• python > 3.6
• pip install jsonpickle this is a non-standard package that is required

• for using NeuralPredictor you would need those packages
• pip install tensorflow
• pip install tflearn
• pip install numpy

The toolkit consists of two parts.

The first part implements remote services (as exchange scraper, history provider, web...) and corresponding connectors. It is highly advised not to touch them.

The second part provides higher abstraction levels over the core and can be subject of customizations.

The bots package implements some basic trading strategies and serves as a showcase of the available API.

All the scripts has to be executed from bot-trading directory.

Every bot must be logged in via a username. It can be either set in bot-trading/bot_trading/configuration.py or passed in as an environment variable USERNAME. The username must be a valid email. Only one controlling connection (i.e. connection allowed to make trades) per username can be done.

The toolkit supports several execution modes.

• launched by python -m bot_trading.run_bot_in_real
• all trades issued are registered by the remote trading server
• runs bot on the most recent trading data

• launched by python -m bot_trading.run_manual_bot_in_real
• all trades issued are registered by the remote trading server
• opens local web trading client that can be accessed on http://127.0.0.1:5522/

• launched by python -m bot_trading.run_bot_in_sandbox
• all trades done by the bot are sandboxed (not registered by the remote trading server)
• runs bot on the most recent trading data

• launched by python -m bot_trading.run_bot_backtest
• all trades done by the bot are sandboxed (not registered by the remote trading server)
• runs bot on historical data (date range can be specified in the launcher script)

• bot-trading/bot_trading/configuration.py
• permanent USERNAME can be specified here (so it does not need to be passed as env var)
• allows to configure some apects of the toolkit
• logging verboseness can be configured here