Initial client-side only Trading Strategy notebook research

[miohtama]

Research and investigate what it takes to run Trading Strategy backtesting framework in the web browser using JupterLite and Pydiode and WebAssembly.

Deliverables

• A client-side execution of the Hello World notebook as a static HTML example (any of raw Pydiode, JupyterLite, others)
• Must run the Python code as a separate WebWorker on background / no blocking the UI
• A separate Github repository with HTML + JavaScript source code (any JS framework is good if needed)
• Necessary changes to our client library as a pull request
• Necessary repackaging and workarounds for any library client depends on
• Necessary server-side modification suggestions, like new HTTP headers needed to support client-side downloads
• Any other suggestions that would make pure client-side development easier

Likely changes needed

• Currently using requests for data download - create a new data transport that is compatible with client-side fetch.
• Make imports optional / ignored just to make it start

Possible problematic libraries

These libraries are likely to cause headache:

• pyarrow
• web3.py

Other issues

If the research finds out this is impossible or very difficult e.g. due to some of the key dependencies being impossible hard to make client-side Python compatible, then drop the work and document all blockers.