dsbro 0.2.1

Your Data Science Bro. One import away.

dsbro

Your Data Science Bro. One import away.

dsbro is an all-in-one Python library for notebook-first data science. It pulls together setup, file I/O, EDA, preprocessing, visualization, metrics, ML baselines, and text helpers into one import with smart defaults and dark-theme visuals.

Installation

pip install dsbro
pip install dsbro[ml]
pip install dsbro[all]

Quick Start

import dsbro
dsbro.setup()
print(dsbro.version())
dsbro.about()

from dsbro import eda
eda.overview(df)

from dsbro import prep
df_small = prep.reduce_memory(df)
df_clean = prep.fill_missing(df)

from dsbro import viz
viz.set_theme("dark")
viz.heatmap(df.corr(numeric_only=True))

from dsbro import ml
results = ml.compare(df, target="target", cv=3)
results.head()

from dsbro import metrics
metrics.regression_report(y_true, y_pred)

Modules

Module	What it does
utils	Notebook setup, seeding, timers, system info, and environment helpers
io	File loading, saving, peeking, searching, and submission utilities
eda	Dataset overview, missing values, correlation, outliers, drift, and profiling
prep	Encoding, scaling, missing-value handling, feature engineering, and memory reduction
viz	Dark-theme charts for tabular analysis and model evaluation
metrics	Quick regression and classification metrics in one place
ml	Model comparison, training, tuning, blending, stacking, and OOF utilities
text	Text cleaning, tokenization, n-grams, word frequency, and TF-IDF features

Why dsbro?

• You stop copy-pasting the same notebook boilerplate for setup, missing values, scaling, and memory reduction.
• You get cleaner charts without writing styling code every time.
• You avoid scattered imports across pandas, seaborn, sklearn, and utility snippets.
• You can benchmark baseline models in one line instead of wiring cross-validation by hand.
• You keep common Kaggle and Colab workflows in one small, consistent package.

For Kaggle Users

Use dsbro to replace the usual notebook starter blocks:

!pip install dsbro[all] -q

import dsbro
dsbro.setup()

Useful first calls:

from dsbro import eda, prep, ml
eda.profile(train_df, target="target")
train_small = prep.reduce_memory(train_df)
leaderboard = ml.compare(train_small, target="target", cv=3)

For Colab Users

Install in the first cell, then keep the rest of the notebook clean:

!pip install dsbro[all] -q

import dsbro
dsbro.setup()

Colab-friendly flow:

from dsbro import io, eda, viz
data = io.load("/content/train.csv")
eda.overview(data)
viz.hist(data, col="target")

Dependencies

Core dependencies:

• numpy
• pandas
• matplotlib
• seaborn
• scikit-learn

Optional extras:

• dsbro[ml]: lightgbm, xgboost, catboost, optuna
• dsbro[plotly]: plotly
• dsbro[all]: all optional extras together

Notebook Example

The project includes a proper tutorial notebook:

• examples/quickstart.ipynb

Development

pytest tests/ -v
ruff check dsbro/ tests/
python -m build

Contributing

We welcome contributions. Read CONTRIBUTING.md before opening a PR.

License

MIT. See LICENSE.

Author

Muhammad Ibrahim Qasmi
Website
GitHub