VIF Trading System - Setup Guide

✓ System Status

• Virtual Environment: Created and activated
• Dependencies: Installed (anthropic, yfinance, pandas, numpy, pyyaml)
• Watchlists: 3 pre-configured (vantage_portfolio, ai_verticals, energy_ai)
• Agents: Ready (watchlist_watcher.py, test_harness.py)
• Configuration: VIF framework config ready (config/vif_config.yml)

Quick Start

1. Activate Virtual Environment

cd c:/Users/marti/vif-trading-system
source venv/Scripts/activate  # Windows
# source venv/bin/activate      # Linux/Mac

2. Test Without API (No Credits Needed)

python agents/test_harness.py

Output: Mock VIF analysis showing the system structure. Results saved to reports/test_analysis.json.

3. Use Live Data (Requires API Credits)

First, add credits to your Anthropic account:

1. Go to https://console.anthropic.com/account/billing/overview
2. Purchase credits (even $5 allows hundreds of analyses)

Then run:

python agents/watchlist_watcher.py --watchlist energy_ai
python agents/watchlist_watcher.py --watchlist ai_verticals
python agents/watchlist_watcher.py --watchlist vantage_portfolio

4. Batch Analyze All Watchlists

python agents/watchlist_watcher.py --all

System Architecture

Three Components

1. Watchlist Parser (watchlist_watcher.py)

• Loads .txt files from watchlists/ directory
• Cleans and deduplicates tickers
• Strips exchange prefixes (NASDAQ:, NYSE:, etc.)

2. Data Fetcher (watchlist_watcher.py)

• Fetches OHLCV data from Yahoo Finance
• Computes RSI, MA, volume indicators
• Caches locally to reduce API calls
• Falls back to mock data if API unavailable

3. VIF Analyst (Claude API)

• Analyzes market data with VIF framework
• Outputs BUY/SELL/HOLD signals with confidence
• Identifies gamma regime, structural levels, kill switches
• Returns JSON with full reasoning

File Structure

vif-trading-system/
├── agents/
│   ├── watchlist_watcher.py      # Main agent (live data + Claude)
│   └── test_harness.py           # Local test (no API needed)
├── watchlists/
│   ├── vantage_portfolio.txt     # 85+ mixed holdings
│   ├── ai_verticals.txt          # 35+ AI/semiconductor
│   └── energy_ai.txt             # 13 energy+AI tickers
├── config/
│   └── vif_config.yml            # VIF parameters & kill switches
├── reports/
│   └── analysis_*.json           # Output reports (auto-generated)
├── data/
│   └── *.pkl                     # Cached market data (auto-generated)
├── .env                          # Your API key (KEEP SECRET)
├── .env.example                  # Template (safe to share)
└── venv/                         # Python virtual environment

Configuration

Edit config/vif_config.yml to customize:

VIF Thresholds:

vif_framework:
  gamma_regime:
    positive_threshold: 0.5       # Adjust gamma detection sensitivity
    negative_threshold: -0.5
  structural_levels:
    lookback_days: 20             # Historical window for support/resistance
  volume:
    ma_period: 20                 # Moving average period
    strong_threshold: 1.5         # 1.5x average = strong volume

Kill Switches (override conditions):

kill_switches:
  k1: "Extreme Volatility (RSI > 80 or < 20)"
  k2: "Gap Risk (5-day range > 10%)"
  k3: "Low Liquidity (Volume < 1M)"
  k4: "News Risk (earnings within 5 days)"
  k5: "Correlation Risk (>0.8 with index)"
  k6: "Technical Breakdown (close below MA)"

Data Fetching:

data_fetching:
  batch_size: 15                  # Tickers per Claude call (token control)
  period_default: "5d"            # Data lookback: 1d, 5d, 1mo, etc.
  cache_ttl_hours: 24             # Cache expiration

Token Budget

Per Watchlist Analysis:

• Input: ~500 tokens (market data summary)
• Output: ~1000 tokens (JSON signals)
• Total: ~1,500 tokens per watchlist

Pricing (Claude Opus 4.7):

• Input: $5 per 1M tokens
• Output: $25 per 1M tokens
• Cost per watchlist: ~$0.02-0.03

Usage Examples:

• Daily (3 watchlists): 45,000 tokens = $0.40/day
• 50 days of daily analysis: $20 budget

Troubleshooting

"Your credit balance is too low"

Solution: Add credits at https://console.anthropic.com/account/billing/overview

"No data for ticker X"

Solution: Yahoo Finance may not have data. Use test_harness.py instead or check ticker symbol.

"Failed to import anthropic"

Solution:

source venv/Scripts/activate
pip install anthropic

"UnicodeEncodeError"

Solution: Already fixed in test_harness.py. Update your script.

Next Steps

1. Add Anthropic API Credits → https://console.anthropic.com/account/billing/overview
2. Run Test: python agents/test_harness.py (verify system works)
3. Run Live: python agents/watchlist_watcher.py --watchlist energy_ai
4. Schedule: Set up daily cron/scheduler for automated analysis
5. Integrate: Export to Slack, email, or dashboard of choice

Advanced Usage

Custom Watchlist

1. Export from TradingView Desktop as .txt
2. Place in watchlists/ directory
3. Run: python agents/watchlist_watcher.py --watchlist mylist

Modify VIF Parameters

Edit config/vif_config.yml thresholds, then re-run analysis.

Analyze Historical Period

python agents/watchlist_watcher.py --watchlist energy_ai --period 1mo

Support

• Claude: Run /help to see available CLI tools
• GitHub: Check our examples in this repo
• Docs: See AGENTS.md and SKILLS.md