Overview
During each test run, two files are generated: componentResults.jsonl and stateComponentResults.json. These files are primarily used to train maintenance and optimizer models and to serve as inputs for those services. They provide a detailed record of the runtime execution, which is essential for debugging tests and training models.
How It Works
The runtime execution is composed of modular, nestable component modules. Each module produces a component result object upon execution, which includes a result, events, and results from any child components. This creates a comprehensive tree of results for the entire run.
UI Display
This information is visible in the UI under a new tab called "Component model". This tab is located on the test detail page (when expanding the action or result tabs) and on the slider page's bottom panel.
The tab contains a table that displays:
- Component name
- Elapsed Time
- Confidence score
If a component's confidence score is less than one, an event with details about the issue is also displayed, providing visibility into what happened during the run.
Limitations
The component results data is intended for backend processes like debugging and model training. While visible in the UI, it may not be stable enough for permanent front-end display, although this is a future goal.
Related Info
The raw component results data can be referenced from the componentResults.jsonl file.