I built a tiny tool to convert Pydantic models to TypeScript. What do you think?

Posted by TheFitDev@reddit | Python | View on Reddit | 17 comments

At work we use FastAPI and Next.js, and I often need to turn Pydantic models into TypeScript for the frontend. Doing it by hand every time was boring, slow, and easy to mess up so I built a small app to do it for me.

Paste your Pydantic models/enums, get clean TypeScript interfaces/types instantly.
Runs 100% in your browser (no server, no data saved)
One-click copy or download a .ts file

What My Project Does

My project is a simple website that converts your Python Pydantic models into clean TypeScript code. You just paste your Pydantic code, and it instantly gives you the TypeScript version. It all happens right in your browser, so your code is safe and never saved. This saves you from having to manually type out all the interfaces, which is boring and easy to mess up.

Target Audience

This is for developers who use FastAPI on the backend and TypeScript (with frameworks like Next.js or React) on the frontend. It's a professional tool meant to be used in real projects to keep the backend and frontend in sync.

Comparison

There are other tools out there, but they usually require you to install them and use your computer's command line. My tool is different because it's a website. You don't have to install anything, which makes it super quick and easy to use for a fast conversion. Plus, because it runs in your browser, you know your code is private.

It’s saved me a bunch of time and keeps backend and frontend in sync. If you do the same stack or use typescript, you might find it handy too.
Github: https://github.com/sushpawar001/pydantic-typescript-converter
Check it out: https://pydantic-typescript-converter.vercel.app/
Would love feedback and ideas!

PS: Not gonna lie I have significantly used AI to build this. (Not vibe coded though)

[-]

coderarun@reddit

Did you look into typespec.io? One possibility is to write models using that and generate both python dataclasses and typescript.

You can convert dataclasses to pydantic using a fquery.pydantic decorator. I shared something about it a few months ago here.

[-]

coderarun@reddit

Completely vibe coded: https://github.com/adsharma/typespec-py

[-]

neums08@reddit

FastAPI generates it's own OpenAPI schema, including request and response models. They even document how to generate an entire typescript client from your openapi.json file.

https://fastapi.tiangolo.com/advanced/generate-clients/

[-]

TheFitDev@reddit (OP)

Yes. I knew this. I felt that approach was complicated

[-]

covmatty1@reddit

What did you think was complicated about running a single command?

[-]

nickcash@reddit

Eh, I get it. It's an extra hop and openapi isn't 1:1 with either pydantic models or typescript types. I could see cases where the direct conversion is more useful.

That said I'd stick with the standard tooling unless I had a really compelling reason.

[-]

Explosive_Squirrel@reddit

Check you this project if you want automatic type generation from OpenAPI schemas: https://openapi-ts.dev

[-]

covmatty1@reddit

If something like this required a person to paste things into a browser and copy out the response, it's a complete non starter for me.

That's way too manual. I want something I can call in a CI pipeline that will generate an NPM package and upload it to a registry for me, so that my front-end project is just referencing a package for all of its models. Otherwise there's just significant duplication that's at a greater risk of drifting apart.

[-]

SignificantPound8853@reddit

It works 100% within the browser, does not require a server, and does not store data, so it is secure and likely to be in high demand!

[-]

akx@reddit

Ah yeah, I'm doing the same thing with https://github.com/akx/typtyp but my project is designed to run as part of your dev process, on your dev machine (or in CI/CD).

[-]

crunk@reddit

This seems like a lot of code for this, I'd guess that implementing this in python itself would be way less, since pydantic itself can parse the models.

[-]