How to define a pydantic.BaseModel that allows both fixed words and a pattern?

I have the following code

from pydantic import BaseModel


class Product(BaseModel):
    name: str
    tags: list[str]  # Accepted tags: "mean", "std", "extra_<float with 1 decimal point>".

The tags is a list of str. But it only accepts certain words.

fixed words: mean and std.
flexible words: it must start with extra_ and be followed by a float with 1 decimal point. For example extra_1.0, extra_2.5. But extra_3.14 is not allowed.

Can you please show me how to implement it?

Thanks.

Solution

EDIT: Thanks to Yurii Motov

1st solution
You can use Union, Literal, Annotated and StringConstraints
(not constr, due to: Pydantic docs, constr will be deleted in 3.0 version)

from typing_extensions import Annotated, Union, Literal 
from pydantic import BaseModel, StringConstraints

class Product(BaseModel):
    name: str
    tags: list[Union[Literal['mean', 'std'],
                  Annotated[str, StringConstraints(pattern=r'exp\_[0-9]\.[0-9]')]]]

if input string doesn't match expectations this error will raise:

Input should be 'mean' or 'std' [type=literal_error,
input_value='template', input_type=str]...
String should match pattern 'exp\_[0-9]\.[0-9]'
[type=string_pattern_mismatch, input_value='template', input_type=str]

or a similar one

2nd solution
Also you can use Annotated and Function validation (docs: Pydantic docs), see here:
My simple example:

def handler(s: str):
    #write the desired handler
    if(not s == 'template'):
        raise TypeError('Not valid string')
    return s

MyStr = Annotated[str, BeforeValidator(handler)]

class Product(BaseModel):
    name: str
    tags: list[MyStr]

# this will raise TypeError, because of int input
# product = Product(name = 'name', tags = [1])

# this will also raise TypeError, that inside handler
# product = Product(name = 'name', tags = ['1','template'])

# this is totally fine
product = Product(name = 'name', tags = ['template','template'])

but you should rewrite handler