fractal.loaders package

Subpackages

Submodules

fractal.loaders.aave module

class fractal.loaders.aave.AaveLoader(reserve_id: str, loader_type: LoaderType, url: str, start_time: datetime = None, resolution: int = 1)[source]

Bases: Loader

extract()[source]

load()[source]

read(with_run: bool = False) → LendingHistory[source]

transform()[source]

class fractal.loaders.aave.AaveV2EthereumLoader(asset_address: str, loader_type: LoaderType, start_time: datetime = None, resolution: int = 1)[source]: Bases: AaveLoader

class fractal.loaders.aave.AaveV3ArbitrumLoader(asset_address: str, loader_type: LoaderType, start_time: datetime = None, resolution: int = 1)[source]: Bases: AaveLoader

fractal.loaders.base_loader module

class fractal.loaders.base_loader.Loader(loader_type: LoaderType, *args, **kwargs)[source]

Bases: ABC

abstract extract()[source]

file_path(*args)[source]

abstract load()[source]

abstract read(with_run: bool = False) → DataFrame[source]

run()[source]

abstract transform()[source]

class fractal.loaders.base_loader.LoaderType(value, names=None, *values, module=None, qualname=None, type=None, start=1, boundary=None)[source]

Bases: Enum

CSV = 1

JSON = 2

PICKLE = 4

SQL = 3

fractal.loaders.binance module

class fractal.loaders.binance.BinanceDayPriceLoader(ticker: str, loader_type: LoaderType, inverse_price: bool = False)[source]

Bases: Loader

extract()[source]

load()[source]

read(with_run: bool = False) → PriceHistory[source]

Reads the price history data from the Binance loader.

Parameters:: with_run (bool, optional) – If True, runs the loader before reading the data. Defaults to False.
Returns:: The price history data.
Return type:: PriceHistory

transform()[source]

class fractal.loaders.binance.BinanceFundingLoader(ticker: str, loader_type: LoaderType = LoaderType.CSV, start_time: datetime = None, end_time: datetime = None)[source]

Bases: Loader

extract()[source]

get_funding_rates(ticker: str, start_time: datetime = None, end_time: datetime = None) → list[source]

Get funding rate history

Parameters:

ticker – str
start_time – datetime
end_time – datetime

Returns:

List of funding rates

Return type:

list

load()[source]

read(with_run: bool = False) → FundingHistory[source]

transform()[source]

class fractal.loaders.binance.BinanceHourPriceLoader(ticker: str, loader_type: LoaderType = LoaderType.CSV, start_time: datetime = None, end_time: datetime = None, inverse_price: bool = False)[source]

Bases: Loader

extract()[source]

get_klines(ticker: str, start_time: datetime = None, end_time: datetime = None) → list[source]

Get Kline/Candlestick Data Klines are uniquely identified by their open time.

Parameters:

ticker – str
start_time – datetime
end_time – datetime

Returns:

List of klines

Return type:

list

load()[source]

read(with_run: bool = False) → PriceHistory[source]

transform()[source]

fractal.loaders.gmx_v1 module

class fractal.loaders.gmx_v1.GMXV1FundingLoader(token_address: str, loader_type: LoaderType)[source]

Bases: Loader

extract()[source]

load()[source]

read(with_run: bool = False) → FundingHistory[source]

transform()[source]

fractal.loaders.structs module

class fractal.loaders.structs.FundingHistory(rates: array, time: array)[source]

Bases: DataFrame

Funding history data structure.

class fractal.loaders.structs.LendingHistory(lending_rates: array, borrowing_rates: array, time: array)[source]

Bases: DataFrame

Lending/borrowing history data structure.

class fractal.loaders.structs.PoolHistory(tvls: array, volumes: array, fees: array, liquidity: array, time: array, prices: array | None = None)[source]

Bases: DataFrame

Pool data structure.

class fractal.loaders.structs.PriceHistory(prices: array, time: array)[source]

Bases: DataFrame

Price history data structure.

class fractal.loaders.structs.RateHistory(rates: array, time: array)[source]

Bases: DataFrame

Staking rate history data structure.

Module contents

class fractal.loaders.AaveV2EthereumLoader(asset_address: str, loader_type: LoaderType, start_time: datetime = None, resolution: int = 1)[source]: Bases: AaveLoader

class fractal.loaders.AaveV3ArbitrumLoader(asset_address: str, loader_type: LoaderType, start_time: datetime = None, resolution: int = 1)[source]: Bases: AaveLoader

class fractal.loaders.ArbitrumGraphLoader(api_key: str, subgraph_id: str, loader_type: LoaderType)[source]

Bases: BaseGraphLoader

Graph Loader with arbitrum gateway

ROOT_URL = 'https://gateway-arbitrum.network.thegraph.com/api'

class fractal.loaders.BaseGraphLoader(root_url: str, api_key: str, subgraph_id: str, loader_type: LoaderType)[source]

Bases: Loader

Base class for The Graph loaders.

Each graph loader should inherit from this class. It should have the following attributes: - subgraph_id: The ID of the subgraph - root_url: The root URL of the graph - api_key: The API key to access

_make_request()[source]: Make a request to the graph with a given query

class fractal.loaders.BinanceDayPriceLoader(ticker: str, loader_type: LoaderType, inverse_price: bool = False)[source]

Bases: Loader

extract()[source]

load()[source]

read(with_run: bool = False) → PriceHistory[source]

Reads the price history data from the Binance loader.

Parameters:: with_run (bool, optional) – If True, runs the loader before reading the data. Defaults to False.
Returns:: The price history data.
Return type:: PriceHistory

transform()[source]

class fractal.loaders.BinanceFundingLoader(ticker: str, loader_type: LoaderType = LoaderType.CSV, start_time: datetime = None, end_time: datetime = None)[source]

Bases: Loader

extract()[source]

get_funding_rates(ticker: str, start_time: datetime = None, end_time: datetime = None) → list[source]

Get funding rate history

Parameters:

ticker – str
start_time – datetime
end_time – datetime

Returns:

List of funding rates

Return type:

list

load()[source]

read(with_run: bool = False) → FundingHistory[source]

transform()[source]

class fractal.loaders.BinanceHourPriceLoader(ticker: str, loader_type: LoaderType = LoaderType.CSV, start_time: datetime = None, end_time: datetime = None, inverse_price: bool = False)[source]

Bases: Loader

extract()[source]

get_klines(ticker: str, start_time: datetime = None, end_time: datetime = None) → list[source]

Get Kline/Candlestick Data Klines are uniquely identified by their open time.

Parameters:

ticker – str
start_time – datetime
end_time – datetime

Returns:

List of klines

Return type:

list

load()[source]

read(with_run: bool = False) → PriceHistory[source]

transform()[source]

class fractal.loaders.ConstantFundingsLoader(rate: float = -0.001, freq: str = 'D', start: str = '2020-01-01', end: str = '2025-01-01')[source]

Bases: Loader

Generates a constant funding rate for a given period.

extract()[source]

load()[source]

read(with_run: bool = False)[source]

transform()[source]

class fractal.loaders.EthereumUniswapV2PoolDataLoader(api_key: str, pool: str, fee_tier: float, loader_type: LoaderType)[source]

Bases: EthereumUniswapV2Loader

Loader for Uniswap V2 PoolData

extract()[source]

load()[source]

read(with_run: bool = False) → PoolHistory[source]

transform()[source]

class fractal.loaders.FundingHistory(rates: array, time: array)[source]

Bases: DataFrame

Funding history data structure.

class fractal.loaders.GMXV1FundingLoader(token_address: str, loader_type: LoaderType)[source]

Bases: Loader

extract()[source]

load()[source]

read(with_run: bool = False) → FundingHistory[source]

transform()[source]

exception fractal.loaders.GraphLoaderException[source]: Bases: Exception

class fractal.loaders.LPMLSimulatedStatesLoader(price_history: PriceHistory, pool_history: PoolHistory, trajectories_number: int = 100, loader_type: LoaderType = LoaderType.PICKLE, seed: int = 420, random_filename: bool = True)[source]

Bases: Loader

A loader class for generating simulated states: using machine learning models.

Generates pool stated (with price_history) for LP strategies. Prices generates as GBM process with drift and volatility,

which are calculated from the historical data in rolling windows.

The pool data is generated using CatBoostRegressor models,: using the price data, log returns, realized volatility, momentum, and moving average difference as predictors.
After prediction, the residuals are added to the predictions,: as samples from the residuals distribution.

The residuals distribution is estimated using the histogram. Also, pool data is smoothed using a rolling mean.

Parameters:

price_history (PriceHistory) – The price history data.
pool_history (PoolHistory) – The pool data.
trajectories_number (int, optional) – The number of trajectories to generate. Defaults to 100.
loader_type (LoaderType, optional) – The type of loader. Defaults to LoaderType.PICKLE.
seed (int, optional) – The seed for random number generation. Defaults to 420.
random_filename (bool, optional) – Whether to use a random filename for saving the data. Defaults to True.

pool_history

The pool data.

Type:: PoolHistory

price_history

The price history data.

Type:: PriceHistory

_data

The merged data of price_history and pool data.

Type:: pd.DataFrame

_simulated_data

The list of simulated data.

Type:: List[pd.DataFrame]

trajectories_number

The number of trajectories to generate.

Type:: int

_dump_filename

The filename for saving the data.

Type:: str

_random

The random number generator.

Type:: random.Random

_np_random

The numpy random number generator.

Type:: numpy.random.Generator

_seed

The seed for random number generation.

Type:: int

delete_dump_file()[source]

extract()[source]

load()[source]

read(with_run: bool = False) → List[LPSimulatedStates][source]

transform()[source]

class fractal.loaders.LPSimulatedStates(tvls: array, volumes: array, fees: array, liquidity: array, time: array, prices: array)[source]

Bases: DataFrame

Pool data structure.

class fractal.loaders.LendingHistory(lending_rates: array, borrowing_rates: array, time: array)[source]

Bases: DataFrame

Lending/borrowing history data structure.

class fractal.loaders.Loader(loader_type: LoaderType, *args, **kwargs)[source]

Bases: ABC

abstract extract()[source]

file_path(*args)[source]

abstract load()[source]

abstract read(with_run: bool = False) → DataFrame[source]

run()[source]

abstract transform()[source]

class fractal.loaders.LoaderType(value, names=None, *values, module=None, qualname=None, type=None, start=1, boundary=None)[source]

Bases: Enum

CSV = 1

JSON = 2

PICKLE = 4

SQL = 3

class fractal.loaders.MonteCarloHourPriceLoader(price_history: PriceHistory, trajectories_number: int = 100, loader_type: LoaderType = LoaderType.PICKLE, seed: int = 420)[source]

Bases: Loader

A class that represents a Monte Carlo hour price loader.

This loader performs Monte Carlo simulation on the price data, obtained from a base loader.

Generates prices using a GBM process with volatility,: and without drift.

Length of the simulation is equal to the length of the base loader data.

price_history

The price history data.

Type:: PriceHistory

trajectories_number

The number of trajectories to simulate.

Type:: int

loader_type

The type of loader used to save and load data.

Type:: LoaderType

seed

The seed value used for random number generation.

Type:: int

extract()[source]: Extracts the price data from the base loader.

transform()[source]: Performs Monte Carlo simulation on the price data.

load()[source]: Saves the simulated price data using the specified loader type.

read(with_run: bool = False): Reads the simulated price data from the saved file.

run(): Executes the entire process of extracting, transforming, and loading the data.

delete_dump_file()[source]

extract()[source]

load()[source]

read(with_run: bool = False) → List[PriceHistory][source]

transform()[source]

class fractal.loaders.PoolHistory(tvls: array, volumes: array, fees: array, liquidity: array, time: array, prices: array | None = None)[source]