CLI Usage

Run Scald from the command line for straightforward AutoML tasks.

Basic Command

scald --train <train.csv> --test <test.csv> --target <column> --task-type <type>

All four parameters are required. Task type must be either classification or regression.

Options

--train specifies the training CSV file path. --test specifies the test CSV file path. --target names the target column in training data. --task-type defines the problem type. --max-iterations controls refinement cycles (default: 5).

Examples

Classification:

scald --train data/titanic_train.csv \
      --test data/titanic_test.csv \
      --target survived \
      --task-type classification \
      --max-iterations 5

Regression:

scald --train data/housing_train.csv \
      --test data/housing_test.csv \
      --target price \
      --task-type regression

Output

Scald creates a session directory with logs, artifacts, and predictions:

sessions/session_20250113_143022/
├── session.log
├── artifacts/
└── predictions.csv

Console output shows iteration progress, final metrics, cost, and execution time.

Configuration

Ensure .env contains API credentials:

OPENROUTER_API_KEY=your_api_key
OPENROUTER_BASE_URL=https://openrouter.ai/api/v1

Help

View all options:

scald --help

Troubleshooting

"API key not found" indicates missing OPENROUTER_API_KEY in .env. "File not found" means incorrect CSV paths. "Invalid task type" requires using classification or regression.

Continue to Python API for programmatic usage.