Updated - implement the following: Updated 8/14:

• [x] A Flask web server to return a web interface, and provide an API for harvest's status. Possibly use socket.io for real-time data.
• [x] A React frontend. For now, it should be very simple, more like a proof-of-concept. Svelte is a good candidate for the framework.
• [ ] Update functions in harvest to automatically log certain actions (like buying stocks) and send the data to the frontend
• [x] #39

Using documentation in _base.py and the example of robinhood.py, implement

• [ ] TDAmeritrade, using robin_stock
• [x] WeBull, using webull
• [x] Alpaca, using alpaca-trade-api-python
• [x] Kraken using python3-krakenex

This issue is marked as a good first issue, because it is supposed to be a litmus test of how easy it is to add a new broker support to Harvest, which is one of its selling point.

One issue with the way the code is organized currently is that each module frequently makes references to each other. For example, the Storage class gets the current time by referencing trader.timestamp, and Trader updates the timestamp by referencing Streamer. This can make the code messy and make the ownership of the attributes unclear.

My current idea is to bundle common runtime attributes like timestamp and watchlist, as well as info like current stock positions, into a single class/dictionary. Every module will share a reference to a single instance of this class/dict.

Currently the functionality of DummyBroker is a little confusing - when used as a streamer, it generates dummy data as its name implies, but when used as a broker, it paper trades. Now that brokers have the mode attribute, DummyBroker should be split into separate brokers as described in the title.

Also, the following name convention is suggested for brokers:

• [Name] for classes that function as streamers && brokers
• [Name]Streamer for classes that only work as streamers
• [Name]Broker for classes that only work as brokers

• [ ] Replace the self.algo in the trader class with a list and replace the set_algo function with a list.
• [ ] Update the trader run function to iterate through the algos and run each

Software like splunk or xsoar have local servers with guis fro users to interface with there system. These servers also make it easier for distributed environments and the ability to run inside a docker container while being able to easily access it outside the container. I think if Harvest had a similar feature, it would be pretty sweet.

I think that this feature is pretty big and probably won't fit into the second milestone, but could be a good third or fourth milestone.

Here are some mockups of how the web app would look:

Harvest currently outputs using the print() function. This is not ideal, since users have no way to control the format of the outputted messages or disable them if necessary. Replace print() statements with info() from python logging module. Link to documentation

Yahoo earnings not working and missing package dependency in setup.py. please add get earnings between dates.

install_requires=[ 'pandas', 'finta', 'pyyaml', 'tqdm', 'pytz' ],

Major Changes

• Create Positions and Position class to manage currently owned assets. It also has helper methods to automatically update attributes such as % profit.
• Create Account class to manage account.
• Create Function class. This class stores various function pointers, so modules can access necessary functions directly rather than daisy chaining references. e.g. self.trader.storage.load is now self.func.load. This is also useful in testing, as we can pass in alternate functions during tests.
• Update code to use the new classes.

Minor Changes

• Changed interval attribute in Trader to watchlist_cfg
• Changed watchlist_global in Trader to watchlist
• Several code quality improvements.

Allow users to specify the date range of backtesting data.

BackTester.start() will take 3 new params:

• start: start date and time of backtest, in format "MM-DD-YYYY:HH:MM:SS", or None to use the max range.
• end: start date and time of backtest, in format "MM-DD-YYYY:HH:MM:SS", or None to use current time
• period: The range of backtesting, in format "{N}DAYS". If start is not specified, start will be set to end - period. If end is not specified, end = start + period.

Some modifications were also made to _base_storage.py

As you can see, there is much more work to be done - many edge cases are not handled, and code needs much more organizing. I'm opening a PR early to get some feedback on the changes I made so far as well as suggestions to better organize the code.

Create a python file that creates a logger and have all files that log import this file. This file should create a logger and configure it.

Suggested logging strategy

• DEBUG: Called at the start of every function and prints the message name and all arguments. Typically not shown to user.
• INFO: Let users know of important things that occur. To give users confidence that the program is running as expected.
• WARNING: Something bad happened and harvest can automatically fix the issue without any user input. For example if the user tries to get stock data from over a period of time that their account allows, we can fix that start time.
• ERROR: Something unexpected happened but can easily be fixed. Like the user tried to add a stock ticker that does not exists or tried to add a stock ticker to a broker that only supports crypto or API call failed, we could try redoing the api call.
• raise Exception: Something really bad happened and the entire system must be shutdown because there is no way to recover. This could be a call to a function that should return something without a solid null case. For example returning an empty list is a fine null case but an empty dictionary or None isn't (since no one checks for the None case).

Let me know if this sounds good or if there is anything I should change.

Describe the bug Sometimes, when placing crypto orders on Robinhood, an error is returned by the API indicating the the number of digits in the order quantity is too much.

To Reproduce For example, if you try to buy 0.00000001 DOGE, the order will fail.

Expected behavior It appears that on Robinhood, the number of digits D_o allowed in the order quantity is N - D_p, where N is some constant and D_p is the number of digits in the current price of the cryptocurrency.

To fix this bug, you must:

• [ ] Investigate the value of N
• [ ] Add checks to buy/sell functions so order quantities are automatically rounded down to the proper number of digits.

Describe the bug For every DataFrame YahooStreamer returns, the latest data has a wrong timestamp. In this case, the timestamp should be 2021-08-01 07:40:00+00:00

To Reproduce Run a Trader with YahooStreamer, and enable debug output. Note the timestamps of the DataFrames returned.