Alchemist is a high-performance, distributed automated trading system designed to seamlessly transition quantitative research into fully automated algorithmic trading strategies.
Built on Ray, Alchemist provides a scalable infrastructure for strategy execution, real-time order management, and seamless integration with market data and broker connections. Whether you’re a quantitative researcher, algorithmic trader, or hedge fund, Alchemist helps you deploy, manage, and scale trading strategies efficiently.
🚀 Fast and Scalable – Built with Ray to handle large-scale trading strategies across multiple processes and nodes.
🔌 Seamless Integration – Connects easily to market data providers, brokers, and exchanges.
📈 Backtesting & Live Trading – Develop strategies, backtest them, and switch to live trading without rewriting code.
⚡ Real-Time Order Management – Handles execution, risk management, and portfolio exposure dynamically.
🔧 Customizable Strategy Framework – A modular system that allows users to develop and optimize their own strategies.
Alchemist is part of a broader quantitative trading ecosystem, providing a structured way to move from strategy research to live trading with minimal friction.
Quantitative traders often struggle with the transition from research to live deployment due to:
- Scalability Issues – Traditional execution frameworks cannot handle large-scale strategies efficiently.
- Infrastructure Complexity – Research environments are often disconnected from live market execution.
- Computational Bottlenecks – High-performance backtesting and execution require parallelized, distributed computing.
- Integration Challenges – Many systems lack easy connectivity with brokers, exchanges, and real-time market data.
Alchemist eliminates these roadblocks by providing a fully integrated, high-performance framework that connects research, backtesting, and execution in one system.
Whether you’re working on statistical arbitrage, market-making, or machine learning-based trading, Alchemist ensures your strategies can transition from development to live execution seamlessly
Install the dependencies using Poetry:
poetry installproducts:
- type: alchemist.products.StockProduct
name: test_pdt
base_currency: USD
exch: test_exch
# - type: FutureProduct
# name: 'YM'
# base_currency: 'USD'
# exch: 'CBOT'
# contract_month: '2024-09'
# TODO: this actually hard-coded for IB. Need to make it more generic. Will make updates to the Account class.
accounts:
- name: 'ib_account'
client_id: 999
host: "127.0.0.1"
port: 4002
acc: DU1474528
data_cards:
- product: test_pdt # this should match the product name to one of the products defined above
freq: "1m"
aggregation: ohlcv
order_ports: [5555, 5556]
# this is optional. monitor is default to be None
# monitor:
# type: TimescaleDBMonitor
# db_url: TIMESCALEDB_URI
# flush_interval: 15.0
gateways:
test_gateway:
type: alchemist.gateways.MockGateway
subscriptions: ["bar"]
zmq_send_port: 5557
zmq_recv_ports: [5555, 5556]
# Uncomment and configure as needed
# ib_gateway:
# type: IbGateway
# subscriptions: ["tick"]
# zmq_send_port: 5558
# zmq_recv_ports: [5555]
# accounts: [/* list account names or details if applicable */]
strategies:
sma_strategy:
type: alchemist.strategies.MovingAverageStrategy
name: sma_strategy
zmq_send_port: 5555
zmq_recv_ports: [5557]
# zmq_recv_ports: [5557, 5558]
n_replicas: 1serve run run_strategies:management_servicelatest 10.35.1e 20250324
Please visit the full documentation for detailed information on setup, configuration, and advanced usage.
pytest tests
# Run tests with coverage report and generate report in html
pytest --cov=alchemist --cov-report=html
# Run tests with coverage report and generate report in console
pytest --cov=alchemist --cov-report=termFor U.S. equities, IB reports the exchange volume divided by 100. So if 3 410 shares traded in that 5s window, IB sends 34.10. source
The TWS “Avg Price” field (which maps to the API’s averageCost) is defined as: “Average price is calculated by dividing your cost (execution price + commission) by the quantity of your position.”
Trader Workstation provides only 5 secs bars for real time. Therefore, the LOWEST resolution in IBGateway is 5s.
That means if you want to trade exactly 1m bars, you need to configure 2 data cards - 1 for 5s bars and the other for 1m bars. Otherwise, next will be called for every 5s bar. (next will be called only at syncronization of 5s and 1m bars. Hence every 1m.)
Alchemist is an independent, high-performance automated trading system built using Ray. While some architectural concepts were inspired by previous open-source projects I was involved in, this system has been built from the ground up with a redesigned architecture.
This project is not affiliated with or derived from any specific open-source repository, and any similarities reflect common design patterns in algorithmic trading systems.
This project includes code adapted from an open-source project licensed under Apache License 2.0:
- The file
zeromq.pyis adapted from PFund, an open-source trading system. - Modifications have been made to fit the architecture of Alchemist.
- The original Apache 2.0 license notice is retained in
zeromq.pyin accordance with open-source licensing requirements.
Alchemist is not affiliated with or endorsed by the authors of Pfund. Any concerns regarding code attribution are welcome, and I am committed to maintaining transparency and compliance with open-source licensing.