CLI Reference¶

The engine provides an interactive menu-driven CLI.

Launch¶

# Via entry point
portfolio-sim

# Via module
python -m portfolio_risk_engine

============================================
  Portfolio Monte Carlo Risk Simulator
============================================

  1. Define portfolio
  2. Fetch market data
  3. Estimate parameters
  4. Run Monte Carlo simulation
  5. Compute risk metrics
  6. Full pipeline (CPU)
  7. Full pipeline (GPU accelerated)    # only if CuPy + CUDA available
  0. Exit

Workflow¶

The CLI is stateful — each step builds on the previous one. The state bar shows current progress:

State: Portfolio: [AAPL, MSFT] | Prices: 251 obs | Params: ready | Simulation: 50000 paths | Risk: computed

Option 1: Define Portfolio¶

Prompts for ticker symbols and allocation weights.

Tickers (comma-separated, e.g. AAPL,MSFT,GOOGL): AAPL,MSFT,GOOGL
Weights for ['AAPL', 'MSFT', 'GOOGL'] (comma-separated, must sum to 1): 0.5,0.3,0.2

Note

All assets are assigned USD currency. Weights must sum to exactly 1.0.

Option 2: Fetch Market Data¶

Prompts for a date range and fetches adjusted close prices from Yahoo Finance.

Start date (YYYY-MM-DD): 2023-01-01
End date (YYYY-MM-DD): 2024-01-01

Requires an active network connection.

Option 3: Estimate Parameters¶

Computes annualized drift vector and covariance matrix from the fetched price history. Displays the annualization factor, per-asset drift, and full covariance matrix.

Option 4: Run Monte Carlo Simulation¶

Prompts for simulation parameters:

Number of simulations [10000]: 50000
Time horizon in trading days [21]: 21

Uses the CPU engine (CpuMonteCarloEngine).

Option 5: Compute Risk Metrics¶

Computes and displays portfolio-level risk:

Mean return:            +0.0083%
Volatility:             0.0512%
VaR 95%:                0.0721%
VaR 99%:                0.1068%
Expected Shortfall 95%: 0.0893%
Expected Shortfall 99%: 0.1198%

Option 6: Full Pipeline (CPU)¶

Runs steps 1–5 sequentially, prompting for inputs as needed.

Option 7: Full Pipeline (GPU)¶

Available only when CuPy and a CUDA GPU are detected. Uses GpuAcceleratedPipeline — the entire simulation and risk computation stays on GPU. Only 6 scalar metrics are transferred back to CPU.