run-experiment

name	run-experiment
description	Run a full Energizados training experiment (validate → ETL → train) and surface the key metrics from the JSON report. Use when you want to kick off a pipeline run and see results.

You are running a full Energizados ML experiment.

Step 1: Confirm Scope

Ask the user (skip if already clear from context):

• Run ETL? (default: yes, unless already processed)
• Run training? (default: yes)
• Custom experiment name? (default: auto-generated timestamp)

Step 2: Validate Config

Run config validation first — fail fast before a long training job:

energizados validate etl,train

If validation fails, stop and show the error. Do NOT proceed to step 3.

Step 3: Run ETL (if requested)

energizados run etl

Show a summary of which ETLs ran and their output paths.

Step 4: Run Training

energizados run train

If a custom name was provided:

energizados run train -n {experiment_name}

Step 5: Find and Surface Results

After training completes, find the most recent run directory:

ls -t output/ | head -1

Then read the JSON report:

cat output/{run_dir}/reports/evaluation/evaluation_report.json

Display a clean metrics summary:

## Experiment Results: {run_dir}

| Metric     | Val   | Test  |
|------------|-------|-------|
| AUC        | ...   | ...   |
| Precision  | ...   | ...   |
| Recall     | ...   | ...   |
| F1         | ...   | ...   |

**Threshold used**: {threshold}
**Model**: {model_type}
**Report**: output/{run_dir}/reports/evaluation/evaluation_report.html

Step 6: Next Steps

Suggest:

• Open the HTML report for interactive charts
• Run /ml-config-reviewer if metrics are below expectations
• Tag the run with /commit if results look good