Auto_Trader

Live paper-only RSI momentum rotation pipeline for NSE equities.

Report Bug · Request Feature

---

Table of Contents

• About
• Features
• Getting Started
  • Prerequisites
  • Installation
• Usage
• Roadmap
• Contributing
• License
• Contact
• Acknowledgments

---

About

Auto_Trader is the live paper-only RSI momentum rotation pipeline for NSE equities. It consumes yfinance-fed historical and intraday prices, generates champion-strategy signals, maintains a simulated ledger, and reports status through Hermes.

No real orders are placed. Zerodha Kite is retired and is not a supported price, login, or execution path; the Kite engine service is masked on the server.

Nightly strategy research lives in the separate Trader_Labs repository.

Features

• ✅ Paper-only RSI momentum rotation (no live orders)
• ✅ yfinance-fed prices (Tickertape REST fallback)
• ✅ Fail-closed signal publication
• ✅ Whole-share target deltas, idempotent signal IDs
• ✅ Atomic state commit before rebuildable output
• ✅ Telegram status delivery via Hermes cron
• 🚧 Live order execution (not supported)

Getting Started

Prerequisites

• Python 3.10+
• Access to NSE data (yfinance)

Installation

git clone https://github.com/The-Great-One/Auto_Trader.git
cd Auto_Trader
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Usage

Live data and execution flow

1. A Mac-side Hermes cron fetches NSE quotes with yfinance (Tickertape REST as fallback) and writes reports/live_prices.json on the server.
2. scripts/rsi_momentum_paper_shadow.py refreshes the paper signal at 10:15 on weekdays. It fails closed on stale/thin/malformed data.
3. scripts/rsi_momentum_paper_ledger.py marks positions to market every five minutes from 09:00–15:59, rebalancing when it sees a newer signal.
4. A Hermes cron reads the latest ledger output and delivers Telegram status.
5. Historical feathers are rebuilt on the Mac with yfinance and synced to the server.

Correct paper execution contract

The shadow publishes a versioned D-close intent with target weights (including inverse-volatility weights), target cash, a stable signal ID, and nullable modeled D+1 opens. The ledger fills whole-share target deltas only from one complete, timezone-aware, strictly fresh paper_quote_snapshot_v1. Fees use actual traded notional, residual cash and target-weight deviations are retained, and consumed signal IDs are idempotent.

Repository contents

• scripts/rsi_momentum_paper_shadow.py — champion signal generator.
• scripts/rsi_momentum_paper_ledger.py — simulated portfolio, MTM, rebalance.
• scripts/rsi_momentum_report.py, scripts/rsi_224466_rotation_lab.py — signal/backtest support.
• scripts/nightly_cleanup.py, scripts/prune_report_clutter.py — cleanup.
• tests/ — safety tests for shadow and ledger.

Generated reports, logs, market data, databases, and secrets are ignored by Git.

Running tests

pytest -q

Do not manually run the ledger or shadow against production state without an explicit operational reason and the shared ledger lock.

Deployment

cd /home/ubuntu/Auto_Trader
git pull --ff-only origin main

See PROJECT_MAP.md for schedules and operational ownership.

Roadmap

• v0.1 – Paper shadow + ledger
• v0.2 – Strategy promotion workflow
• v1.0 – Validated champion promotion

See the open issues for the full list of proposed features and known bugs.

Contributing

Contributions are welcome. Please read CONTRIBUTING.md and follow our Code of Conduct.

License

Distributed under the MIT License. See LICENSE for more information.

Contact

Sahil Goel – @The-Great-One – sahilgoel187@gmail.com

Project link: https://github.com/The-Great-One/Auto_Trader

Acknowledgments

• Img Shields
• Contributor Covenant