How to debug Fastapi openapi generation error

Question

I spend some time going over this error but had no success.

File "C:\Users\ebara.conda\envs\asci\lib\site-packages\fastapi\openapi\utils.py", line 388, in get_openapi flat_models=flat_models, model_name_map=model_name_map

File "C:\Users\ebara.conda\envs\asci\lib\site-packages\fastapi\utils.py", line 28, in get_model_definitions model_name = model_name_map[model]

KeyError: <class 'pydantic.main.Body_login_access_token_api_v1_login_access_token_post'>

The problem is that I'm trying to build a project with user authentication from OpenAPI form to create new users in database.

I've used backend part of this template project https://github.com/tiangolo/full-stack-fastapi-postgresql

Everything works except for Authentication like here.

@router.post("/login/access-token", response_model=schemas.Token)
def login_access_token(
    db: Session = Depends(deps.get_db), form_data: OAuth2PasswordRequestForm = Depends()) -> Any:

When I add this part form_data: OAuth2PasswordRequestForm = Depends() - and go to /docs page - this error appears (Failed to load API definition. Fetch error. Internal Server Error /openapi.json)

.

The server itself runs in normal mode, but it can't load the open API. If I remove the aforementioned formdata part - then everything works smoothly, but without Authorisation. I tried to debug it, but I have no success. I think it might be connected to a dependency graph or some start-up issues, but have no guess how to trace it back.

Here is the full working example which will reproduce the error. The link points out the code which causes the problem. If you will comment out lines 18-39 - the docs will open without any problems. https://github.com/BEEugene/fastapi_error_demo/blob/master/fastapi_service/api/api_v1/endpoints/login.py

Any ideas on how to debug or why this error happens?

Answer 1

It seems that in my case - the main error issue was that I was an idiot.

As said, if you will comment out lines 18-39 - the docs will open without any problems. But, you will notice this warning:

UserWarning: Duplicate Operation ID read_users_api_v1_users__get for function read_users at ...\fastapi_error\fastapi_service\api\api_v1\endpoints\users.py warnings. Warn(message)

I've started to compare all the files and it is appeared that I included router to the fastapi twice:

import logging
from fastapi import FastAPI
from starlette.middleware.cors import CORSMiddleware

from fastapi_service.api.api_v1.api import api_router
from fastapi_service.core.config import settings
from fastapi_service.core.event_handlers import (start_app_handler,
                                                 stop_app_handler)

log = logging.getLogger(__name__)

def get_app(mode="prod") -> FastAPI:

    fast_app = FastAPI(title=settings.PROJECT_NAME,
                       version=settings.APP_VERSION,
                       debug=settings.IS_DEBUG)
                       # openapi_url=f"{settings.API_V1_STR}/openapi.json")
    #  first time when I included the router
    fast_app.include_router(api_router, prefix=f"{settings.API_V1_STR}")
    fast_app.mode = mode
    logger = log.getChild("get_app")
    logger.info("adding startup")
    fast_app.add_event_handler("startup", start_app_handler(fast_app))
    logger.info("adding shutdown")
    fast_app.add_event_handler("shutdown", stop_app_handler(fast_app))

    return fast_app


app = get_app()

# Set all CORS enabled origins
if settings.BACKEND_CORS_ORIGINS:
    app.add_middleware(
        CORSMiddleware,
        allow_origins=[str(origin) for origin in settings.BACKEND_CORS_ORIGINS],
        allow_credentials=True,
        allow_methods=["*"],
        allow_headers=["*"],
    )
#  second time when I included the router
app.include_router(api_router, prefix=settings.API_V1_STR)

So, if you comment out (or just delete) the second router introduction - the app will work normally.

It seems, that the answer to my question on how to debug this error - is to find the point where the bug appears in fastapi and compare the values in it to the version where there is no error. In my case the number of keys in different dictionaries differed in the function get_model_definitions.