rbot (Rusty bot framwrok for crypto trading)

rbot is a python crypt trading bot framework written in Rust.

The feature is;

Your Bot can be executed in three mode backtest, dry_run and production without modification.
TICK BASED backtesting.
Written in Rust and Provided as Python plugin. So you can write your own bot in Python with super efficent Rust back end.

For 0.4.0 Beta release (This is beta release)

Change:

OHLCVのTimestampの型をi64からDateTimeへ変更(->Agentに影響あり）
約定ログの保存方式をSQLiteからparquetへ変更（ファイルサイズが５分の１へ縮小）
注意：過去にダウンロードした約定ログは使えません。DBディレクトリごと削除し、再度ダウンロードしてください。
DownloadとBackテストにProgressBarを導入

Fix:

#244 立ち上げ直後に最初のTickでon_clockがよびだされていたのを修正
#255 Session.ohlcvで指定した本数よりも多く足が計算されていたのを修正。

Limitations:

バックテスト以外のDryRun,RealRunは修正後テストしていません。
一旦このβリリースでは動かないものと考えてください。

the execution sample

In the backtest mode, you can run with Jupyter notebook. So, you can make such analysis below. This example is provided in the github, which can be executed on the Google colab!! Please try.

sample1: basil bot sample2: breakout bot

Architecture

The only you have to make is Agent. The other class will support you Agent will work.

Session will provide abstruction layer to the Market. Agent only interact with Session for ordering, fetching market information and current order status.
Market is the API and DB wrapper for each trading pair in the exchange. At this version Bybit and Biannce exchange are supportd.
Runner makes session to run back_test, dry_run or real_run.
Logger stores the result of execution. That provide the result in polars DataFrame.

install

$ pip install rbot

download historical data

create market object

from rbot import Bybit, BybitConfig
exchange = Bybit(production=True)

config = BybitConfig.BTCUSDT          # use BTC/USDT pair
market = exchange.open_market(config)

you can find the locatin of data base.

market.file_name

In case you want to alter the location of db, you can specify the path by environment variable RBOT_DB_ROOT.

enable order

Ordering is disabled by default. You can enable it by setting enable_order_with_my_own_risk to True.

exchange.enable_order_with_my_own_risk = True

download archive data

market.download_archive(
    ndays=1,        # specify from past days
    force=False,    # if false, the cache data will be used.
    verbose=True    # verbose to print download progress.
)

get OHLCV data

ohlcv = market.ohlcv(
    start_time=0,           # start time in unix timestamp(microseconds)
    end_time=0,             # end time in unix timestamp(microseconds)
    window_sec=60           # ohlc bar window size in seconds
)

The OHLCV data is a polars data frames with the following columns:

timestamp (in asc order)
open
high
low
close
volume (sum of volume in order size)
count (number of ticks)

if you want to use it as pandas dataframe, you can convert it by the following code.

pandas_df = ohlcv.to_pandas(use_pyarrow_extension_array=True)

Creating skelton bot

You can make Bot class with any kind of names. Only rules is you must implement on_init, on_tick, on_clock, and on_update, if you want to recieve such event(It is not necessary to implement all of them).

Each method receives session object. You can use it to get market information, place order, cancel order, etc.

Example:

class SkeltonAgent:      # you can use any names for trading bot agent / クラス名は任意です
    def on_init(self, session):
        """
        Bot initialization process. Bot initialization time. Called once at the start of the bot.
        It is best place to seting up session.clock_interval_sec which is interval of on_clock call.
        Args:
            session: Session class 
        """
        session.clock_interval_sec = 60 * 60 * 1        # 1時間ごとにon_clockを呼び出す

    def on_tick(self, session, side, price, size):
        """
        If you implement this method, you can receive all tick data from exchange.
        Args:
            session: Session object (that can be used to order and get market information)
            side: "Sell" or "Order"
            price: executed price of tick
            size: executed size of the tick
        """
        pass

    def on_clock(self, session, clock):
        """
        If you implement this method and seting up session.clock_interval_sec, 
        you can receive clock event in specified interval.
        Args:
            session: Session object(that can be used to order or get market information) 
            clock: Unix time stamp in micro seconds.
        """
        pass

    def on_update(self, session, updated_order):
        """
        If your order's status is changed, this method is called.
        Args:
            session: Session object
            updated_order: Updated order
        """
        pass

Session API

In your on_init, on_tick, on_clock, and on_update, you can use session object to get market information, place order, cancel order, etc.

OHLCV data

ohlcv = session.ohlcv(
    interval=60,        # ohlc bar window size in seconds
    count=10,           # number of bars to generate
)

The OHLCV data is a polars data frames with the following columns:

timestamp (in asc order)
open
high
low
close
volume (sum of volume in order size)
count (number of ticks)

order book

bid, ask = session.board

bid, ask are polars dataframes with the following columns:

price: price of the order
size: size of the order
sum: cumulative size of the order

place order

market order

size = 0.001
market_order = session.market_order("BUY", size)

limit order

price = 50000.0
size = 0.001
sell_limit_order = session.limit_order("SELL", price, size)

cancel order

cancelled_order =session.cancel_order(config, id_to_cancel)

order queue status

You can get the order queue status(Limit orders that is not fullfied)

buy_orders = session.buy_orders
sell_orders = session.sell_orders

expire order

Expire the order in the order queue.

expire_time = 60 * 60   # expire older than 1H
session.expire_order(expire_time)

position

calculate psudo-position from the session starts.

position = session.position

Running bot

You can run Agent without modification in three modes; backtest, dry_run, and production.

backtest

from rbot import Runner

agent = SkeltonAgent()
runner = Runner()

session = runner.back_test(
                exchange=exchange,  # exchange object
                market=market,    # market object
                agent=agent,      # agent object
                start_time=0,    # start time in unix timestamp(microseconds)` 0 means from the beginig of DB
                end_time=0,      # end time in unix timestamp(microseconds) 0 means to the end of the DB
                verbose=True     # verbose to print progress.
            )

dry run

from rbot import Runner

agent = SkeltonAgent()
runner = Runner()

session = runner.dry_run(
                exchange=exchange,  # exchange object
                market=market,    # market object
                agent=agent,      # agent object
                execute_time = 60,  # Time(Seconds) to execute
                verbose=True     # verbose to print progress.
            )

real run(production)

from rbot import Runner

agent = SkeltonAgent()
runner = Runner()

session = runner.real_run(
                exchange=exchange,  # exchange object
                market=market,    # market object
                agent=agent,      # agent object
                #execute_time = 60,  # execute time, if not set, it runs forever
                verbose=True,
                log_file="skelton_bot.log"
            )

analize bot performance

you can get orders dataframe list by session.orders.

log_id
symbol
update_time
create_time
status
order_id
client_order_id
order_side
order_type
order_price
order_size
remain_size
transaction_id
execute_price
execute_size
quote_vol
commission
commission_asset
is_maker
message
commission_home
commission_foreign
home_change
foreign_change
free_home_change
free_foreign_change
lock_home_change
lock_foreign_change
open_position
close_position
position
profit
fee
total_profit
sum_profit

note

for real run mode, you can retreive log as below

from rbot import Logger

log = Logger()
log.restore("skelton_bot.log")

Changes from release-0.2

Exchange and market is separated. So you must make Exchange(such as Bybit) first, then open_market to build market object.
Market#download is separated into three Market#download_archive, Market#download_latest and Market#donwload_gap
Runner#real_run and Runner#dry_run now have no_download optional flag.
Support Bybit exchange.

links

rbot github

Example github API manuals

yasstake/rbot