daily-signal-composite

name	daily-signal-composite
description	Run the Signal Aggregator to compute a unified 0–100 composite score per ETF by combining macro events, news sentiment, NAV Z-scores, FII/DII flows, ML predictions, and anomaly regimes. Use when user asks for ETF buy/sell recommendations, signal dashboard, or composite view.

Skill: Daily Signal Composite

Aggregates 6 signal sources into a unified 0–100 composite score per ETF with BUY/ACCUMULATE/HOLD/TRIM/AVOID actions.

Trigger

Use this skill when the user asks:

• "What should I buy/sell today?"
• "Which ETFs have the strongest signals?"
• "Run the signal aggregator"
• "Give me a composite view of all ETFs"
• "What does the signal dashboard show?"

Commands

# Compute and display
python src/main.py signals

# Compute + persist to ClickHouse (market_data.signal_composite)
python src/main.py signals --save

# Verbose (show per-source debug info)
python src/main.py signals --save -v

6 Signal Sources (Weighted)

Each source is a SignalSource subclass in src/agents/signal_sources.py (Strategy pattern). All 6 run in parallel via ThreadPoolExecutor. Adding a 7th pillar = 1 new class + 1 list append in run_signal_aggregation().

Source	Class	Weight	Data
Macro net signal	MacroSignalSource	25%	8-theme geopolitical → ETF directional impact sum
ETF news sentiment	SentimentSignalSource	15%	Positive/negative ratio from last 7 days in DB
NAV premium Z-score	ValuationSignalSource	15%	Discount = buy opportunity, premium = overpriced
FII/DII flow momentum	FlowSignalSource	25%	5-day institutional net flows
ML prediction	MLSignalSource	15%	LightGBM 5-day forward return (GOLDBEES)
Anomaly regime	GARCHAnomalySource	5%	Flash Crash → contrarian boost; Blow-off → dampening

All DB reads go through MarketDataRepository (src/db/repository.py) — no raw SQL in source classes.

Action Thresholds

Score Range	Action
≥ 75	BUY
60–74	ACCUMULATE
40–59	HOLD
25–39	TRIM
< 25	AVOID

18 ETFs Covered

GOLDBEES, NIFTYBEES, BANKBEES, ITBEES, JUNIORBEES, SILVERBEES, CPSEETF, LIQUIDBEES, LIQUIDCASE, GILT5YBEES, MON100, MAFANG, HNGSNGBEES, AUTOBEES, PHARMABEES, PSUBNKBEES, MID150BEES, SMALL250

DB Storage

Results saved to market_data.signal_composite table. View in Streamlit UI → 🎛️ Signals tab.

Source File

src/agents/signal_aggregator.py