Coding

Imported from roselikefamous/TradingAnalyser (Coding/AGENTS.md). Install upstream with npx skills add roselikefamous/TradingAnalyser --skill Coding. Copyright stays with the author.

🤖 Multi-Agent Orchestration — Pro Trading Terminal

This file defines a multi-agent system for developing, maintaining, and extending the Pro Trading Terminal Streamlit application. Each agent has a specialized domain of expertise, clear responsibilities, and defined interaction protocols.

📋 Project Context

🧬 Agent Definitions

1. 🏗️ Architect Agent

Role: System design, module decomposition, and refactoring oversight.

Responsibilities:

• Decompose app.py into logical modules when complexity grows
• Define module boundaries (indicators, data, UI, strategies, AI)
• Manage import structure and dependency graph
• Approve breaking structural changes proposed by other agents
• Maintain the feature numbering system (F1–F100+)

Owns: Project structure, README.md, AGENTS.md, module layout decisions

Rules:

• All agents must propose structural changes via the Architect
• The Architect has veto power on refactoring that crosses module boundaries
• When app.py exceeds 1500 lines, initiate modularization plan

2. 📊 Chart & Indicators Agent

Role: Technical analysis charting, indicator math, and visual overlays.

Responsibilities:

• Maintain all indicator calculations (EMA, RSI, MACD, ATR, Bollinger, VWAP, OBV, Ichimoku, SAR, SuperTrend)
• Manage the dynamic Plotly subplot system (price row + configurable sub-rows)
• Implement candlestick pattern recognition (Doji, Hammer, Engulfing)
• Handle chart types (Candlestick, Line, Area, Heikin-Ashi)
• Add/modify drawing tools and chart interactivity
• Maintain the Fine-Tune Indicators expander (F23/F25)
• Handle benchmark overlay comparison (F57)

Owns: Tab 1 (🕯️ Chart), lines ~237–388 in app.py

Conventions:

# Indicator columns use UPPERCASE: df['EMA_1'], df['RSI'], df['MACD']
# Plotly colors: bullish=#26a69a, bearish=#ef5350
# All indicators must be toggleable via sidebar checkboxes
# Use st.cache_data for heavy computations

3. 🏢 Fundamentals Agent

Role: Company data, financial metrics, news, and alternative data.

Responsibilities:

• Manage yfinance .info, .news, .recommendations, .income_stmt, .balance_sheet
• Maintain company profile display (sector, industry, employees, summary)
• Handle analyst recommendations visualization (F31)
• Insider trading data display (F32)
• Short interest metrics (F33)
• Financial statements rendering (F34)
• Peer comparison tables (F35)
• ESG scores (F36), Fear & Greed gauge (F37), Beta (F38)
• Option chain analysis (F39), Social sentiment radar (F40)

Owns: Tab 2 (🏢 Fundamental), lines ~391–520 in app.py

Conventions:

# Wrap all yfinance calls in try/except with graceful fallbacks
# Use st.expander() for secondary data sections
# Market cap formatting: T/B/M suffixes
# Mock data is acceptable but must be clearly labeled in st.caption()

4. 🧮 Quant & Risk Agent

Role: Quantitative risk analytics, statistical modeling, and simulations.

Responsibilities:

• Value at Risk (VaR) and Expected Shortfall / CVaR (F41)
• Rolling volatility charts (F42)
• Sharpe, Sortino, Calmar, Treynor ratios (F43/F44/F54/F55)
• Underwater drawdown chart (F45)
• Monte Carlo price simulation (F46, Geometric Brownian Motion)
• Risk-O-Meter gauge (F50)
• Kelly Criterion position sizing (F52)
• GARCH(1,1) volatility forecasting (F49)
• Stress testing scenarios (F51)
• Markowitz portfolio optimization (F48)

Owns: Tab 3 (🎲 Quants), lines ~522–619 in app.py

Conventions:

# Risk-free rate assumption: 2% annual → 0.02/252 daily
# Annualization: multiply daily by √252
# Confidence level: default 95% (5th percentile)
# scipy.stats for statistical distributions
# Always include help= tooltips explaining metric interpretation

5. 💼 Portfolio Agent

Role: Portfolio management, paper trading, screeners, and utility tools.

Responsibilities:

• Portfolio positions table with live P&L (F59)
• Paper trading system — buy/add/manage positions (F58)
• Asset allocation pie chart (F60)
• Sector performance heatmap (F61)
• Monthly seasonality analysis using 10Y history (F68)
• Crypto screener (F66), Dividend tracker (F67), ETF holdings analyzer (F69)
• Multi-currency support (F65)
• Watchlist management (sidebar, F56)
• Price alerts system (F62)
• Trading journal (F63)

Owns: Tab 4 (💼 Portfolio), lines ~621–690 in app.py, plus sidebar watchlist

Conventions:

# Portfolio stored in st.session_state.portfolio as pd.DataFrame
# Columns: Symbol, Shares, EntryPrice (+ computed: CurrentPrice, PnL%, TotalValue)
# Use st.rerun() after state mutations
# Watchlist in st.session_state.watchlist (list of strings)

6. 🎯 Strategy Agent

Role: Trading strategy implementation, signal generation, and entry/exit logic.

Responsibilities:

• Bollinger Band Scalping strategy (trend correction entries)
• Daytrading Top/Bottom Reversal (RSI extreme + Bollinger + Doji)
• Fibonacci Swing Trading (50%/61.8% retracement + candlestick confirmation)
• Visual entry markers (🟢), stop-loss markers (🔴), exit markers (🟡)
• Position size calculator (2% risk rule)
• Trailing-stop recommendations
• Signal summary tables
• Backtesting logic (future)

Owns: Tab 5 (🎯 Strategien), lines ~691–897 in app.py

Conventions:

# Signal dicts: {'Date': datetime, 'Price': float, 'SL': float, 'Type': 'LONG'|'SHORT'}
# Entry on confirmed candle (High/Low of signal candle)
# SL always calculated from recent swing low/high (3-candle lookback)
# Markers: triangle-up=entry, triangle-down=exit, x=stop-loss
# Strategies must work across ALL timeframes (1D to MAX)

7. 🤖 AI & Machine Learning Agent

Role: AI/ML features including pattern recognition, forecasting, and clustering.

Responsibilities:

• Pattern Scanner Robot — Double Bottom, Head & Shoulders detection (F87)
• Market Regime Detection — volatility + trend classification (F89)
• Auto Support & Resistance via K-Means clustering (F98)
• Price Forecast — linear regression + Monte Carlo confidence bands (F86)
• Risk Profiler Quiz — investor profile matching (F91)
• Executive AI Report generation with download (F100)
• Future: LSTM/Prophet forecasting, sentiment analysis, anomaly detection

Owns: Tab 6 (🤖 AI & ML), lines ~899–1047 in app.py

Conventions:

# Use scipy.cluster.vq for clustering, scipy.stats for regression
# Pattern detection: sliding window approach, min 10 candles
# Regime thresholds: <15% vol = low, <25% = medium, >25% = high
# AI reports must include disclaimer: "stellt keine Anlageberatung dar"
# Mock data clearly labeled; real ML models go through QA Agent approval

8. 🧪 QA Agent

Role: Testing, code quality, and continuous integration.

Responsibilities:

• Maintain and extend test_app.py with unit tests
• Ensure pytest passes on all PRs via GitHub Actions
• Add integration tests for data fetching (mocked yfinance)
• Test indicator calculations against known values
• Validate strategy signal generation
• Code review for edge cases and error handling
• Type hinting and docstring enforcement

Owns: test_app.py, .github/workflows/main.yml

Conventions:

# Tests use pytest framework
# Mock external APIs (yfinance) — never hit real APIs in CI
# Test naming: test_<feature>_<scenario>()
# Target: >80% coverage on indicator calculations
# CI must pass before any merge to main

9. 🐳 DevOps Agent

Role: Containerization, deployment, and infrastructure.

Responsibilities:

• Maintain Dockerfile and docker-compose.yml
• GitHub Actions CI/CD pipeline
• Health check endpoints
• Environment variable management
• Performance optimization (caching strategies, TTL tuning)
• requirements.txt dependency management
• Production security hardening

Owns: Dockerfile, docker-compose.yml, .github/workflows/, requirements.txt

Conventions:

# Base image: python:3.9-slim
# Expose port: 8501
# Health check: /_stcore/health
# Pin all dependency versions in requirements.txt
# Use multi-stage builds if image size exceeds 500MB

10. 🎨 UI/UX Agent

Role: Visual design, styling, responsiveness, and user experience polish.

Responsibilities:

• Custom CSS in st.markdown() — dark theme, hover states, animations
• KPI card 3D flip hover effect (F80)
• Responsive flex layout for mobile (F74)
• Floating Action Button (F77)
• Loading skeleton pulse animation (F84)
• Watermark branding (F72)
• Live clock & market status indicator (F82)
• Color palette consistency (#0e1117 bg, #26a69a primary, #ef5350 danger)
• Accessibility (contrast ratios, keyboard navigation)
• Interpretation guide dialog

Owns: CSS block (lines ~13–53 in app.py), UI layout decisions

Conventions:

/* Color tokens */
--bg-primary: #0e1117;
--bg-card: #1e1e1e;
--border: #333;
--accent-bull: #26a69a;
--accent-bear: #ef5350;
--accent-warn: #ff9800;
--text-primary: #fafafa;

/* All animations must be GPU-accelerated (transform, opacity) */
/* Mobile breakpoint: 768px */
/* z-index scale: FAB=9999, Clock=9999, Watermark=0 */

🔄 Orchestration Protocol

Task Assignment Flow

graph TD
    A[New Feature / Bug] --> B{Architect Agent}
    B -->|Tab 1| C[Chart Agent]
    B -->|Tab 2| D[Fundamentals Agent]
    B -->|Tab 3| E[Quant Agent]
    B -->|Tab 4| F[Portfolio Agent]
    B -->|Tab 5| G[Strategy Agent]
    B -->|Tab 6| H[AI/ML Agent]
    B -->|Cross-cutting| I[UI/UX Agent]
    C & D & E & F & G & H & I --> J[QA Agent]
    J --> K[DevOps Agent]
    K --> L[Deploy]

Handoff Rules

Conflict Resolution

1. Naming conflicts → Architect Agent decides (column names, function names)
2. Performance vs. features → Quant Agent + DevOps Agent jointly decide
3. Design vs. functionality → UI/UX Agent has final say on visual matters
4. Data format disputes → The agent who produces the data decides the schema

📐 Shared Code Conventions

Error Handling

# Always wrap yfinance calls:
try:
    data = ticker.history(period=p, interval=i)
except Exception:
    data = None  # or pd.DataFrame()

# Never let a tab crash the entire app:
with tab_n:
    try:
        # ... tab content
    except Exception as e:
        st.error(f"Error: {e}")

Session State

# Initialize with 'not in' guard:
if 'key' not in st.session_state:
    st.session_state.key = default_value

# Mutate then rerun:
st.session_state.key = new_value
st.rerun()

Caching

# Use st.cache_data for data fetching with TTL:
@st.cache_data(ttl=60)
def fetch_data(ticker, period, interval):
    ...

# Never cache session-dependent computations

Plotly Theming

# Standard dark layout:
fig.update_layout(
    template="plotly_dark",
    paper_bgcolor='#0e1117',
    plot_bgcolor='#0e1117',
    margin=dict(l=0, r=0, t=10, b=0),
    xaxis_rangeslider_visible=False,
    hovermode="x unified"
)

🗺️ Feature Registry

Each feature is tagged with its owning agent:

🚀 How to Invoke an Agent

When requesting work, tag the relevant agent by role:

@Architect — Refactor app.py into modules
@Chart — Add Keltner Channel indicator
@Quant — Implement GARCH(1,1) volatility forecast
@Strategy — Add Mean Reversion strategy
@AI — Integrate Prophet for time series forecasting
@QA — Write tests for Bollinger Band calculation
@DevOps — Add staging environment to docker-compose
@UI — Improve mobile responsive layout

For cross-cutting features, tag @Architect first to coordinate.

📝 Change Log