🔧 Command Line Tools

RexF provides powerful command-line tools for experiment analysis, automation, and integration with other tools.

Installation and Setup

The CLI tools are automatically installed with RexF:

pip install rexf

# Verify installation
rexf-analytics --help

Main Command: rexf-analytics

The rexf-analytics command provides comprehensive experiment analysis capabilities.

Basic Usage

# Show help
rexf-analytics --help

# Quick summary of all experiments
rexf-analytics --summary

# List all experiments
rexf-analytics --list

Common Options

All commands support these common options:

# Specify database file (default: experiments.db)
rexf-analytics --database custom.db --summary

# Limit number of results
rexf-analytics --list --limit 10

# Verbose output for debugging
rexf-analytics --list --verbose

# Filter by experiment name
rexf-analytics --list --experiment-name my_experiment

Core Commands

List Experiments

Display experiment information in various formats:

# Basic list (table format)
rexf-analytics --list

# JSON format
rexf-analytics --list --format json

# CSV format (great for Excel/analysis)
rexf-analytics --list --format csv

# Save to file
rexf-analytics --list --format csv --output experiments.csv

# Filter by experiment name
rexf-analytics --list --experiment-name hyperparameter_search

# Show only recent experiments
rexf-analytics --list --limit 20

Query Experiments

Use expressions to find specific experiments:

# Find high-accuracy experiments
rexf-analytics --query "accuracy > 0.9"

# Find fast experiments
rexf-analytics --query "training_time < 60"

# Complex queries
rexf-analytics --query "accuracy > 0.8 and param_learning_rate < 0.01"

# Parameter queries (use param_ prefix)
rexf-analytics --query "param_batch_size >= 64"

# Save query results
rexf-analytics --query "accuracy > 0.9" --output high_accuracy.csv

Generate Summary

Get quick overview statistics:

# Overall summary
rexf-analytics --summary

# Summary for specific experiment
rexf-analytics --summary --experiment-name ml_experiment

# Summary in JSON format
rexf-analytics --summary --format json

Example output:

📊 Experiment Summary
=====================

Total Experiments: 127
Successful: 119 (93.7%)
Failed: 8 (6.3%)

Average Duration: 145.2 seconds
Total Runtime: 5.1 hours

Top Experiments by accuracy:
1. Run a1b2c3d4: 0.9847
2. Run e5f6g7h8: 0.9823
3. Run i9j0k1l2: 0.9801

Generate Insights

Get intelligent analysis and recommendations:

# Generate insights
rexf-analytics --insights

# Insights for specific experiment
rexf-analytics --insights --experiment-name optimization_experiment

# Save insights to file
rexf-analytics --insights --output insights_report.json

Compare Experiments

Compare multiple experiments side-by-side:

# Compare best experiments
rexf-analytics --compare --best 5

# Compare specific experiments
rexf-analytics --compare --run-ids abc123,def456,ghi789

# Compare experiments from query
rexf-analytics --query "accuracy > 0.95" --compare

Web Dashboard

Launch the web dashboard from command line:

# Launch dashboard
rexf-analytics --dashboard

# Custom host and port
rexf-analytics --dashboard --host 0.0.0.0 --port 9090

# Don't open browser automatically
rexf-analytics --dashboard --no-browser

Advanced Usage

Batch Analysis

Process multiple databases or perform batch operations:

# Analyze multiple databases
for db in experiments_*.db; do
    echo "=== Analysis for $db ==="
    rexf-analytics --database "$db" --summary
done

# Export all data for external analysis
rexf-analytics --list --format csv --output all_experiments.csv
rexf-analytics --insights --output insights.json

Automation and Scripting

Use CLI tools in scripts and automation:

#!/bin/bash

# Automated experiment monitoring script

# Check for new experiments
RECENT_COUNT=$(rexf-analytics --query "start_time > '$(date -d '1 hour ago' --iso-8601)'" --format json | jq length)

if [ "$RECENT_COUNT" -gt 0 ]; then
    echo "Found $RECENT_COUNT new experiments"

    # Generate insights for recent experiments
    rexf-analytics --insights --output "insights_$(date +%Y%m%d_%H%M%S).json"

    # Check for high-performing experiments
    HIGH_PERF=$(rexf-analytics --query "accuracy > 0.95" --format json | jq length)

    if [ "$HIGH_PERF" -gt 0 ]; then
        echo "🎉 Found $HIGH_PERF high-performing experiments!"
        # Send notification, update dashboard, etc.
    fi
fi

Integration with Other Tools

Git Integration

Track experiment results with Git:

# Export current results
rexf-analytics --list --format csv --output experiments_$(git rev-parse --short HEAD).csv

# Add to Git for version tracking
git add experiments_$(git rev-parse --short HEAD).csv
git commit -m "Experiment results for commit $(git rev-parse --short HEAD)"

Continuous Integration

Use in CI/CD pipelines:

# GitHub Actions example
- name: Run Experiments and Analyze
  run: |
    python run_experiments.py
    rexf-analytics --summary
    rexf-analytics --query "accuracy < 0.8" --format json > failing_experiments.json

- name: Upload Results
  uses: actions/upload-artifact@v3
  with:
    name: experiment-results
    path: |
      experiments.db
      failing_experiments.json

External Analysis Tools

Export data for analysis in R, Python, MATLAB:

# For R analysis
rexf-analytics --list --format csv --output experiments_for_r.csv

# For Python/Pandas
rexf-analytics --list --format json --output experiments.json

# For MATLAB
rexf-analytics --list --format csv --output experiments.csv

Output Formats

JSON Format

Structured data perfect for programmatic processing:

rexf-analytics --list --format json

[
  {
    "run_id": "abc123def456",
    "experiment_name": "ml_experiment",
    "status": "completed",
    "start_time": "2024-01-15T10:30:00",
    "duration": 145.2,
    "parameters": {
      "learning_rate": 0.01,
      "batch_size": 32
    },
    "metrics": {
      "accuracy": 0.9547,
      "loss": 0.0453
    }
  }
]

CSV Format

Tabular data ideal for Excel and data analysis:

rexf-analytics --list --format csv

run_id,experiment_name,status,start_time,duration,param_learning_rate,param_batch_size,metric_accuracy,metric_loss
abc123def456,ml_experiment,completed,2024-01-15T10:30:00,145.2,0.01,32,0.9547,0.0453

Table Format

Human-readable tables for terminal display:

rexf-analytics --list --format table

┌──────────┬─────────────────┬───────────┬─────────────────────┬──────────┬──────────┐
│ Run ID   │ Experiment      │ Status    │ Start Time          │ Duration │ Accuracy │
├──────────┼─────────────────┼───────────┼─────────────────────┼──────────┼──────────┤
│ abc123de │ ml_experiment   │ completed │ 2024-01-15 10:30:00 │ 145.2s   │ 0.9547   │
│ def456gh │ ml_experiment   │ completed │ 2024-01-15 11:15:00 │ 132.1s   │ 0.9423   │
└──────────┴─────────────────┴───────────┴─────────────────────┴──────────┴──────────┘

Filtering and Search

Experiment Name Filtering

# Show only specific experiments
rexf-analytics --list --experiment-name hyperparameter_search

# Works with all commands
rexf-analytics --summary --experiment-name ml_experiment
rexf-analytics --insights --experiment-name optimization_study

Status Filtering

# Show only completed experiments
rexf-analytics --query "status == 'completed'"

# Show failed experiments for debugging
rexf-analytics --query "status == 'failed'" --format table

Time-based Filtering

# Recent experiments (last 24 hours)
rexf-analytics --query "start_time > '$(date -d '1 day ago' --iso-8601)'"

# Experiments from specific date range
rexf-analytics --query "start_time between '2024-01-01' and '2024-01-31'"

Performance and Optimization

Large Databases

For databases with many experiments:

# Use limits to avoid overwhelming output
rexf-analytics --list --limit 100

# Use specific queries instead of listing all
rexf-analytics --query "start_time > '2024-01-01'" --limit 50

# Export in chunks for very large datasets
rexf-analytics --query "start_time between '2024-01-01' and '2024-01-07'" --output week1.csv

Caching and Speed

# Use specific queries for faster results
rexf-analytics --query "experiment_name == 'ml_experiment' and status == 'completed'"

# Limit output fields for faster processing
rexf-analytics --list --format csv --output minimal.csv

Common Use Cases

Daily Experiment Review

#!/bin/bash
echo "=== Daily Experiment Review ==="
echo ""

echo "Recent experiments (last 24h):"
rexf-analytics --query "start_time > '$(date -d '1 day ago' --iso-8601)'" --format table

echo ""
echo "Best performers today:"
rexf-analytics --query "start_time > '$(date -d '1 day ago' --iso-8601)' and accuracy > 0.9"

echo ""
echo "Any failures to investigate:"
rexf-analytics --query "start_time > '$(date -d '1 day ago' --iso-8601)' and status == 'failed'"

Research Paper Data Export

# Export specific experiment set for paper
rexf-analytics --query "experiment_name == 'paper_experiments' and accuracy > 0.0" \
               --format csv --output paper_results.csv

# Generate insights for methodology section
rexf-analytics --insights --experiment-name paper_experiments \
               --output paper_insights.json

Hyperparameter Optimization Monitoring

# Monitor optimization progress
rexf-analytics --query "experiment_name == 'hyperopt' and status == 'completed'" \
               --format table

# Find best parameters found so far
rexf-analytics --query "experiment_name == 'hyperopt'" --compare --best 5

# Export for external optimization tools
rexf-analytics --query "experiment_name == 'hyperopt'" \
               --format json --output hyperopt_results.json

Troubleshooting

Common Issues

Database not found:

# Check current directory
ls -la *.db

# Specify full path
rexf-analytics --database /full/path/to/experiments.db --summary

No experiments found:

# Verify database has data
rexf-analytics --list --limit 1

# Check experiment names
rexf-analytics --summary

Query syntax errors:

# Use proper syntax for parameter queries
rexf-analytics --query "param_learning_rate > 0.01"  # Correct
# rexf-analytics --query "learning_rate > 0.01"      # Wrong

# Use quotes for string values
rexf-analytics --query "status == 'completed'"       # Correct
# rexf-analytics --query "status == completed"       # Wrong

Performance Issues

# For large databases, use specific queries
rexf-analytics --query "start_time > '2024-01-01'" --limit 100

# Export data in smaller chunks
rexf-analytics --query "experiment_name == 'specific_exp'" --output chunk.csv

Next Steps

• 📊 Web Dashboard - Interactive visualization and monitoring

• 🔍 Advanced Features - Advanced analysis and exploration

• api/cli - Detailed CLI API reference