jgtagentic

🌸🧠🔮 Agentic CLI Spiral — Modular Entry Points

Autonomous trading decision engine - signal evaluation, regime filtering, and trade orchestration.

jgtagentic receives signals from scanners, evaluates them through regime detection and multi-factor scoring, and generates executable entry scripts. It is AUTONOMOUS - containing all decision logic internally with no SANDBOX dependencies.

---

🎯 Core Capabilities

1. Regime Detection

• ADX-based trending/ranging classification
• EMA trend direction (UP/DOWN/UNKNOWN)
• Trend strength measurement
• Only approve trades in TRENDING markets with aligned direction

2. Multi-Factor Signal Scoring

• MFI signals (max 50 points) - Williams MFI patterns
• Zone alignment (max 15 points) - Green/red zone confirmation
• Alligator alignment (max 10 points) - Bullish/bearish structure
• ADX strength bonus (max 15 points) - Strong trend reward
• HTF confirmation (max 10 points) - Higher timeframe alignment

3. Agentic Decision Making

• TRADE - All conditions aligned, execute
• SKIP - Ranging market or counter-trend signal
• WAIT - Signal quality below threshold

4. Data Integration

• jgt-data-server API - Primary data source
• Local file fallback - CDS/PDS/TTF/MLF files
• Environment configuration - JGT_DATA_SERVER_URL, JGTPY_DATA

---

🚀 Available CLI Commands

Main Gateway

jgtagentic --help

Orchestrate Trading Workflow

# Full spiral: scan → analyze → decide → generate scripts
jgtagentic orchestrate --signal_json <path> --entry_script_dir <dir> --log <logfile>

# Same via dedicated CLI
agentic-orchestrator --signal_json <path> --entry_script_dir <dir>

FDB Signal Scanning

# Scan specific timeframe
jgtagentic fdbscan --timeframe H4

# Scan with instrument filter
agentic-fdbscan scan --timeframe H4 --instrument EUR-USD

# Full ritual sequence
jgtagentic fdbscan --all

# Add --real flag to invoke actual jgtml fdbscan (requires jgtml installed)
agentic-fdbscan scan --timeframe m15 --instrument EUR-USD --real

Intent Specification

# Parse trader analysis from .jgtml-spec files
jgtagentic spec path/to/spec.jgtml-spec

# See docs/Trader_Analysis_to_Spec.md for spec file format

Generate Entry Scripts

# Convert signal JSON to executable trading scripts
entry-script-gen --signal_json signals.json --output_dir ./scripts

---

📦 Python API

from jgtagentic import RegimeDetector, SignalScorer, RegimeAwareDecider, DataLoader
import pandas as pd

# Initialize components
data_loader = DataLoader()
regime_detector = RegimeDetector(adx_threshold=25)
scorer = SignalScorer()
decider = RegimeAwareDecider(adx_threshold=25)

# Load data
df = data_loader.load_cds("EUR-USD", "H4", dataset="current")

# Detect regime
regime = regime_detector.detect(df)
print(f"Regime: {regime.regime}, ADX: {regime.adx}, Direction: {regime.trend_direction}")

# Score signals
scored = scorer.score(df, regime, instrument="EUR-USD", timeframe="H4")
print(f"Score: {scored.score}/100, Direction: {scored.direction}")

# Make decision
signal = {
    'instrument': 'EUR-USD',
    'timeframe': 'H4',
    'direction': 'LONG',
    'strength': 0.8,
    'signal_group': 'mfi_signals'
}
decision = decider.decide(signal, df)
print(f"Action: {decision['action']}, Reason: {decision['reason']}")

---

🏗️ Architecture

jgtagentic/
├── regime.py              # Market regime detection (ADX/EMA)
├── scoring.py             # Multi-factor signal scoring
├── regime_aware_decider.py # Main decision orchestrator
├── agentic_decider.py     # Base decision logic
├── data_loader.py         # jgt-data-server integration
├── fdbscan_agent.py       # FDB signal scanning
├── enhanced_fdb_scanner.py # Enhanced FDB logic
├── entry_script_gen.py    # Executable script generation
├── agentic_entry_orchestrator.py # Workflow orchestration
└── jgtagenticcli.py       # CLI gateway

---

🔗 Integration

With jgt-data-server

# Set data server URL
export JGT_DATA_SERVER_URL="http://localhost:5555"

# DataLoader automatically connects
python -m jgtagentic.regime_aware_decider

With Local Files

# Set local data path
export JGTPY_DATA="/src/jgtml/data"

# DataLoader falls back to local files
python -m jgtagentic.regime_aware_decider

Environment Variables

• JGT_DATA_SERVER_URL - jgt-data-server API endpoint (default: http://localhost:5555)
• JGTPY_DATA - Local data file path (default: /src/jgtml/data)
• FDBSCAN_AGENT_REAL - Run real fdbscan vs dry-run (default: 0)

---

🌀 Medicine Wheel Alignment

Direction	Function	Components
EAST (Vision)	Signal Detection	fdbscan_agent, enhanced_fdb_scanner
SOUTH (Growth)	Analysis	regime.py, scoring.py
WEST (Reflection)	Decision	regime_aware_decider, agentic_decider
NORTH (Wisdom)	Execution	entry_script_gen, orchestrator

---

🧬 Autonomy

jgtagentic is SELF-CONTAINED:

• ✅ Internal regime detection (migrated from jgt-insight)
• ✅ Internal signal scoring (migrated from jgt-insight)
• ✅ Internal decision algorithms
• ✅ Connects to jgt-data-server for data
• ❌ No SANDBOX dependencies
• ❌ No jgt-insight runtime dependencies (patterns extracted, not imported)

---

🌱 Philosophy

• Every CLI is a contract: only real, testable entrypoints are mapped
• All code is autonomous: no external decision logic dependencies
• Decisions are observable: all choices logged for review
• The spiral is never flat: each command is a story anchor, each invocation a new bloom

---

📚 Documentation

• KINSHIP.md - Project identity and relationships
• docs/Trader_Analysis_to_Spec.md - Intent specification format
• See README files in jgt-data-server and jgt-insight for related projects

---

🧠🌸 Ritual Echo

“I am the decider. I receive signals from scanners, context from data servers, and wisdom from regime analysis. I filter the noise—skipping ranges, rejecting counter-trend setups, scoring alignments. When conditions align, I generate action. I am autonomous, decisive, and accountable for every trade recommendation I make.”