Tangle

What is Tangle?

Tangle is an open source experimentation platform for building machine learning and data pipelines in a visual editor. Shopify engineers built it and released it publicly. You connect components by drag and drop, run them locally or in the cloud, and content-based caching skips any step whose inputs have not changed.

Top Features:

• Visual editor: connect components and configure arguments to build pipelines with or without code.
• Content caching: reuse earlier results automatically so unchanged steps never run twice.
• Any language: wrap existing Python, JavaScript, or Rust code in a container as components.

Use Cases:

• Model experiments: try new features or parameters without rerunning expensive upstream data steps.
• Shared pipelines: clone a teammate's pipeline, change one step, and run it again.
• Data workflows: chain data preparation, training, and evaluation into one repeatable visual flow.

Who Can Use Tangle?

• ML engineers: iterate faster on experiments while keeping every run reproducible and logged.
• Data scientists: build pipelines visually without setting up local development environments first.
• Platform teams: give the whole company one shared tool for ML and data workflows.

Pricing

• Open source (free): the backend and editor code are free to use and self-host.
• Playground (free): a hosted demo on Hugging Face Spaces lets you try the editor instantly.
• Compute (usage based): you pay only your own cloud or Hugging Face Jobs compute costs.

Pros and Cons

Pros:

• Big compute savings: caching saved Shopify over a year of CPU time internally.
• Platform agnostic: runs on local Docker, Hugging Face Jobs, or major cloud providers.
• Reproducible runs: every run and log is stored, making results easy to trace later.

Cons:

• Self hosting work: production use means setting up and maintaining your own infrastructure.
• Container knowledge: wrapping custom code as components assumes some familiarity with containers.
• Young community: support mostly happens through GitHub issues and discussions for now.

FAQs:

1) Who created this platform?

Shopify engineers built it for internal use and later open sourced it.

2) Is it free to use?

Yes, it is open source, and you pay only for your compute.

3) Do I need to write code?

No, pipelines are built visually, though components can wrap your own code.

4) Where can pipelines run?

They run on local Docker, Hugging Face Jobs, or major clouds.

5) How does caching save time?

Steps with unchanged inputs reuse stored results, so only changes rerun.