CROAK Installer

🐸 CROAK - Computer Recognition Orchestration Agent Kit

The official installer and CLI for CROAK, an agentic framework for object detection model development.

Quick Start

# Install from npm
npx croak-cv init

# Or install globally
npm install -g croak-cv
croak init

What is CROAK?

CROAK is an open-source agentic framework that guides you through the entire object detection pipeline:

• Data Preparation - Scan, validate, and prepare your datasets
• Annotation - Integration with vfrog.ai for professional labeling
• Training - Guided model training with GPU support (local or Modal.com)
• Evaluation - Comprehensive model analysis and diagnostics
• Deployment - Deploy to cloud (vfrog.ai) or edge (TensorRT/ONNX)

CLI Commands

| Command | Description | |---------|-------------| | croak init | Initialize CROAK in current directory | | croak doctor | Check environment and dependencies | | croak upgrade | Upgrade to latest version | | croak help | Show help |

croak init

Initializes a new CROAK project with interactive configuration:

croak init

Options:

• -y, --yes - Skip prompts, use defaults
• --name <name> - Set project name
• --no-vfrog - Skip vfrog.ai integration
• --no-modal - Skip Modal.com GPU setup

What it creates:

your-project/
├── .croak/
│   ├── config.yaml          # Project configuration
│   ├── pipeline-state.yaml  # Pipeline progress tracking
│   ├── agents/              # Agent definitions
│   ├── workflows/           # Workflow specifications
│   ├── knowledge/           # Knowledge base
│   └── contracts/           # Handoff contracts
├── data/
│   ├── raw/                 # Raw images
│   └── processed/           # Processed datasets
├── training/
│   ├── configs/             # Training configurations
│   ├── scripts/             # Training scripts
│   └── experiments/         # Experiment outputs
├── evaluation/
│   └── reports/             # Evaluation reports
└── deployment/
    └── edge/                # Edge deployment packages

croak doctor

Checks your environment for compatibility:

croak doctor

Checks performed:

• Python 3.10+ installation
• Required Python packages (ultralytics, torch, etc.)
• NVIDIA GPU availability
• Modal.com configuration
• vfrog.ai API key
• Git installation

Options:

• --fix - Attempt to fix issues automatically

croak upgrade

Upgrades CROAK to the latest version:

croak upgrade

Options:

• --check - Check for updates without installing

Requirements

• Node.js 18.0.0 or higher
• Python 3.10 or higher
• Git (recommended)
• NVIDIA GPU (optional - can use Modal.com for cloud GPU)

Environment Variables

| Variable | Description | Required | |----------|-------------|----------| | VFROG_API_KEY | vfrog.ai API key for annotation | For annotation | | MODAL_TOKEN_ID | Modal.com token (set via modal setup) | For cloud GPU |

Configuration

After initialization, edit .croak/config.yaml to customize:

version: "1.0"

project:
  name: "my-detection-project"
  task_type: "detection"

training:
  framework: "ultralytics"
  architecture: "yolov8s"
  epochs: 100
  batch_size: 16

compute:
  provider: "modal"  # or "local"
  gpu_type: "T4"

tracking:
  backend: "mlflow"  # or "wandb"

Next Steps After Init

1. Add your images to data/raw/

2. Scan your data:

   croak scan

3. Prepare your dataset:

   croak prepare

4. Train your model:

   croak train

5. Evaluate results:

   croak evaluate

6. Deploy:

   croak deploy

Troubleshooting

Python not found

Ensure Python 3.10+ is installed and in your PATH:

python3 --version

vfrog API key not working

1. Verify the key at https://vfrog.ai/settings/api
2. Ensure the environment variable is set:

   echo $VFROG_API_KEY

No GPU detected

CROAK will automatically use Modal.com for cloud GPU training. Run:

pip install modal
modal setup

Links

• Documentation: https://github.com/vfrog-ai/croak
• Issues: https://github.com/vfrog-ai/croak/issues
• vfrog.ai: https://vfrog.ai

License

MIT License - see LICENSE for details.

🐸 CROAK — by vfrog.ai