analyst Agent v4.0

• EMA(12/26) crossover detection
• Crossover timing: Fresh crossover BUY at score 60+, already-above BUY at score 65+
• Gradual scoring: RSI ramp 40→75 = 0→20 pts, Volume ramp 0.5→3x = 0→20 pts
• Volume Momentum Override: +10 pts when Vol>2.5x + RSI>75 (momentum, not overbought)
• ATR(14) for SL/target computation
• VCP (Volatility Contraction Pattern) tightness ratio
• 5 candlestick patterns: bullish engulfing, marubozu, hammer, three white soldiers, breakout candle
• Support/resistance: swing high/low detection on recent candles
• Sector correlation penalty: -3.0 to -8.0 if sector advance < 10.0%
• Composite scoring (0-100) with widget score + pattern bonus + sector penalty + momentum override

BQS — Breakout Quality Score v1.0

• Status: Disabled (A/B logging only)
• Momentum Speed (30 pts): RSI acceleration + EMA slope + volume surge
• Room to Run (25 pts): distance to resistance + ATR-based headroom
• Volume Confirmation (25 pts): volume ratio vs 20-day avg + buying pressure
• Widget Confluence (20 pts): multi-widget agreement across categories
• A/B: both BQS and legacy scores logged for every stock analysis

Signal Generator v2.0

• Entry: CMP within zone (0.3% lower / 1.5% upper)
• SL: tighter of ATR-based (1.5x ATR) or support-based
• T1: 1:1 R:R — book 50% | Minimum T1: 0.7%
• T2: 1:2 R:R — exit remaining | No cap (human decides on approval)
• RS filter: DISABLED
• Rejects: SL too tight (<0.2%), T1 too small (<0.7%), score below phase threshold

Risk Agent v1.2

• Position sizing: 1.0% capital risk per trade
• Hard limits: max 500 qty, 200,000 INR value, 3.0% drawdown
• Minimum R:R 1.5 enforced
• Phase-aware position caps
• Anti-martingale: 0.5x size after 2 SL, pause 60min after 3 SL

Exit Agent v4.0

• Entry fill verification: waits for COMPLETE, handles partial fills, cancels unfilled after 120s
• SL orders: SL with limit price (trigger - 1%), tick-rounded to 0.05. 3x retry with 2s delay. If all 3 fail → emergency market exit.
• SL modify: Single API call (no cancel+replace gap). Falls back to cancel+replace with 3x retry if modify rejected.
• Entry orders: LIMIT at CMP + 0.3% buffer, tick-rounded to 0.05
• Entry cancelled/rejected: orphaned SL order auto-cancelled (prevents phantom sells)
• All prices: Entry, SL, T1, T2 tick-rounded to 0.05 grid at signal generation AND recalculation
• SL status handling: CANCELLED → re-place, REJECTED → re-place or market exit
• Position persistence: saved to DB BEFORE SL placement (crash recovery)
• T1 hit: sell 50%, cancel old SL, place new SL at breakeven + 0.1%
• T2 hit: sell remaining, cancel SL, record exit
• ATE (Adaptive Trail Exit): Disabled (fixed T1/T2)
  1. Phase 1 PROTECT: SL = Entry - 2×ATR
  2. Phase 2 BREAKEVEN: SL → Entry + 0.1% (after +1% gain)
  3. Phase 3 TRAIL: SL = High - 1.5×ATR (after +2% gain)
  4. Phase 4 TIME: SL = High - 1×ATR (after 14:30 IST)
• Trailing SL (legacy): after T1, trail at highest_price - 1.5×ATR. Only moves up.
• Position reconciliation: compares system positions with Kite every 30s. Detects manual exits + orphans.
• Approval dedup: same symbol replaces existing PENDING approval (upsert, not duplicate)
• WebSocket exits: T1/T2 thresholds registered with ticker for sub-second detection. Falls back to 30s REST polling.
• Circuit detection: skips exit actions for stocks at upper/lower circuit. Resumes when released.
• Batch LTP (single API call for all positions per poll)
• EOD: cancel all SL → market sell all → record exits → reconcile
• Uses actual fill prices from Kite (average_price) for P&L accuracy

Market Phase Engine v1.0

• 9 phases: PRE_OPEN, PRE_MARKET_SCAN, WARMUP, MORNING, MID_MORNING, LUNCH, AFTERNOON, EXIT_ONLY, CLOSED
• PRE_MARKET_SCAN (09:09): Chartink scanning begins (no trades)
• WARMUP: captures Nifty 50 open price (also captured on login + pipeline failsafe)
• Dynamic score thresholds and position caps per phase
• LUNCH: scan + queue for approval (max 1 position, human decides in manual mode)
• EXIT_ONLY: cancel SL + force square-off + reconciliation

Data Fetcher v2.1

• Priority cascade: WebSocket Ticker → Kite REST → yfinance
• yfinance fallback enables full pipeline without Kite auth
• Rate-limited Kite REST (3 req/sec via semaphore)
• Nifty 50 tracking for intraday RS calculation

Auto-Calibration Engine v1.0

• Computes win rates from actual trade outcomes (min 10 trades required)
• Phase thresholds: disable if win rate < 20.0%, raise +5.0 if < 35.0%, lower -3.0 if > 60.0%
• Widget score: raise min from 5 to 7 if score 5-7 band win rate < 30.0%
• Overrides are HARD (not soft suggestions) — applied directly to phase_config
• Logged to calibration_log table for audit trail
• Manual trigger: POST /calibrate

P&L Enforcement v1.0

• Protect gains: if P&L > 2,000 after 3+ trades → raise min_score by 5.0 (only take high-conviction trades)
• Limit losses: if P&L < -2,000 after 3+ trades → halve position size (0.5x)
• No enforcement in first 2 trades (too early to judge the day)
• HARD enforcement — not advisory, directly applied to execution

Trade Journaling v1.0

• Every approval card has a "Why?" input field
• Reason saved to DB on both APPROVE and REJECT
• Captures human reasoning for post-hoc analysis
• Enables learning: "WHY did I approve trades that lost?" / "WHY did I skip winners?"

Human-in-the-Loop v1.1

• Pipeline change: Signal → Queue (PENDING) → Dashboard card → User APPROVE/REJECT → Execute
• Manual mode (current): every signal requires human approval before execution
• Semi-auto: auto-execute if score ≥ 85.0, queue the rest
• Auto: full autonomous, no queue (only after proven track record)
• On APPROVE (Option D Hybrid):
  1. Re-fetch live price from Kite/yfinance
  2. Directional validation: reject if price moved +1.5% up (chasing) or -0.3% down (failed breakout)
  3. Recalculate Entry/SL/T1/T2 from live price using original ATR
  4. Re-validate: SL tightness (>0.2%), T1 (>0.7%)
  5. If all pass → place order with recalculated levels
  6. If stale → show "setup no longer valid"
• TTL: 5 min — no action = auto-expires
• Security: approve/reject requires API key auth + pipeline lock (prevents duplicate orders)
• Rollout plan: Manual (wk1-2) → Semi-auto (wk3-4) → Auto (month2+)
• All decisions (approved/rejected/expired) logged to DB for performance review
• Current mode: manual

Watchdog v1.0

• Heartbeat on every successful pipeline/warmup cycle
• Tracks consecutive failures, alerts after 3
• Stale detection: heartbeat > 5 min old = unhealthy
• POST /reconcile for manual post-market check