Skip to content

CLI Reference

DeepFabric's command-line interface provides a modular set of tools that support complex workflows through focused, single-purpose commands.

Command Overview

  • generate

    Complete dataset generation from YAML configuration

    Reference

  • validate

    Configuration validation and problem detection

    Reference

  • visualize

    Topic graph visualization and analysis

    Reference

  • upload-hf

    Hugging Face Hub integration and publishing

    Reference

  • upload-kaggle

    Kaggle integration and publishing

    Reference

  • upload

    DeepFabric Cloud integration and publishing (experimental)

    Reference

  • import-tools

    Import tool definitions from MCP servers

    Reference

  • evaluate

    Model evaluation on tool-calling tasks

    Reference

  • info

    Version and environment information

    Reference

Global Options

All commands support common options for help and version information:

deepfabric --help     # Show command overview
deepfabric --version  # Display version information

Individual commands provide detailed help:

deepfabric generate --help
deepfabric validate --help

Command Composition

Modular Workflow

The modular design enables sophisticated workflows through command composition:

Complete workflow
# Validate configuration
deepfabric validate config.yaml

# Generate the dataset
deepfabric generate config.yaml

# Visualize topic structure (if using graphs)
deepfabric visualize topic_graph.json --output structure.svg

# Upload to Hugging Face
deepfabric upload-hf dataset.jsonl --repo username/dataset-name

Each command operates independently, allowing selective re-execution when iterating on specific aspects of your generation process.

Error Handling

All commands provide detailed error messages with actionable guidance for resolution. Error categories include:

  • Configuration problems
  • Authentication issues
  • API failures
  • File system problems

Exit Codes

Commands use consistent exit codes where success returns 0 and various error conditions return non-zero values, enabling reliable scripting and automated workflows.

Configuration Override Patterns

Many commands accept configuration file arguments along with parameter overrides:

Override config options
deepfabric generate config.yaml \
  --temperature 0.9 \
  --num-samples 50 \
  --provider anthropic \
  --model claude-sonnet-4-5

This pattern supports rapid iteration during development while maintaining reproducible baseline configurations.