Add Initial Support for Post-Processing Functions

[johnwalz97]

Internal Notes for Reviewers

Initial support for post-processing functions along with an example notebook with a lot of examples.

External Release Notes

Adding support for post-processing functions to allow users to modify test outputs without modifying test code.

[github-actions]

PR Summary

This pull request introduces several enhancements to the ValidMind project, focusing on post-processing capabilities and raw data handling. Key changes include:

1. Post-Processing Functions: The run_test function now accepts a post_process_fn parameter, allowing users to define custom post-processing functions that modify the TestResult object after a test is run. This enables users to customize test outputs, such as modifying tables, adding/removing tables, and creating figures from tables.

2. Raw Data Handling: A new RawData class is introduced to encapsulate raw data produced during test execution. This class allows for inspection and serialization of raw data, providing a structured way to handle additional data that is not directly displayed in test results.

3. Notebook Addition: A new Jupyter notebook, post_processing_functions.ipynb, is added to demonstrate the use of post-processing functions with ValidMind tests. The notebook includes examples of modifying tables, adding/removing tables, creating figures from tables, and more.

4. Python Version Update: The GitHub Actions workflow is updated to use Python 3.11 instead of 3.8, ensuring compatibility with the latest Python features and improvements.

5. ROC Curve Enhancement: The ROCCurve test now returns both a RawData object and a Plotly figure, allowing for more flexible post-processing and visualization of ROC curve data.

6. Code Refactoring: Various code improvements, including the addition of a HumanReadableEncoder for better JSON serialization of numpy arrays, and refactoring of figure handling logic to support multiple figure types.

These changes enhance the flexibility and usability of the ValidMind library, allowing for more customized and detailed test result handling.

Test Suggestions

• Test the new post-processing function capability by creating a custom function that modifies a test result and verify the changes.
• Verify that the RawData class correctly encapsulates and serializes raw data from a test.
• Run the post_processing_functions.ipynb notebook to ensure all examples execute without errors and produce the expected outputs.
• Test the updated ROCCurve test to ensure it returns both the RawData and the Plotly figure correctly.
• Check the GitHub Actions workflow to confirm it runs successfully with Python 3.11.

[AnilSorathiya]

@johnwalz97 Awesome 👍
There are a couple of things:

1. Are we using key anywhere? If not why don't we create separate ticket get rid off it?
2. Do we need ref_id for RawData to access it on platform/frontend? Since we are using ref_id for tables and figures.

[johnwalz97]

@johnwalz97 Awesome 👍

There are a couple of things:

1. Are we using key anywhere? If not why don't we create separate ticket get rid off it?

2. Do we need ref_id for RawData to access it on platform/frontend? Since we are using ref_id for tables and figures.

Thanks for reviewing @AnilSorathiya

1. Thats a great idea, I can add a ticket for that.
2. We don't need it yet since ref_id is only used to link figures to test results on the backend (since they are logged stored separately in the DB). If we ever start storing raw_data outside of the test_results table than we can add it.

[AnilSorathiya]

left one last comment otherwise it looks good.