eda-wizard 0.1.0

Zero-config automated EDA: pass any folder, get charts, stats, and a full HTML report.

⬡ EDA Wizard — Zero-Config Automated EDA

Pass any folder. Get a full EDA report. No setup.

import edawizard
edawizard.eda("any/folder")

---

What it does

EDA Wizard recursively scans a folder, detects every supported file type, runs comprehensive exploratory data analysis on each one, and generates a self-contained report.html plus organized CSVs — all in one call.

Supported file types

Category	Formats
Tabular	CSV, TSV, Excel (.xlsx/.xls), JSON, Parquet, Feather, Pickle
Video	MP4, AVI, MKV, MOV, WMV, FLV, WebM, MPEG, 3GP, TS …
Audio	MP3, WAV, FLAC, AAC, OGG, M4A, WMA, OPUS, AIFF …
Image	PNG, JPG, JPEG, WebP, TIFF, BMP, GIF

---

Installation

Core (tabular data only — zero extra deps):

pip install eda-wizard

With full media support (video + audio + image):

pip install "eda-wizard[full]"

Pick what you need:

pip install "eda-wizard[video]"   # adds opencv-python-headless
pip install "eda-wizard[audio]"   # adds mutagen
pip install "eda-wizard[image]"   # adds Pillow

Video fallback: If opencv-python-headless is not installed, EDA Wizard automatically tries ffprobe (part of FFmpeg) if available on your PATH.

---

Usage

import edawizard

# Scan a folder — all supported files analysed automatically
edawizard.eda("path/to/any/folder")

# Custom output folder name
edawizard.eda("my_data", output_name="eda_results")

# Disable row sampling (analyse all rows, slower on huge files)
edawizard.eda("my_data", max_rows=0)

# Silent mode
edawizard.eda("my_data", verbose=False)

---

Output structure

PyEDA_Output/
├── report.html               ← self-contained interactive report (open in browser)
├── summary.csv               ← overview of all files
│
├── sales/                    ← one folder per tabular file
│   ├── summary_stats.csv
│   ├── missing_values.csv
│   ├── correlation_matrix.csv
│   └── value_counts/
│       ├── region.csv
│       └── category.csv
│
├── media_video/
│   └── metadata.csv          ← FPS, resolution, duration, codec, bitrate per file
│
├── media_audio/
│   └── metadata.csv          ← sample rate, channels, bitrate, duration per file
│
└── media_image/
    └── metadata.csv          ← width, height, megapixels, mode, DPI per file

---

What's in the report

Tabular datasets

• Shape, memory usage, dtype breakdown
• Missing value counts + heatmap (per column)
• Duplicate row detection
• Descriptive statistics (mean, std, percentiles, skew, kurtosis)
• Outlier detection (IQR method) with fences
• Correlation matrix heatmap + top correlated pairs
• Distribution histograms + KDE per numeric column
• Box plots per numeric column
• Top-value bar charts per categorical column
• Zero / negative counts per numeric column

Video datasets

• Duration distribution + total hours
• FPS distribution (discrete bar or continuous histogram)
• Resolution scatter plot (width × height bubble map)
• Aspect ratio breakdown
• Bitrate distribution
• Codec distribution

Audio datasets

• Duration distribution + total hours
• Sample rate (Hz) distribution — discrete bar chart
• Channel layout (Mono / Stereo / Surround) pie chart
• Bitrate distribution
• Codec distribution

Image datasets

• Resolution scatter map
• Megapixel distribution
• Color mode breakdown (RGB, RGBA, Grayscale …)
• Aspect ratio breakdown
• File size distribution

---

How media extraction works

Video  →  opencv-python-headless  →  ffprobe (subprocess)  →  file size only
Audio  →  mutagen                 →  ffprobe (subprocess)  →  file size only
Image  →  Pillow                  →  file size only

EDA Wizard degrades gracefully: if a library isn't installed it tries the next option and prints a tip at startup.

---

Requirements

Package	Purpose	Optional?
pandas	Data loading & stats	No
numpy	Numerics	No
matplotlib	Chart rendering	No
seaborn	Chart styling	No
openpyxl	Excel .xlsx	No
pyarrow	Parquet / Feather	No
opencv-python-headless	Video metadata	Yes (pip install "eda-wizard[video]")
mutagen	Audio metadata	Yes (pip install "eda-wizard[audio]")
Pillow	Image metadata	Yes (pip install "eda-wizard[image]")

---

Contributing

git clone https://github.com/SMFRafin/eda-wizard
cd eda-wizard
pip install -e ".[dev]"
pytest

PRs welcome! Especially for:

• Additional file type support
• LLM-generated insight summaries
• More chart types
• Performance improvements for very large datasets

---

License

MIT