Dynamic / recursive dependency resolution based on configuration with unknown sub-dependencies

[saevarb]

First Check

• I added a very descriptive title here.
• I used the GitHub search to find a similar question and didn't find it.
• I searched the FastAPI documentation, with the integrated search.
• I already searched in Google "How to X in FastAPI" and didn't find any information.
• I already read and followed all the tutorial in the docs and didn't find an answer.
• I already checked if it is not related to FastAPI but to Pydantic.
• I already checked if it is not related to FastAPI but to Swagger UI.
• I already checked if it is not related to FastAPI but to ReDoc.

Commit to Help

• I commit to help with one of those options 👆

Example Code

class EnvDependency:
    def __init__(self, env_var: str):
        self.env_var = env_var

    def __call__(self):
        import os

        return os.getenv(self.env_var)


class SomeDependency:
    def __init__(self, name: str):
        self.name = name

    def __call__(self):
        return self.name


ADependency = Annotated[Callable[[], None], Depends(SomeDependency("a"))]
BDependency = Annotated[Callable[[], None], Depends(SomeDependency("b"))]


def get_dep_a(dep: ADependency):
    return "A"


def get_dep_b(dep: BDependency):
    return "B"


import os

os.environ["CONFIG"] = "a"


# The problem with this setup is that I have to anticipate ahead of time that
# I will either need dependency A or dependency B
# This is roughly what we are already doing
def configurable_dependency(
    config_value: Annotated[str, Depends(EnvDependency("CONFIG"))],
    a_dep: ADependency,
    b_dep: BDependency,
):
    print("config_value", config_value)
    if config_value == "a":
        return get_dep_a(a_dep)
    else:
        return get_dep_b(b_dep)


def configurable_dependency_ideal(
    config_value: Annotated[str, Depends(EnvDependency("CONFIG"))],
):
    ...
    # dynamically resolve the dependency(using e.g. importlib) at runtime
    # and just return the dependency, letting FastAPI resolve it further
    if config_value == "a":
        return ADependency
    else:
        return BDependency


ConfigurableDependency = Annotated[
    Callable[[], None], Depends(configurable_dependency)
]

Description

Hi.

We are trying to build a modular application which can be deployed for different customers in different configurations. The idea is that we will have a library of various dependencies for each different customer and situation. Customer 1 may require dependencies A, B and C and customer 2 may only require dependency C.

A developer would build the customer-specific module which would use the dependencies it requires, and then this module would be loaded and created as a dependency at runtime based on, say, environment variables.

The problem with our current approach is that we will always need to know all the dependencies that may be required ahead of time, if I'm not mistaken. I've attempted to demonstrate this in the example code.

Ideally, it would be possible for us to have a dependency which itself returns a Depends(dynamic_dependency), and then FastAPI would realize that it needs to resolve this dependency further and automatically resolve all the parameters to dynamic_dependency(since they would all be Annotated[something, Depends(..)]).

I've pored over the docs and also searched online as well as in the repo for something similar, but I've not found something exactly like this. Is there a way to accomplish this? Can I perhaps manually invoke the FastAPI dependency resolution functionality somehow?

Operating System

Linux

Operating System Details

No response

FastAPI Version

0.104.1

Pydantic Version

1.10.13

Python Version

3.11.7

Additional Context

No response