Skip to content

Latest commit

 

History

History
76 lines (55 loc) · 5.53 KB

README.md

File metadata and controls

76 lines (55 loc) · 5.53 KB

Panda Patrol Panda Patrol

License Python Downloads

Gain greater visibility and context into your data pipelines with dashboards, alerting, silencing and other features built around your existing data tests and data profiling tools — within each step of your data pipelines. Add less than 5 lines of code and just run your pipelines as you normally would. Panda Patrol will take care of the rest.

Questions and feedback

Email: ivanzhangofficial@gmail.com

Table of Contents

Integrations

  • Custom python data pipeline python
  • Airflow airflow
  • Dagster dagster
  • Prefect prefect
  • dbt core (>=1.5) dbt-core
  • dbt cloud dbt-cloud

For examples of each integration, see examples.

Features

This section describes the features of Panda Patrol at a high level. See demo for a short walkthrough of each feature. See wiki to learn how to implement each feature and more details.

AI-Generated Data Tests

Don't know what data tests to write? No problem. Panda Patrol can generate data tests for you. Just pass in the headers, a preview of the data, and optional additional context.

General Data Tests

Want to get started with a few quick, easy, general, and important data tests? Panda Patrol comes pre-built with a few data tests that run on your data. The best part? It only takes one function call to run these tests.

Anomaly Detection

Want to quickly check for anomalies in a column? Panda Patrol can do that for you. Panda Patrol uses the ECOD anomaly detection model from the pyod open source anomaly detection library. Just pass in the excepted distribution of the column and the current distribution of the column. Panda Patrol detect and surface any anomalies. Even better, customize your own anomaly detection model and pass it in to Panda Patrol.

Data Test Results

Write Python-based data tests right within your pipelines. Panda Patrol will store the results of each data test — the test code, logs, return value, start time, end time, exception (if any), and more — in a database. These results can be tracked in a general dashboard (with high level context like test status) and a dashboard for each pipeline run (with all the context w.r.t. the test).

Data Test Parameters

Data changes all the time. Your data tests should change to accomodate these changes. With Panda Patrol, you can pass in parameters to your data tests and later change these parameters on the frontend — with just one function call.

Monitor Data Pipeline Steps

Monitor each step of your pipeline so that you know each step is running as expected. Panda Patrol will store the start time, end time, and status of each step in a database. This gives you a high-level overview of your pipeline and allows you to drill down into each step to see more details.

Alerting

Be notified when your data tests fail. Configure your own email and Slack settings to receive alerts. Alerts provide all the details you see in the dashboards so you get all the context you need to debug pipelines.

Silencing

Want to skip and silence a data test? No problem. Silencing a data test is as easy as clicking a button and choosing a time.

Data Profiles

Using a custom data profiling tool? Or an open-source tool like ydata-profiling? Store data profiles (that are in JSON or HTML format) and check them to see what your data looks like at each step of your pipeline.

The best part? Panda Patrol can be fully self-hosted; this repository contains its backend and frontend code. You can run it on your own infrastructure and have full control over your data. No need to worry about data privacy and security.

Demo

See demo here: https://www.loom.com/share/0468aef48b1843f381146399f1652b81?sid=107df0c0-3e53-4d3c-b9f2-1159d3f23bdf

Getting Started

Check out the Quickstart guide to get started.

You can also look at examples of how Panda Patrol fits into your data pipeline. For example, if you use dagster, see examples/dagster for a guide on how to get started with dagster. All guides should take no longer than 10 minutes to complete. See examples for all examples.

For documentation on how to use Panda Patrol and more details on each feature, see the wiki.