CLI Commands Reference¶

Quick reference for Arc commands.

Command Overview¶

/ml - Machine learning operations (data, model, evaluate, predict, jobs)
/sql - SQL queries
/config - Configuration
/report - Generate reports
/help - Show help
/clear - Clear context

Starting Arc¶

arc chat                      # Interactive mode
arc --help                    # Show help
arc --version                 # Show version

ML Commands¶

/ml data¶

Create data processing pipeline.

/ml data --name <name>
         --instruction "<description>"
         --source-tables <tables>

Example:

/ml data --name processed_features
         --instruction "Normalize features and split data"
         --source-tables raw_data

/ml model¶

Train a model.

/ml model --name <name>
          --instruction "<description>"
          --data-table <table>
          [--plan-id <id>]

Example:

/ml model --name churn_predictor
          --instruction "Binary classifier"
          --data-table processed_features

/ml evaluate¶

Evaluate trained model.

/ml evaluate --model-id <id>
             --data-table <table>
             [--metrics "<metrics>"]

Available metrics: - Classification: accuracy, precision, recall, f1, auc - Regression: mse, rmse, mae, r2

/ml predict¶

Make predictions.

/ml predict --model <name>
            --data <table>
            [--output <table>]

/ml jobs¶

Manage jobs.

/ml jobs list              # List all jobs
/ml jobs status <job_id>   # Show job details

SQL Commands¶

Execute SQL queries:

/sql <query>

Common examples:

/sql SHOW TABLES
/sql DESCRIBE table_name
/sql SELECT * FROM table_name LIMIT 10
/sql CREATE TABLE new AS SELECT * FROM old WHERE condition

Configuration¶

/config

Interactive prompts for: - API Key - Base URL - Model name

Settings saved to ~/.arc/user-settings.json.

Alternative: Use environment variables:

export ARC_API_KEY="your-key"
export ARC_BASE_URL="https://api.openai.com/v1"
export ARC_MODEL="gpt-4"

Other Commands¶

/report              # Generate ML workflow report
/help                # Show available commands
/clear               # Clear conversation context
exit                 # Exit Arc

Natural Language¶

You can also use plain English instead of commands:

"Clean up data in customers table"
"Build a model to predict churn"
"Evaluate my model on test data"

Need Help?¶

Run /help in Arc for interactive help
Run arc --help for command-line options
See User Guides for detailed workflows