Add capability to deploy models remotely

[jakeichikawasalesforce]

Remote Deploy

This PR adds the capability to deploy models to a remote server by creating a client with remote_server=true. See this PR's docs/tabpy-tools.md for usage.

• Pass deploy to _remote_deploy if client is connected to a remote server.
• In _remote_deploy generate a python script to run on the TabPy server: (i) create client instance, (ii) add credentials if necessary, (iii) add source code, (iv) deploy function with user's parameters.
• Then pass to _evaluate_remote_script, which will use the TabPy server's /evaluate endpoint to run the script.

If TABPY_EVALUATE_ENABLE is set to False in the config, this remote deployment will not work, which is the intended behavior (otherwise this would provide a hack where the user can upload arbitrary code and then execute it with the query endpoint).

Other Fixes:

• Add user's docstring to _gen_endpoint, which apparently was not being set at all (always defaulting to "no docstring found"). This uses CustomQueryObject's get_doc_string.
  • However, was running into issues with GitHub's Windows CI/CD Runner, so only fixing this on Mac and Linux for now. This seems to work when ran locally on Windows. Needs more investigation.
• Add time.sleep(interval) to beginning of _wait_for_endpoint_deployment; this helps prevent a 500 error response when the model did indeed deploy.
• Now deleting an endpoint's stage directory as well, when removing an endpoint.
• Include all local log files in .gitignore.

[pep8speaks]

Hello @jakeichikawasalesforce! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2024-11-25 21:44:20 UTC