Logging System¶

The platform includes a dual-stream logging system in src/core/logging/logger.py: 1. Interactive Console Stream: Real-time output formatted with ANSI color codes. 2. Persistent File Stream: Timestamped log files saved in src/core/logs/.

1. Colorized Terminal Output¶

The ColoredFormatter colors output by severity level so you can spot errors or milestones instantly:

Level	Color	Meaning
`DEBUG`	Blue	Detailed internal diagnostic messages (data shapes, step creation).
`INFO`	Green	Major milestones (data loaded, training complete, evaluation scores).
`WARNING`	Yellow	Recoverable issues or fallbacks (e.g. using default split ratio).
`ERROR`	Red	Failures or exceptions (e.g. unconfigured model type).
`CRITICAL`	Magenta	Fatal crashes.

2. Setting the Log Level¶

You control the console verbosity with the --logging-level command-line flag:

# Detailed step-by-step diagnostic logs
uv run python -m main --config configs/mvp.yaml --logging-level DEBUG

# Standard progress logs (default)
uv run python -m main --config configs/mvp.yaml --logging-level INFO

# Warnings and errors only
uv run python -m main --config configs/mvp.yaml --logging-level WARNING

3. Persistent File Logs¶

Every execution creates a timestamped log file in src/core/logs/:

src/core/logs/
├── logs_2026_10_14_23_28.log
├── logs_2026_10_15_00_12.log
└── ...

Each log line records:

YYYY-MM-DD HH:MM:SS : [LEVEL] : module_name : [function_name] : line_number : message

4. Automatic MLflow Upload¶

When the run finishes, the Orchestrator.__exit__ method automatically takes the active log file from src/core/logs/ and uploads it to MLflow under the logs/ artifact folder:

self.mlflow_tracker.log_artifact(
    console.handlers[1].baseFilename,
    f"logs/{self.mlflow_tracker.run.info.run_name}.log",
)

You can view the full log file directly from the MLflow web UI under the Artifacts tab!