Move settings to pydantic-settings

[samuelcolvin]

Removing settings from pydantic as settings will be a separate package pydantic-settings.

We might make it installable via pydantic[settings], not sure yet if this is a good idea?

skip change file check.

[samuelcolvin]

Unfortunately we have a slight problem - someone has already taken the package name pydantic-settings, I've emailed them asking if they could change their package name.

[samuelcolvin]

otherwise should be ready.

[hramezani]

Unfortunately we have a slight problem - someone has already taken the package name pydantic-settings, I've emailed them asking if they could change their package name.

Do you have other options in your mind?

[samuelcolvin]

He has replied and kindly agreed to release the name, he's just getting his 2fa reset.

[PrettyWood]

He has replied and kindly agreed to release the name, he's just getting his 2fa reset.

Great!

---

According to the description of author's package

but from my point it’s missing some useful features:

• Overriding settings values by environment variables even for nested fields
• Providing detailed information about value location inside a loaded file or environment variable, which helps to point user mistake
• Documenting model fields isn’t feels comfortable, but it’s really essential to write comprehensive documentation for application settings

I reckon it'd be great to work together with author on the future of (official) pydantic-settings

[samuelcolvin]

Ye, I'd love him to contribute, he says he has no plans to continue to work on the package though.

But maybe joining the team here will tempt him 😉.

[samuelcolvin]

See https://github.com/pydantic/pydantic-settings, I've moved the existing logic to a new package.

[caniko]

+1 for pyndatic[settings]

[zmievsa]

@samuelcolvin you probably already know this but better be safe than sorry:

If the person does not reply or does not want to give you the package, you can try filling out a PEP 541 request here.

[ofek]

Can we look into finally supporting this? 🙂 #1676

[samuelcolvin]

Can we look into finally supporting this? 🙂 #1676

Yea definitely, V2 is an opportunity to fix all the messy stuff we couldn't fix in minor releases.

[ofek]

pydantic/pydantic-settings#1

[danields761]

@Ovsyanka83, relax, I'll transfer pypi name via my good will, and Samuel didn't sent me any blackmails ;)

@PrettyWood , nice to hear that! If some of my original ideas could be integrated into new pydantic-settings, I'll gladly contribute

@samuelcolvin , is there features of mine pydantic-settings library, which you would like to pull into new one?

[zmievsa]

@danields761 I apologize :)

I was writing this in case the package was name-squatted by someone. Thanks for transferring it!

[samuelcolvin]

@danields761, thanks so much.

Yes, I think so. Like @PrettyWood said, we definitely want to have better support for environment variables filling nested fields.

I also think it would be really useful to provide details on where field are populated from as you mention.

But I need to go through this in more depth and write a proposal, please keep your pydantic-settings repo in place so I can review it further while writing the proposal.

[samuelcolvin]

Oh, forgot to add @danields761, hapy to make you a collaborator on the pydantic/pydantic-settings repo if you'd like.

[danields761]

@samuelcolvin , that would be nice, thanks!

Please, let me know, when you'll be ready

[samuelcolvin]

Thanks @danields761, I've sent you an email.

[hmeine]

FWIW, I have been using https://github.com/danields761/pydantic-settings/ mostly for its with_attrs_docs decorator.

It greatly reduces the visual noise / clutter when documenting fields that otherwise do not even need a Field().

Hence, I would love to see this come upstream (not caring so much about the actual syntax / import).

[samuelcolvin]

Thanks @hmeine, without looking into with_attrs_docs, I'd be happy to try and support it.

Could you create an issue on https://github.com/pydantic/pydantic-settings so we don't forget this.

[maartenbreddels]

We might make it installable via pydantic[settings]

This could have made it possible to make libraries compatible with pydantic 1 and 2. Now we need to add the following requirement:

• "pydantic-settings; python_version > '3.6'",

Which makes our libraries implicitly dependent on pydantic 2.

If we could add pydantic[settings], pip would just ignore it for pydantic 1 (ok, it gives a warning, but it makes it possible for users to still install pydantic 1 is needed for some reason).

See also #6497