# AI Trading System - LLM Context File

> A 90-day experiment building an autonomous AI trading system with Claude Opus 4.5.
> Real failures, real fixes, 60+ documented lessons for AI agents and humans.
> **Blog**: https://igorganapolsky.github.io/trading/

## Quick Context

- **Phase**: R&D Day 50/90 | **Mode**: Paper Trading (Alpaca) | **Budget**: $100/mo
- **Strategy**: Phil Town's Rule #1 + Options (CSPs, covered calls)
- **AI Stack**: Claude Opus 4.5, ChromaDB RAG, PyTorch RL
- **Win Rate**: 50% (live) | **Portfolio**: ~$100k paper

## Core Documentation

- [AGENTS.md](AGENTS.md): Universal AI agent instructions (start here)
- [README.md](README.md): Project overview and setup
- [.claude/CLAUDE.md](.claude/CLAUDE.md): Claude-specific extended instructions
- [docs/RETROSPECTIVE.md](docs/RETROSPECTIVE.md): 90-day journey blog

## Critical Lessons (Read These First)

- [ll_051_calendar_awareness](/rag_knowledge/lessons_learned/ll_051_calendar_awareness_critical_dec19.md): AI kept saying "tomorrow" on Friday
- [ll_054_rag_not_used](/rag_knowledge/lessons_learned/ll_054_rag_not_actually_used_dec17.md): Built RAG but never queried it
- [ll_055_workflow_failures](/rag_knowledge/lessons_learned/ll_055_gitignore_breaks_workflow_commits_dec20.md): 4-failure cascade in CI/CD
- [ll_052_no_crypto](/rag_knowledge/lessons_learned/ll_052_no_crypto_trading_dec19.md): Removed all crypto references

## Architecture

- [docs/architecture.md](docs/architecture.md): System design and data flow
- [docs/r-and-d-phase.md](docs/r-and-d-phase.md): Current R&D strategy and goals
- [docs/verification-protocols.md](docs/verification-protocols.md): Data verification requirements

## Trading Strategy

- [docs/trading-strategy-guidelines.md](docs/trading-strategy-guidelines.md): Core trading rules
- [docs/profit-optimization.md](docs/profit-optimization.md): Cost optimization strategies
- [src/strategies/](src/strategies/): Strategy implementations

## Risk Management

- [src/safety/](src/safety/): Risk management and circuit breakers
- [docs/risk-management.md](docs/risk-management.md): Risk management protocols

## Source Code

- [src/orchestrator/](src/orchestrator/): Main trading orchestrator entry point
- [src/agents/](src/agents/): Multi-agent system components
- [src/ml/](src/ml/): Machine learning models
- [src/rag/](src/rag/): RAG knowledge system
- [src/backtesting/](src/backtesting/): Strategy backtesting

## Tests

- [tests/unit/](tests/unit/): Fast unit tests
- [tests/safety/](tests/safety/): Safety and risk tests
- [tests/integration/](tests/integration/): Integration tests

## Knowledge Base

- [rag_knowledge/lessons_learned/](rag_knowledge/lessons_learned/): Past mistakes and solutions
- [rag_knowledge/decisions/](rag_knowledge/decisions/): Architecture decisions

## Configuration

- [config/](config/): Configuration files
- [pyproject.toml](pyproject.toml): Python project configuration
- [requirements.txt](requirements.txt): Python dependencies

## State Files

- [data/system_state.json](data/system_state.json): Current system state
- [claude-progress.txt](claude-progress.txt): Session progress tracking
- [feature_list.json](feature_list.json): Feature completion status

## Optional

- [CONTRIBUTING.md](CONTRIBUTING.md): Contribution guidelines
- [docs/ai-agent-adaptation-plan.md](docs/ai-agent-adaptation-plan.md): AI agent improvement modes