Hi team, I'm new to using prefect so bare with me. I'm getting the attached error when trying to run the remote deploy (managed via github actions) from prefect cloud however the same code works when I use a local deploy. openai looks to be installed to me when looking at the generated logs. Furthermore I'm pretty sure this worked when I tried it yesterday.

hi @Emmanuel Quarless - that second picture where you show deps installed, where is that? relatedly, where / how are you specifying what dependencies should exist in your remote runtime?

Hi @Nate - that's the output in githb actions after running the "Prefect Deploy" step defined in the third picture (github actions workflow)

right ok that’s what i suspected! so you essentially need to include some information in your deployment on runtime deps, and there’s a couple ways to do that depending on your runtime. what kind of work pool are you using?

Here's some relevant information on deploying Prefect flows with containerized runtimes and dependency management: 1. Dependencies at Deploy Time: It's recommended to install dependencies at deploy time to build a parameter schema, which enhances the UI experience with forms. However, if you're using a containerized runtime (like Docker, Kubernetes, or ECS), it's typical to build an image with all necessary dependencies and set that image as a job variable on your work pool. For more details, you can refer to Prefect YAML. 2. Building Docker Images: When using Docker, you can build the image using the

docker build

command and push it to a container registry. Prefect provides guidelines on how to manage this process, including injecting environment variables and running the container. More information is available in the Docker Deployment Guide. 3. Managing Work Pools: Work pools allow you to manage infrastructure for flow runs. You can version control your job templates and update them through CI/CD pipelines. Prefect supports various work pool types, and you can find more about them in the Work Pools Documentation. 4. Prefect Blocks: Blocks in Prefect store typed configurations and are useful for managing parameters and infrastructure settings across deployments. For more on blocks and their uses, see Prefect Blocks. If you have further questions or need more specific details, feel free to ask.

oops i triggered marvin again bc i edited my message 🙂

thanks for getting back to be btw! so in my that last pip install in my third image wouldn't do the job? and the workpool is

Prefect:managed

I kinda wanted to avoid any Dockerisation https://docs.prefect.io/v3/deploy/infrastructure-concepts/deploy-ci-cd#build-deployments-via-ci-cd

oh yeah if you’re using the managed pool things are easier check out this example • define your flow • define your deployment in a `prefect.yaml` • use the cli to deploy (you can do this in CI as you are now)

» prefect --no-prompt deploy -n uses-managed-work-pool

• trigger and observe!

» prefect deployment run 'load-titanic-dataset/uses-managed-work-pool' --watch
Creating flow run for deployment 'load-titanic-dataset/uses-managed-work-pool'...
Created flow run 'resilient-fox'.
└── UUID: 54686b95-0be3-4f0e-80fc-9b3a184d6e31
└── Parameters: {}
└── Job Variables: {}
└── Scheduled start time: 2025-02-07 19:28:12 CST (now)
└── URL: <https://app.prefect.cloud/account/xxx/workspace/cafa2ffa-f6cc-4ed6-ab76-eaa4ba1ad40e/runs/flow-run/xxx>
Watching flow run 'resilient-fox'...
19:28:13.202 | INFO    | prefect - Flow run is in state 'Scheduled'
19:28:18.304 | INFO    | prefect - Flow run is in state 'Scheduled'
19:28:23.419 | INFO    | prefect - Flow run is in state 'Scheduled'
19:28:28.523 | INFO    | prefect - Flow run is in state 'Pending'
19:28:33.633 | INFO    | prefect - Flow run is in state 'Pending'
19:28:38.734 | INFO    | prefect - Flow run is in state 'Pending'
19:28:43.877 | INFO    | prefect - Flow run is in state 'Pending'
19:28:49.000 | INFO    | prefect - Flow run is in state 'Pending'
19:28:54.098 | INFO    | prefect - Flow run is in state 'Pending'
19:28:59.185 | INFO    | prefect - Flow run is in state 'Completed'
Flow run finished successfully in 'Completed'.

ahh I follow you now thanks! Ye defining dependencies in the prefect.yaml file seems like it'll be a bit cumbersome so I'll probably take your advice and just build an image haha. (I'd also like to confirm it works once i put openai under the work pool's pip dependencies 🔥)