Crypto Trader Bot with AI (v1.05)

Overview

This project combines machine-learning price forecasting with automated trade execution and conversational control surfaces. It supports:

• Crypto trading with Binance market data, Capital.com execution, and a Telegram assistant.
• Index option trading on Zerodha Kite via a modular, agent-driven orchestrator.

Both tracks share a focus on explainable automation, configurable risk controls, and paper/live mode flexibility.

Key Features

• AI Forecasting: PyTorch LSTM predicts directional moves for ETH/SOL/BTC using recent Binance OHLC data.
• Human-in-the-loop Execution: Telegram inline keyboards require explicit confirmation before every order; auto-close prompts are issued when counter signals appear.
• Broker Integrations: Capital.com (crypto) and Zerodha Kite (index options) connectors with dry-run fallbacks.
• Strategy Intelligence Layer: Sentiment analysis, market-context tagging, and optional RAG-augmented LangGraph agent choose the intraday playbook.
• Trade Logging & Reporting: SQLite persistence, backtesting hooks, and scheduled reporting stubs keep performance auditable.

Architecture Snapshot

• main.py — End-to-end crypto bot: loads model, fetches Binance data, produces signals, mediates Telegram UX, and dispatches to Capital.com.
• trading_bot.py — Zerodha orchestrator that authenticates, selects strategies, and coordinates order and position agents.
• agents.py — Order execution (isolated worker pattern) and position management (stop-loss, trailing-stop logic).
• strategy_factory.py, indicators.py, indicator_calculator.py — Technical indicator calculators and strategy registry.
• langgraph_agent.py, sentiment_agent.py, market_context.py — Market-intel inputs for the LangGraph-driven strategy selector.
• backtester.py, reporting.py — Offline evaluation and reporting utilities.
• state.json, output/ (if present) — Persisted runtime state and generated reports/backtests.

Some modules referenced in the codebase (e.g., rag_service.py) are optional or may be supplied privately. Stub them if you plan to run the full workflow.

Setup

1. Python Environment
   • pip install -r requirements.txt
   • GPU support is optional; CPU works for inference.
2. Secrets
   • Create a .env file holding Binance, Capital.com, and Telegram credentials for main.py.
   • Populate config.yaml with Zerodha keys, trading flags, and strategy settings for trading_bot.py.
3. Model Artifact
   • Place price_predictor.pt in the project root or retrain/export using your own pipeline.
4. Database
   • The crypto bot autogenerates trades.db; ensure the process has write access.

Telegram Workflow (Crypto Bot)

• /show — Run AI analysis and receive BUY/SELL/HOLD suggestions with inline actions.
• /status — Summarize the most recent open position and P&L.
• /close — Close the active trade (requires stored deal_id).
• /capital — Display Capital.com connection diagnostics.
• /symbols — List common ETH epic formats for Capital.com.
• /test — Run a health check across data, model, broker, and Telegram subsystems.

Configuration Highlights

Default trading flags (editable via .env or config.yaml):

• Underlying instrument, timeframe, and lot sizing.
• Risk per trade, stop-loss thresholds, and maximum trades per day.
• Paper trading toggles and natural-language prompt overrides.
• RAG controls (minimum trading days, enable/disable retrieval augmentation).

Safety & Disclaimer

Algorithmic trading carries significant financial risk. Use paper/demo modes until you validate strategies, confirm broker credentials, and stress test failure paths. The authors assume no responsibility for trading outcomes.

Developer: @Perfect