Design Brainstorming...

[asmodehn]

For the record, aiokraken originates from @dantimofte Its simplicity and clarity make it understandable quickly. See : https://github.com/dantimofte/aiokraken

However it does not seem usable as is in an application context.

While we should be careful not to add too many useless features, some seem mandatory if anyone want to use this package safely, given that it is meant to be used in a sensitive environment:

• configuration (keys should never be in source)
• ledger (record of trades) should be done at the lowest level possible in an application, so why not in the package interfacing with the exchange (this one)
• managing currencies as currencies (using decimal, maybe discriminating units)
• Very basic CLI (following kraken WebUI for concepts), but no more. Beware attempting to cater to every user/usage...
• more ?

Will NOT do :

• large scale storage of data. While potentially useful, it can grow quickly and can as well be done by another package using aiokraken.
• any complex operation that can be done as well in another place.
• No GUI or TUI or userfriendly CLI commands. Can be done elsewhere (script or whatever).
• No fancy logging processing, just basic stuff for human/dev consumption, mostly debugging.
• more ?

[asmodehn]

Design update : Actually, will do:

• basic visual output component (plots) in library, but with limitation :
  1. Only show what was used in computation (mostly for understanding computation results and debugging)
  2. we should try to keep that logic separated as much as possible (like using pandas-ta for analysis computation...)
• multiple ways to use and interact with aiokraken (terminal, ipython, notebook, ...) still TBD.

Actually will not do just yet:

• manage currencies as currencies... would require lots of changes, and doesn't seem needed at the moment (?).

[asmodehn]

Package structure refined :

aiokraken/                   # main api package, structured like a domain model. Providing data indexed by asset, pairs or time.
aiokraken/rest             # rest api subpackage, handling Kraken's REST API. also contains schemas and self tests.
aiokraken/websocket  # websocket api subpackage, handling Kraken's WSS API. also contains schemas and self tests.
aiokraken/model         # subpackage holding all common models and timeindexed dataframes
aiokraken/layouts        # bokeh content to be able to visually debug our models...

• We aim to obtain a generic interface for the user, independent of krakens internal, as much as suitable.
• We will provide only a minimal TA functionality over our OHLC model (leveraging pandas-ta), but not much more in here. Use this package to build more complex strategies by yourself.
• We want to provide a minimal interactive TUI via the __main__ scripts in the various package, where it is meaningful. It should help newcomers to understand the package.