kaggle-utils-dataset 0.1.0b0

Automated workflow for dataset discovery and downloading in ML/DL and data analysis projects.

kaggle_utils (beta 0.1.0)

Utilities for checking, downloading, and extracting Kaggle datasets within ML/DL and data analysis workflows.

---

Features

• Check whether a dataset directory exists and list contained files
• Download and unzip Kaggle datasets using Kaggle CLI
• Configurable dataset root directory via environment variable
• Lightweight, reusable design for notebook and local workflows

---

Requirements

• Python 3.10+
• Kaggle CLI installed and authenticated

Install Kaggle CLI

pip install kaggle

Ensure the kaggle command is available in your PATH.

Configure your Kaggle API token:

1. Download kaggle.json from your Kaggle account
2. Place it in:

~/.kaggle/kaggle.json

Ensure correct permissions:

chmod 600 ~/.kaggle/kaggle.json

---

Installation

Local (Editable Mode – Recommended for Development)

From the project root:

pip install -e .

---

Configuration

Dataset root directory defaults to:

datasets

Override it via environment variable:

macOS / Linux:

export KAGGLE_INPUT_DIR="datasets"

Windows (PowerShell):

setx KAGGLE_INPUT_DIR "datasets"

---

Usage

1️⃣ Check Local Dataset

from kaggle_utils import check

result = check("titanic")

if not result:
    print("Dataset not found or empty.")
else:
    print(result)

Example return:

{
    "datasets/titanic": ["train.csv", "test.csv"]
}

---

2️⃣ Download Dataset

from kaggle_utils import download

files = download(owner="heptapod", dataset="titanic")

print(files)

If the dataset does not exist or the CLI fails,
a RuntimeError will be raised.

---

Error Handling

• check() → returns empty dict {} if dataset is missing
• download() → raises RuntimeError on failure
• Kaggle Notebook environments are not supported

---

Project Structure

kaggle_utils/
  __init__.py
  config.py
  check.py
  download.py

---

Design Philosophy

• Separation of concerns
• Minimal external dependencies
• Clear failure handling strategy
• Designed for local ML experimentation

---

Roadmap

beta 0.2.0

• download_if_not_exists() orchestration helper
• Logging support
• Retry mechanism

1.0.0

• Stable API
• PyPI publication

---

License

Distributed under the MIT License. See the LICENSE for more information.

---