The cryptoast library provides functionalities for retrieving, storing, and analyzing data. Potenially it also provides functionalities for taking actions based on this data; Either through inferred signals' thresholds or by using a reinforcement learning model on top of it.
Optional third-party apis used are:
- BINANCE for data retrieval and order execution
- AWS for data storage (optional, local storage can be used)
- pandas
- numpy
- gym
- ta
- s3fs
- tensorflow
- scikit-learn
- python-binance
- stable-baselines
A number of packages and subsequent modules are available:
Object for asset-specific data storage, retrieval, and analysis (through ta indicators and derived signals).
Object for regrouping multiple Kline objects, and representing them as a readable list of official asset acronyms.
Management object for all available assets.
In-house backtester, feedable to a reinforcement-learning agent.
Single Asset Environment for reinforcement-learning purposes.
pip install -e cryptoast
Local storage update and average signal example.
from binance.client import Client
from cryptoast.base.kline import Kline
client = Client('key', 'secret')
kline = Kline('ETHUSDT')
kline.update(client, store=True)
if kline.signals.iloc[-1, :].eq(1).mean() > .5:
print('buy')
Notebook showing cryptoast's core functionalities.
Testing notebook showing mainly: initialization, data update, dummy buy-sell signal computations and rl implementation tryouts.
Exhaustive testing notebook showing amongst other: initialization, data update, single-asset buy-sell signal computation tryouts, multi-asset buy-sell signal computation tryouts, rl implementation tryouts, categorical modeling tryouts, etc.
No relation whatsoever to http://cryptoast.fr