Skip to content

Visualise your Kedro data and machine-learning pipelines and track your experiments.

License

Apache-2.0, Unknown licenses found

Licenses found

Apache-2.0
LICENSE.md
Unknown
COPYING.txt
Notifications You must be signed in to change notification settings

kedro-org/kedro-viz

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Kedro-Viz


Kedro-Viz Pipeline Visualisation

Data Science Pipelines. Beautifully Designed
Live Demo: https://demo.kedro.org/


CircleCI npm version Python Version PyPI version License DOI code style: prettier

Introduction

Kedro-Viz is an interactive development tool for building data science pipelines with Kedro. Kedro-Viz also allows users to view and compare different runs in the Kedro project.

Features

  • ✨ Complete visualisation of a Kedro project and its pipelines
  • 🎨 Supports light & dark themes out of the box
  • 🚀 Scales to big pipelines with hundreds of nodes
  • 🔎 Highly interactive, filterable and searchable
  • 🔬 Focus mode for modular pipeline visualisation
  • 📊 Rich metadata side panel to display parameters, plots, etc.
  • 📊 Supports all types of Plotly charts
  • ♻️ Autoreload on code change
  • 🧪 Supports tracking and comparing runs in a Kedro project
  • 🎩 Many more to come

Installation

There are two ways you can use Kedro-Viz:

  • As a Kedro plugin (the most common way).

    To install Kedro-Viz as a Kedro plugin:

    pip install kedro-viz
  • As a standalone React component (for embedding Kedro-Viz in your web application).

    To install the standalone React component:

    npm install @quantumblack/kedro-viz

Usage

CLI Usage

To launch Kedro-Viz from the command line as a Kedro plugin, use the following command from the root folder of your Kedro project:

kedro viz

A browser tab opens automatically to serve the visualisation at http://127.0.0.1:4141/.

Kedro-Viz also supports the following additional arguments on the command line:

Usage: kedro viz [OPTIONS]

  Visualise a Kedro pipeline using Kedro-Viz.

Options:
  --host TEXT               Host that viz will listen to. Defaults to
                            localhost.

  --port INTEGER            TCP port that viz will listen to. Defaults to
                            4141.

  --browser / --no-browser  Whether to open viz interface in the default
                            browser or not. Browser will only be opened if
                            host is localhost. Defaults to True.

  --load-file FILE          Path to load the pipeline JSON file
  --save-file FILE          Path to save the pipeline JSON file
  --pipeline TEXT           Name of the registered pipeline to visualise. If not
                            set, the default pipeline is visualised

  -e, --env TEXT            Kedro configuration environment. If not specified,
                            catalog config in `local` will be used

  --autoreload              Autoreload viz server when a Python or YAML file change in
                            the Kedro project

  -h, --help                Show this message and exit.

Experiment Tracking usage

To enable experiment tracking in Kedro-Viz, you need to add the Kedro-Viz SQLiteStore to your Kedro project.

This can be done by adding the below code to settings.py in the src folder of your Kedro project.

from kedro_viz.integrations.kedro.sqlite_store import SQLiteStore
from pathlib import Path
SESSION_STORE_CLASS = SQLiteStore
SESSION_STORE_ARGS = {"path": str(Path(__file__).parents[2] / "data")}

Once the above set-up is complete, tracking datasets can be used to track relevant data for Kedro runs. More information on how to use tracking datasets can be found here

Notes:

  • Experiment Tracking is only available for Kedro-Viz >= 4.0.2 and Kedro >= 0.17.5
  • Prior to Kedro 0.17.6, when using tracking datasets, you will have to explicitly mark the datasets as versioned for it to show up properly in Kedro-Viz experiment tracking tab. From Kedro >= 0.17.6, this is done automatically:
train_evaluation.r2_score_linear_regression:
  type: tracking.MetricsDataSet
  filepath: ${base_location}/09_tracking/linear_score.json
  versioned: true

Standalone React component usage

To use Kedro-Viz as a standalone React component, import the component and supply a data JSON as prop:

import KedroViz from '@quantumblack/kedro-viz';

const MyApp = () => (
  <div style={{ height: '100vh' }}>
    <KedroViz data={json} />
  </div>
);

The JSON can be obtained by running:

kedro viz --save-file=filename.json

We also recommend wrapping the Kedro-Viz component with a parent HTML/JSX element that has a specified height (as seen in the above example) in order for Kedro-Viz to be styled properly.

Feature Flags

Kedro-Viz uses features flags to roll out some experimental features. The following flags are currently in use:

Flag Description
sizewarning From release v3.9.1. Show a warning before rendering very large graphs (default true)
expandAllPipelines From release v4.3.2. Expand all modular pipelines on first load (default false)

To enable or disable a flag, click on the settings icon in the toolbar and toggle the flag on/off.

Kedro-Viz also logs a message in your browser's developer console to show the available flags and their values as currently set on your machine.

Maintainers

Kedro-Viz is maintained by the kedro team and a number of contributors from across the world.

Contribution

If you want to contribute to Kedro-Viz, please check out our contributing guide.

License

Kedro-Viz is licensed under the Apache 2.0 License.

Citation

If you're an academic, Kedro-Viz can also help you, for example, as a tool to visualise how your publication's pipeline is structured. Find our citation reference on Zenodo.