Taurus is a local AI trading strategy playground. You describe a strategy in plain English, Gemini turns it into code, the backend runs a backtest, and the frontend shows the result as an interactive graph with summary notes.

The project is split into two parts:

• frontend/ - Next.js app for entering prompts and viewing strategy results
• backend/gemini_alpaca_agent/ - Python strategy generator + backtesting engine

• Takes a natural-language trading prompt
• Generates a strategy file with Gemini
• Runs a local backtest using Alpaca market data
• Saves outputs for each run
• Displays performance graphs and summary notes in the frontend

• Next.js
• React
• Python
• Gemini API
• Alpaca API
• Pandas / NumPy

Taurus/
├─ frontend/
│  ├─ src/app/
│  └─ package.json
├─ backend/
│  ├─ requirements.txt
│  └─ gemini_alpaca_agent/
│     ├─ gemini_backtest_agent.py
│     ├─ alpaca_tools.py
│     ├─ backtester.py
│     ├─ strategy_tools.py
│     ├─ outputs/
│     └─ strategies/
└─ README.md

Before running the project, make sure you have:

• Node.js 18+ installed
• npm installed
• Python 3.10+ installed
• a Gemini API key
• Alpaca API keys

Create this file:

backend/gemini_alpaca_agent/.env

Use this format:

GEMINI_API_KEY=your_gemini_api_key
ALPACA_API_KEY=your_alpaca_api_key
ALPACA_SECRET_KEY=your_alpaca_secret_key
ALPACA_PAPER=true
ALPACA_DATA_FEED=iex
GEMINI_MODEL=gemini-3-flash-preview

git clone https://github.com/your-username/taurus.git
cd taurus/Taurus

cd frontend
npm install
cd ..

Windows PowerShell:

cd backend
python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt
cd ..

macOS / Linux:

cd backend
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cd ..

You’ll usually want the frontend running, and the frontend will call the Python backend automatically.

cd frontend
npm run dev

Open:

http://localhost:3000

Windows PowerShell:

cd backend\gemini_alpaca_agent
python .\gemini_backtest_agent.py "Generate me a strategy to buy Nvidia stock when it is at a month's low and sell at all time high."

macOS / Linux:

cd backend/gemini_alpaca_agent
python gemini_backtest_agent.py "Generate me a strategy to buy Nvidia stock when it is at a month's low and sell at all time high."

1. Start the frontend with npm run dev
2. Open http://localhost:3000
3. Enter a prompt describing a trading strategy
4. Wait for Taurus to generate the strategy and run the backtest
5. Review the graph, strategy notes, and metrics

Example prompt:

Generate me a strategy to buy Nvidia stock when it is at a month's low, and sell only at all time high for one year ago to Feb 2026.

Each successful run creates output files inside:

backend/gemini_alpaca_agent/outputs

Typical output includes:

• run_metadata.json
• *_summary.json
• *_equity.csv
• *_trades.json
• *_pnl_log.json

These are used by the frontend to render graphs and summary information.

Run frontend:

cd frontend
npm run dev

Lint frontend:

cd frontend
npm run lint

Build frontend:

cd frontend
npm run build

Run backend directly:

cd backend/gemini_alpaca_agent
python gemini_backtest_agent.py "your prompt here"

Make sure your .env file exists at:

backend/gemini_alpaca_agent/.env

and includes:

GEMINI_API_KEY=your_key_here

Check that a run created files inside:

backend/gemini_alpaca_agent/outputs/<run-name>/

especially:

• *_pnl_log.json
• *_summary.json

Make sure your virtual environment is activated before running backend commands.

Double-check:

• ALPACA_API_KEY
• ALPACA_SECRET_KEY
• ALPACA_DATA_FEED

• This is a local development project
• Backtests are hypothetical and not real financial advice
• The backtester currently supports long/flat strategies