/alpaca-trade-api-python

Python client for Alpaca's trade API

Primary LanguagePythonApache License 2.0Apache-2.0

PyPI version CircleCI Updates Python 3

alpaca-trade-api-python

alpaca-trade-api-python is a python library for the Alpaca Commission Free Trading API. It allows rapid trading algo development easily, with support for both REST and streaming data interfaces. For details of each API behavior, please see the online API document.

- To use the Streaming abilities go to section called StreamConn

Note this module supports only python version 3.6 and above, due to the async/await and websockets module dependency.

Install

We support python 3.x. If you want to work with python < 3.7 please note that these package dropped support for python <3.7 for the following versions:

pandas >= 1.2.0
numpy >= 1.20.0
scipy >= 1.6.0

The solution - manually install these package before installing alpaca-trade-api. e.g:

pip install pandas==1.1.5 numpy==1.19.4 scipy==1.5.4
$ pip3 install alpaca-trade-api

Example

In order to call Alpaca's trade API, you need to sign up for an account and obtain API key pairs. Replace <key_id> and <secret_key> with what you get from the web console.

REST example

import alpaca_trade_api as tradeapi

api = tradeapi.REST('<key_id>', '<secret_key>', base_url='https://paper-api.alpaca.markets') # or use ENV Vars shown below
account = api.get_account()
api.list_positions()

please note the exact format of the dates

Example Scripts

Please see the examples/ folder for some example scripts that make use of this API

API Document

The HTTP API document is located at https://docs.alpaca.markets/

API Version

API Version now defaults to 'v2', however, if you still have a 'v1' account, you may need to specify api_version='v1' to properly use the API until you migrate.

Authentication

The Alpaca API requires API key ID and secret key, which you can obtain from the web console after you sign in. You can pass key_id and secret_key to the initializers of REST or StreamConn as arguments, or set up environment variables as outlined below.

Alpaca Environment Variables

The Alpaca SDK will check the environment for a number of variables that can be used rather than hard-coding these into your scripts.

Environment default Description
APCA_API_KEY_ID=<key_id> Your API Key
APCA_API_SECRET_KEY=<secret_key> Your API Secret Key
APCA_API_BASE_URL=url https://api.alpaca.markets (for live)
https://paper-api.alpaca.markets (for paper)
Specify the URL for API calls, Default is live, you must specify this to switch to paper endpoint!
APCA_API_DATA_URL=url https://data.alpaca.markets Endpoint for data API
APCA_RETRY_MAX=3 3 The number of subsequent API calls to retry on timeouts
APCA_RETRY_WAIT=3 3 seconds to wait between each retry attempt
APCA_RETRY_CODES=429,504 429,504 comma-separated HTTP status code for which retry is attempted
POLYGON_WS_URL wss://socket.polygon.io/stocks Endpoint for streaming polygon data. You likely don't need to change this unless you want to proxy it for example
POLYGON_KEY_ID Your Polygon key, if it's not the same as your Alpaca API key. Most users will not need to set this to access Polygon.
DATA_PROXY_WS When using the alpaca-proxy-agent you need to set this environment variable as described here

REST

The REST class is the entry point for the API request. The instance of this class provides all REST API calls such as account, orders, positions, and bars.

Each returned object is wrapped by a subclass of the Entity class (or a list of it). This helper class provides property access (the "dot notation") to the json object, backed by the original object stored in the _raw field. It also converts certain types to the appropriate python object.

import alpaca_trade_api as tradeapi

api = tradeapi.REST()
account = api.get_account()
account.status
=> 'ACTIVE'

The Entity class also converts the timestamp string field to a pandas.Timestamp object. Its _raw property returns the original raw primitive data unmarshaled from the response JSON text.

Please note that the API is throttled, currently 200 requests per minute, per account. If your client exceeds this number, a 429 Too many requests status will be returned and this library will retry according to the retry environment variables as configured.

If the retries are exceeded, or other API error is returned, alpaca_trade_api.rest.APIError is raised. You can access the following information through this object.

  • the API error code: .code property
  • the API error message: str(error)
  • the original request object: .request property
  • the original response objecgt: .response property
  • the HTTP status code: .status_code property

API REST Methods

Rest Method End Point Result
get_account() GET /account and Account entity.
get_order_by_client_order_id(client_order_id) GET /orders with client_order_id Order entity.
list_orders(status=None, limit=None, after=None, until=None, direction=None, nested=None) GET /orders list of Order entities. after and until need to be string format, which you can obtain by pd.Timestamp().isoformat()
submit_order(symbol, qty, side, type, time_in_force, limit_price=None, stop_price=None, client_order_id=None, order_class=None, take_profit=None, stop_loss=None, trail_price=None, trail_percent=None) POST /orders Order entity.
get_order(order_id) GET /orders/{order_id} Order entity.
cancel_order(order_id) DELETE /orders/{order_id}
cancel_all_orders() DELETE /orders
list_positions() GET /positions list of Position entities
get_position(symbol) GET /positions/{symbol} Position entity.
list_assets(status=None, asset_class=None) GET /assets list of Asset entities
get_asset(symbol) GET /assets/{symbol} Asset entity
get_barset(symbols, timeframe, limit, start=None, end=None, after=None, until=None) GET /bars/{timeframe} Barset with limit Bar objects for each of the the requested symbols. timeframe can be one of minute, 1Min, 5Min, 15Min, day or 1D. minute is an alias of 1Min. Similarly, day is an alias of 1D. start, end, after, and until need to be string format, which you can obtain with pd.Timestamp().isoformat() after cannot be used with start and until cannot be used with end.
get_aggs(symbol, timespan, multiplier, _from, to) GET /aggs/ticker/{symbol}/range/{multiplier}/{timespan}/{from}/{to} Aggs entity. multiplier is the size of the timespan multiplier. timespan is the size of the time window, can be one of minute, hour, day, week, month, quarter or year. _from and to must be in YYYY-MM-DD format, e.g. 2020-01-15.
get_last_trade(symbol) GET /last/stocks/{symbol} Trade entity
get_last_quote(symbol) GET /last_quote/stocks/{symbol} Quote entity
get_clock() GET /clock Clock entity
get_calendar(start=None, end=None) GET /calendar Calendar entity
get_portfolio_history(date_start=None, date_end=None, period=None, timeframe=None, extended_hours=None) GET /account/portfolio/history PortfolioHistory entity. PortfolioHistory.df can be used to get the results as a dataframe

Rest Examples

Using submit_order()

Below is an example of submitting a bracket order.

api.submit_order(
    symbol='SPY',
    side='buy',
    type='market',
    qty='100',
    time_in_force='day',
    order_class='bracket',
    take_profit=dict(
        limit_price='305.0',
    ),
    stop_loss=dict(
        stop_price='295.5',
        limit_price='295.5',
    )
)
Using get_barset()
import pandas as pd
NY = 'America/New_York'
start=pd.Timestamp('2020-08-01', tz=NY).isoformat()
end=pd.Timestamp('2020-08-30', tz=NY).isoformat()
print(api.get_barset(['AAPL', 'GOOG'], 'day', start=start, end=end).df)

# Minute data example
start=pd.Timestamp('2020-08-28 9:30', tz=NY).isoformat()
end=pd.Timestamp('2020-08-28 16:00', tz=NY).isoformat()
print(api.get_barset(['AAPL', 'GOOG'], 'minute', start=start, end=end).df)

please note that if you are using limit, it is calculated from the end date. and if end date is not specified, "now" is used.
Take that under consideration when using start date with a limit.


StreamConn

The StreamConn class provides WebSocket-based event-driven interfaces. Using the on decorator of the instance, you can define custom event handlers that are called when the pattern is matched on the channel name. Once event handlers are set up, call the run method which runs forever until a critical exception is raised. This module itself does not provide any threading capability, so if you need to consume the messages pushed from the server, you need to run it in a background thread.

We provide 2 price data websockets. The polygon data, and the alpaca data. We default to Alpaca data, and one must explicitly specify the polygon data stream in order to use that. It is done by passing the data_stream keyword to the __init__() function of StreamConn (options: 'alpacadatav1', 'polygon')

This class provides a unique interface to the two interfaces, both Alpaca's account/trade updates events and Polygon's price updates. One connection is established when the subscribe() is called with the corresponding channel names. For example, if you subscribe to trade_updates, a WebSocket connects to Alpaca stream API, and if AM.* given to the subscribe() method, a WebSocket connection is established to Polygon's interface. If your account is enabled for Alpaca Data API streaming, adding alpacadatav1/ prefix to T.<symbol>, Q.<symbol> and AM.<symbol> will also connect to the data stream interface.

The run method is a short-cut to start subscribing to channels and running forever. The call will be blocked forever until a critical exception is raised, and each event handler is called asynchronously upon the message arrivals.

The run method tries to reconnect to the server in the event of connection failure. In this case, you may want to reset your state which is best in the connect event. The method still raises an exception in the case any other unknown error happens inside the event loop.

The msg object passed to each handler is wrapped by the entity helper class if the message is from the server.

Each event handler has to be a marked as async. Otherwise, a ValueError is raised when registering it as an event handler.

conn = StreamConn()

@conn.on(r'^trade_updates$')
async def on_account_updates(conn, channel, account):
    print('account', account)

@conn.on(r'^status$')
async def on_status(conn, channel, data):
    print('polygon status update', data)

@conn.on(r'^AM$')
async def on_minute_bars(conn, channel, bar):
    print('bars', bar)

@conn.on(r'^A$')
async def on_second_bars(conn, channel, bar):
    print('bars', bar)

# blocks forever
conn.run(['trade_updates', 'AM.*'])

# if Data API streaming is enabled
# conn.run(['trade_updates', 'alpacadatav1/AM.SPY'])

You will likely call the run method in a thread since it will keep running unless an exception is raised.

StreamConn Method Description
subscribe(channels) Request "listen" to the server. channels must be a list of string channel names.
unsubscribe(channels) Request to stop "listening" to the server. channels must be a list of string channel names.
run(channels) Goes into an infinite loop and awaits for messages from the server. You should set up event listeners using the on or register method before calling run.
on(channel_pat) As in the above example, this is a decorator method to add an event handler function. channel_pat is used as a regular expression pattern to filter stream names.
register(channel_pat, func) Registers a function as an event handler that is triggered by the stream events that match with channel_path regular expression. Calling this method with the same channel_pat will overwrite the old handler.
deregister(channel_pat) Deregisters the event handler function that was previously registered via on or register method.

Debugging

Websocket exceptions may occur during execution. It will usually happen during the consume() method, which basically is the websocket steady-state.
exceptions during the consume method may occur due to:

  • server disconnections
  • error while handling the response data

We handle the first issue by reconnecting the websocket every time there's a disconnection. The second issue, is usually a user's code issue. To help you find it, we added a flag to the StreamConn object called debug. It is set to False by default, but you can turn it on to get a more verbose logs when this exception happens. Turn it on like so StreamConn(debug=True)

Logging

You should define a logger in your app in order to make sure you get all the messages from the different components.
It will help you debug, and make sure you don't miss issues when they occur.
The simplest way to define a logger, if you have no experience with the python logger - will be something like this:

import logging
logging.basicConfig(format='%(asctime)s %(message)s', level=logging.INFO)

Polygon API Service

Alpaca's API key ID can be used to access Polygon API, the documentation for which is found here. This python SDK wraps their API service and seamlessly integrates it with the Alpaca API. alpaca_trade_api.REST.polygon will be the REST object for Polygon.

The example below gives AAPL daily OHLCV data in a DataFrame format.

import alpaca_trade_api as tradeapi

api = tradeapi.REST()
# all of these examples work
aapl = api.polygon.historic_agg_v2('AAPL', 1, 'day', _from='2019-01-01', to='2019-02-01').df
aapl = api.polygon.historic_agg_v2('AAPL', 1, 'day', _from=datetime.datetime(2019, 1, 1), to='2019-02-01').df
aapl = api.polygon.historic_agg_v2('AAPL', 1, 'day', _from=datetime.date(2019, 1, 1), to='2019-02-01').df
aapl = api.polygon.historic_agg_v2('AAPL', 1, 'day', _from=pd.Timestamp('2019-01-01'), to='2019-02-01').df
# timestamp should be in milliseconds datetime.datetime(2019, 1, 1).timestamp()*1000 == 1546293600000
aapl = api.polygon.historic_agg_v2('AAPL', 1, 'day', _from=1546293600000, to='2019-02-01').df

and here's a minute example usage:

import pytz
NY = 'America/New_York'

start = pytz.timezone(NY).localize(datetime(2020,1,2,9,30)).timestamp()*1000  # timestamp in micro seconds
# another alternative will be: start = pd.Timestamp('2020-01-02 09:30', tz=NY).value/1e6
end = pytz.timezone(NY).localize(datetime(2020,1,2,16,0)).timestamp()*1000
df = api.polygon.historic_agg_v2('AAPL', 1, 'minute', _from=start, to=end).df

polygon/REST

It is initialized through the alpaca REST object.

REST Method Description
exchanges() Returns a list of Exchange entity.
symbol_type_map() Returns a SymbolTypeMap object.
historic_trades_v2(symbol, date, timestamp=None, timestamp_limit=None, reverse=None, limit=None) Returns a TradesV2 which is a list of Trade entities. date is a date string such as '2018-2-2'. The returned quotes are from this day only. timestamp is an integer in Unix Epoch nanoseconds as the lower bound filter, exclusive. timestamp_limit is an integer in Unix Epoch nanoseconds as the maximum timestamp allowed in the results. limit is an integer for the number of ticks to return. Default and max is 50000.
TradesV2.df Returns a pandas DataFrame object with the ticks returned by historic_trades_v2.
historic_quotes_v2(symbol, date, timestamp=None, timestamp_limit=None, reverse=None, limit=None) Returns a QuotesV2 which is a list of Quote entities. date is a date string such as '2018-2-2'. The returned quotes are from this day only. timestamp is an integer in Unix Epoch nanoseconds as the lower bound filter, exclusive. timestamp_limit is an integer in Unix Epoch nanoseconds as the maximum timestamp allowed in the results. limit is an integer for the number of ticks to return. Default and max is 50000.
QuotesV2.df Returns a pandas DataFrame object with the ticks returned by the historic_quotes_v2.
historic_agg_v2(self, symbol, multiplier, timespan, _from, to, unadjusted=False, limit=None) Returns an AggsV2 which is a list of Agg entities. AggsV2.df gives you the DataFrame object.
- multiplier is an integer affecting the amount of data contained in each Agg object.
-timespan is a string affecting the length of time represented by each Agg object. It is one of the following values: minute, hour, day, week, month, quarter, year.
- _from is an Eastern Time timestamp string/object that filters the result for the lower bound, inclusive. we accept the date in these formats: datetime.datetime, datetime.date, pd.Timestamp, datetime.timestamp, isoformat string (YYYY-MM-DD).
- to is an Eastern Time timestamp string that filters the result for the upper bound, inclusive. we support the same formats as the _from field.
- unadjusted can be set to true if results should not be adjusted for splits.
- limit is an integer to limit the number of results. 3000 is the default and max value.
The returned entities have fields relabeled with the longer name instead of shorter ones. For example, the o field is renamed to open.
Aggs.df Returns a pandas DataFrame object with the ticks returned by historic_agg_v2.
daily_open_close(symbol, date) Returns a DailyOpenClose entity.
last_trade(symbol) Returns a Trade entity representing the last trade for the symbol.
last_quote(symbol) Returns a Quote entity representing the last quote for the symbol.
condition_map(ticktype='trades') Returns a ConditionMap entity.
company(symbol) Returns a Company entity if symbol is string, or a dict[symbol -> Company] if symbol is a list of string.
dividends(symbol) Returns a Dividends entity if symbol is string, or a dict[symbol -> Dividends] if symbol is a list of string.
splits(symbol) Returns a Splits entity for the symbol.
earnings(symbol) Returns an Earnings entity if symbol is string, or a dict[symbol -> Earnings] if symbol is a list of string.
financials_v2(symbol, limit, report_type, sort) Returns an Financials entity if symbol is string, or a dict[symbol -> Financials] if symbol is a list of string.
news(symbol) Returns a NewsList entity for the symbol.

Running Multiple Strategies

There's a way to execute more than one algorithm at once.
The websocket connection is limited to 1 connection per account.
For that exact purpose this project was created
The steps to execute this are:

  • Run the Alpaca Proxy Agent as described in the project's README
  • Define this env variable: DATA_PROXY_WS to be the address of the proxy agent. (e.g: DATA_PROXY_WS=ws://127.0.0.1:8765)
  • If you are using the Alpaca data stream, make sure you you initiate the StreamConn object with the container's url, like so: data_url='http://127.0.0.1:8765'
  • execute your algorithm. it will connect to the servers through the proxy agent allowing you to execute multiple strategies

Raw Data vs Entity Data

By default the data returned from the api or streamed via StreamConn is wrapped with an Entity object for ease of use.
Some users may prefer working with raw python objects (lists, dicts, ...).
You have 2 options to get the raw data:

  • Each Entity object as a _raw property that extract the raw data from the object.
  • If you only want to work with raw data, and avoid casting to Entity (which may take more time, casting back and forth)
    you could pass raw_data argument to Rest() object or the StreamConn() object.

Support and Contribution

For technical issues particular to this module, please report the issue on this GitHub repository. Any API issues can be reported through Alpaca's customer support.

New features, as well as bug fixes, by sending a pull request is always welcomed.