📊 B-vista

Visual, Scalable, and Real-Time Exploratory Data Analysis for pandas DataFrames.

B-Vista is a high-performance EDA (Exploratory Data Analysis) toolkit that connects your pandas DataFrame to a dynamic, modern browser UI powered by Flask and WebSockets. Ideal for Jupyter notebooks, Google Colab, Kaggle, or standalone use.

🔧 Getting Started

Installation

Install via PyPI:

pip install bvista

Or via Conda:

conda install -c conda-forge bvista

Launch in Notebook

import bvista
import pandas as pd

df = pd.read_csv("your_dataset.csv")
bvista.show(df)

By default, B-Vista will auto-launch a backend server and open the interface inline (Jupyter) or in the browser.

🌟 Core Concepts

Sessions

Each DataFrame uploaded creates a unique session. Sessions are isolated, reusable, and persistent while the backend is alive.

Backend Auto-Start

B-Vista auto-launches a Flask server the first time you call bvista.show(df). No extra configuration needed.

Notebook vs Browser

• In Jupyter/Colab, B-Vista embeds via iframe.
• In Python scripts or terminals, the default is to open in your system browser.

📊 Main Function: bvista.show()

bvista.show(
    df,                  # pandas DataFrame
    name=None,           # Optional session name
    session_id=None,     # Reconnect to previous session
    open_browser=True,   # Force open in web browser
    silent=False         # Suppress logs
)

Examples

bvista.show(df)

bvista.show(df, name="marketing_campaign")

bvista.show(session_id="abc123")  # Reconnect to saved session

🔄 Session Management

Upload New Dataset

Each call to show(df) creates a new session unless a session_id is provided.

Reconnect to Old Session

bvista.show(session_id="session_id_you_saved")

💡 Key Features

B-Vista gives you:

• Descriptive statistics (skew, kurtosis, normality)
• Correlation heatmaps (7+ methods)
• Distributions with auto binning + log scaling
• Missing value visualizations (matrix, heatmap, MCAR/MAR/NMAR diagnostics)
• Smart cleaning tools (mean/mode/knn/autoencoder...)
• Column transforms (rename, reorder, format)
• Duplicate detection and resolution
• Cell-level editing with real-time WebSocket sync

🔍 See full breakdown →

📒 API Reference

bvista.show()

Launches the UI interface. Must pass either:

• A pandas DataFrame (df), or
• A session ID (session_id)

bvista.start_backend()

Programmatically start the backend server if needed.

from bvista.server_manager import start_backend
start_backend()

📎 Integration

Google Colab

!pip install bvista --quiet
import bvista, pandas as pd
df = pd.read_csv("...")
bvista.show(df)

Kaggle

Works out-of-the-box. Use same pattern as Colab.

🛠️ Troubleshooting

🛋️ Deployment

Docker Usage

docker pull baciak/bvista:latest
docker run --platform linux/amd64 -p 8501:5050 baciak/bvista:latest

Then visit http://localhost:8501 in your browser.

🔹 Final Notes

• Source:
• PyPI:
• Docker:
• License:

✨ consider starring the repo or sharing with your team!