Based on the research and GitHub issues, I can provide you with several solutions to prevent Prefect from interfering with your FastAPI application's logging setup: 1. Use Environment Variables The simplest solution is to set the following environment variable before importing Prefect:

python
import os
os.environ["PREFECT_LOGGING_CONFIGURE_ROOT_LOGGER"] = "False"

# Now import Prefect
from prefect import flow  # or whatever you're importing

2. Custom Logging Configuration You can create a custom logging configuration file at

~/.prefect/logging.yml

that preserves your existing logging setup. Here's a minimal example:

yaml
version: 1
disable_existing_loggers: False  # This is important!

# Your existing handlers and formatters here...

loggers:
    prefect:
        level: "WARNING"  # Or whatever level you prefer
        handlers: ["console"]  # Define your handlers
        propagate: False  # Prevent propagation to root logger

root:
    level: WARNING
    handlers: ["console"]

3. Initialize FastAPI Logging After Prefect If the above solutions don't work, you can try re-initializing your FastAPI logging setup after importing Prefect:

python
import os
os.environ["PREFECT_LOGGING_LEVEL"] = "WARNING"  # Optional: set Prefect's log level

from prefect import flow  # Import Prefect first
import logging
import fastapi
from fastapi.logger import logger as fastapi_logger

# Reconfigure FastAPI logging
logging.basicConfig(
    level=<http://logging.INFO|logging.INFO>,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
)

app = fastapi.FastAPI()

The issue you're experiencing is documented in GitHub issue #9831, which specifically deals with Prefect's interference with uvicorn's logging. Important Notes: 1. The first solution (using

PREFECT_LOGGING_CONFIGURE_ROOT_LOGGER=False

) is the most straightforward and recommended approach. 2. Make sure to set any logging configuration before importing Prefect. 3. If you're using uvicorn's logging features, you might need to explicitly reconfigure uvicorn's logging after setting up Prefect. Would you like me to provide more details about any of these solutions or help you implement a specific approach for your use case?

After thoroughly investigating the source code and related issues, I can now provide a more accurate answer about handling Prefect's logging interference: 1. Current State (Prefect 3.x) The issue you're experiencing is a known problem, and there isn't currently a simple environment variable to disable Prefect's logging configuration completely. The logging setup happens during module import. 2. Working Solutions A. Custom Logging Configuration File Create a

~/.prefect/logging.yml

file that preserves your existing logging setup:

yaml
version: 1
disable_existing_loggers: False  # Critical!

handlers:
    console:
        class: logging.StreamHandler
        level: INFO
        formatter: standard

formatters:
    standard:
        format: "%(asctime)s - %(name)s - %(levelname)s - %(message)s"

loggers:
    prefect:
        level: WARNING
        handlers: [console]
        propagate: False  # Prevent propagation to root logger

    # Your FastAPI logger configuration
    fastapi:
        level: INFO
        handlers: [console]
        propagate: False

    uvicorn:
        level: INFO
        handlers: [console]
        propagate: False

root:
    level: INFO
    handlers: [console]

B. Reconfigure Logging After Import

python
# First, import Prefect
from prefect import flow

# Then reconfigure logging
import logging
logging.basicConfig(
    level=<http://logging.INFO|logging.INFO>,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
    force=True  # This is important!
)

# Now set up FastAPI
from fastapi import FastAPI
app = FastAPI()

3. Related GitHub Issues - There's an open feature request (#10051) to add a way to disable Prefect's logging configuration completely. - The issue is also discussed in #9831 specifically regarding uvicorn logging. 4. Best Practice The most reliable current solution is to use a combination of a custom logging configuration file and explicit reconfiguration after Prefect import. This gives you the most control over your logging setup. Would you like me to provide more specific details about implementing any of these solutions for your particular use case?