Skip to content

Coordinate and use AI in your dataflow with prefect-openai

[!WARNING] prefect-openai is no longer actively maintained by PrefectHQ. We encourage the community use marvin instead. For more details, please see the Maintenance Status section below.


Maintenance Status

prefect-openai has been a valuable part of the Prefect ecosystem. Due to shifts in our strategic priorities, we have decided to discontinue the active maintenance of this library. While we will not be updating the code or addressing issues, the existing codebase will remain accessible for archival purposes. We appreciate the support and contributions from our community.

The prefect-openai collection makes it easy to leverage the capabilities of AI in your flows. Check out the examples below to get started!

Getting Started

Summarize tracebacks with GPT3

Tracebacks--it's quintessential in programming. They are a record of every line of code leading to the error, to help us, humans, determine what's wrong with the program and find a solution.

However, tracebacks can be extraordinarily complex, especially for someone new to the codebase.

To streamline this process, we could add AI to the mix, to offer a more human-readable summary of the issue, so it's easier for the developer to understand what went wrong and implement a fix.

After installing prefect-openai and saving an OpenAI key, you can easily incorporate OpenAI within your flows to help you achieve the aforementioned benefits!

from prefect import flow, get_run_logger
from prefect_openai import OpenAICredentials, CompletionModel

def summarize_traceback(traceback: str) -> str:
    logger = get_run_logger()
    openai_credentials = OpenAICredentials.load("openai-credentials")
    completion_model = CompletionModel(
    prompt = f"Summarize cause of error from this traceback: ```{traceback}```"
    summary = completion_model.submit_prompt(traceback).choices[0]["text"]"Summary of the traceback: {summary}")
    return summary

if __name__ == "__main__":
    traceback = """
        ParameterBindError: Error binding parameters for function 'summarize_traceback': missing a required argument: 'traceback'.
        Function 'summarize_traceback' has signature 'traceback: str) -> str' but received args: () and kwargs: {}.

12:29:32.085 | INFO    | Flow run 'analytic-starling' - Finished text completion using the 'text-curie-001' model with 113 tokens, creating 1 choice(s).
12:29:32.089 | INFO    | Flow run 'analytic-starling' - Summary of the traceback:     
This error is caused by the missing argument traceback. The function expects a traceback object as its first argument, but received nothing.
Notice how the original traceback was quite long and confusing.

On the flip side, the Curie GPT3 model was able to summarize the issue eloquently!

Built-in decorator

No need to build this yourself, prefect-openai features a built-in decorator to help you automatically catch and interpret exceptions in flows, tasks, and even vanilla Python functions.

import httpx
from prefect_openai.completion import interpret_exception

def example_func():
    resp = httpx.get("")


Create a story around a flow run name with GPT3 and DALL-E

Have you marveled at all the AI-generated images and wondered how others did it?

After installing prefect-openai and saving an OpenAI key, you, too, can create AI-generated art.

Here's an example on how to create a story and an image based off a flow run name.

from prefect import flow, task, get_run_logger
from prefect.context import get_run_context
from prefect_openai import OpenAICredentials, ImageModel, CompletionModel

def create_story_from_name(credentials: OpenAICredentials, flow_run_name: str) -> str:
    Create a fun story about the flow run name.
    text_model = CompletionModel(
        openai_credentials=credentials, model="text-curie-001", max_tokens=288
    text_prompt = f"Provide a fun story about a {flow_run_name}"
    story = text_model.submit_prompt(text_prompt).choices[0].text.strip()
    return story

def create_image_from_story(credentials: OpenAICredentials, story: str) -> str:
    Create an image associated with the story.
    image_model = ImageModel(openai_credentials=credentials, size="512x512")
    image_result = image_model.submit_prompt(story)
    image_url =[0]["url"]
    return image_url

def create_story_and_image_from_flow_run_name() -> str:
    Get the flow run name and create a story and image associated with it.
    context = get_run_context()
    flow_run_name ="-", " ")

    credentials = OpenAICredentials.load("openai-credentials")
    story = create_story_from_name(credentials=credentials, flow_run_name=flow_run_name)
    image_url = create_image_from_story(credentials=credentials, story=story)

    story_and_image = (
        f"The story about a {flow_run_name}: '{story}' "
        f"And its image: {image_url}"
    return story_and_image


Saving an OpenAI key

It's easy to set up an OpenAICredentials block!

  1. Head over to
  2. Login to your OpenAI account
  3. Click "+ Create new secret key"
  4. Copy the generated API key
  5. Create a short script, replacing the placeholders (or do so in the UI)
from prefect_openai import OpenAICredentials`

Congrats! You can now easily load the saved block, which holds your OpenAI API key:

from prefect_openai import OpenAICredentials

Visit Flow Run Name Art to see some example output!


For more tips on how to use tasks and flows in a Collection, check out Using Collections!


Install prefect-openai with pip:

pip install prefect-openai

Requires an installation of Python 3.7+.

We recommend using a Python virtual environment manager such as pipenv, conda or virtualenv.

These tasks are designed to work with Prefect 2.0. For more information about how to use Prefect, please refer to the Prefect documentation.


If you encounter any bugs while using prefect-openai, feel free to open an issue in the prefect-openai repository.

If you have any questions or issues while using prefect-openai, you can find help in either the Prefect Discourse forum or the Prefect Slack community.

Feel free to star or watch prefect-openai for updates too!


If you'd like to help contribute to fix an issue or add a feature to prefect-openai, please propose changes through a pull request from a fork of the repository.

Here are the steps:

  1. Fork the repository
  2. Clone the forked repository
  3. Install the repository and its dependencies:
    pip install -e ".[dev]"
  4. Make desired changes
  5. Add tests
  6. Insert an entry to
  7. Install pre-commit to perform quality checks prior to commit:
    pre-commit install
  8. git commit, git push, and create a pull request