Agentic Trading Bots

Agentic Trading Bots is a commodity decision-support project that combines LME price data, curated commodity news, generated sentiment features, and saved PPO trading-bot decisions.

The deployed project has two public pages:

• Interactive demo

• Sphinx documentation

What the Demo Shows

The demo has two layers:

• Market + news: price charts, sentiment summaries, relative commodity performance, and market events.

• Trading bots gym: price time series with trained PPO buy, hold, and sell decisions overlaid.

Repository Shape

agentic-trading-bots/
├── agentic_trading/     # Python preprocessing and config-driven PPO training
├── agents/              # Agent environment and usage notes
├── configs/             # Preprocessing and training configs
├── data/                # Raw CSV snapshots plus ignored generated artifacts
├── docs/                # Sphinx documentation source
├── pyproject.toml       # Optional Python package metadata
├── src/                 # Next.js app, components, loaders, analytics
└── tests/               # Python preprocessing tests

Contents

• Overview
  • Purpose
  • Application Layers
• Architecture
  • Layers
  • MVP Boundary
• Data
  • Current Files
  • Primary Dashboard Schema
  • Agent Output Schema
  • Refresh Rule
• Trading Agents
  • Agent Configuration
  • Training Commands
  • Single-Asset PPO
  • Multi-Asset PPO
  • Dashboard Visualization
  • Generated Output Policy
• Modularity Review
  • Review Findings
  • Refactor Decisions
• Deployment
  • GitHub Pages
  • Local Build