Thanks for reaching out! In Prefect 3.x, a deployment captures “how to fetch and run your code” at the time you deploy. If code changes aren’t showing up until you delete and redeploy, it usually means your deployment is still pointing at an old source (old Git ref, old container image, or code that isn’t present on the worker machine). Common causes and fixes: - Using Git but pinned to an old commit - Cause: Your deployment’s pull steps are pinned to a specific commit SHA, so new code isn’t pulled. - Fix: Redeploy with the new commit SHA, or configure `from_source(...)`/pull steps to use a branch ref if you want “always latest” behavior. - Docs: Deployments, Deploy via Python, Deployment versioning - Using Docker/Kubernetes and the code is baked into an image - Cause: Your deployment references an existing image tag; unless you rebuild/push a new image and point the deployment to it, runs will keep using the old code. - Fix: Build/push a new image for each change (e.g., tag with a commit SHA) and redeploy so the deployment points to the new tag. Avoid reusing “latest” unless your infra always pulls fresh images. - Tip (Kubernetes): If you must use “latest”, ensure imagePullPolicy is Always and that you actually pushed the new image. - Local Process work pool with local files - Cause: The worker loads code from the machine it runs on. If the worker host doesn’t have the updated files (or the deployment uses pull steps that aren’t fetching the latest), runs will use old code. - Fix: Make sure the updated files exist on the worker host, or configure Git pull steps to fetch the latest on each run. - Older Prefect version - In older 3.x releases there were fixes around pull step updates; upgrade to the latest 3.x to rule this out. - Release notes: 3.2 notes Quick checks: - Inspect what your deployment actually uses (entrypoint, pull_steps, image):

prefect deployment inspect "your-flow-name/your-deployment-name"

- Re-run your deployment command instead of deleting;

prefect deploy

will create a new deployment version pointing at your latest config/code:

prefect deploy

To help pinpoint this for you, could you share: - How you’re deploying:

prefect deploy

CLI, or

flow.from_source(...).deploy(...)

in Python? - Your work pool type: Process, Docker, or Kubernetes? - Where your code lives for runs: local filesystem, Git branch/commit, or container image? - Your Prefect version (

prefect version

) and a snippet of

prefect deployment inspect

output (entrypoint, pull_steps, job_variables/image)? With those details I can give you exact steps to make changes reflect without having to delete the deployment each time.