@Marvin Does Prefect 3 not include newly-created deployments in prefect.yaml by default? I've been using code-based deployments for a while, but just tried going back to creating one via "prefect deploy". It created it successfully in my prefect instance, but doesn't prompt me to see if I want to save the deployment code for later, and doesn't edit prefect.yaml

✅ thought for 116.6 seconds

@Marvin if I have an existing prefect.yaml (created by "prefect init", with no deployments specified), shouldn't it pick that up and prompt me to ask if I want to save? I'm running from an interactive terminal.

✅ thought for 8.3 seconds

@Marvin when did this change? it used to prompt me any time there was a new deployment added, and then run without prompting on future runs. Is there another CLI utility that we can use to quickly add things with the proper schema? @Nate I think so. The last time I used the CLI for creating deployments was in the 2.x days, where I remember it adding them for me.

✅ thought for 213.4 seconds

2.x shouldn't have been that different, unless you had some custom harness for the CLI (unless there's a branch of the interactive CLI that i'm forgetting, which is possible) editing a shared/checked-in

prefect.yaml

and then

prefect --no-prompt deploy --all

in CI is my general recommendation the thing that often trips people up is that

prefect.yaml

isn't the exhaustive state on the server, its just a declaration of what you want to exist when you run

prefect deploy

, but someone else might have their own

specific-prefect.yaml

file. some orgs like different norms, ie certain teams have their own

prefect.yaml

vs strictly keeping all things in one

prefect.yaml

per server/workspace

@Nate That should work well for CI/CD. I'm just trying to find a way to guide some new team members through that "cli wizard" style process for picking the entrypoint, generating schedule, etc. We' re looking to consolidate a bunch of single-workflow repos into one big monorepo, and it looked like going back to the YAML setup and CLI may be a better option than our custom code deployment. It'd be nice to be able to trigger that "wizard" and save the output as well-formatted YAML whenever they add a new pipeline to the repo

gotcha! yea

It'd be nice to be able to trigger that "wizard" and save the output as well-formatted YAML whenever they add a new pipeline to the repo

this should happen today? ie if your

prefect.yaml

exists, and you specify a new deployment by answering the wizard's questions, it should ask you to persist it as a new entry in the yaml.. if it doesn't that'd be unexpected to me and i can look into it!

@Nate ahh ok that's where I'm confused. because it's generating a deployment in my local test server (local prefect OSS instance running 3.7.1), but not prompting me to save the output into Prefect.yaml, and not modifying the deployments section. That's where I was confused, because it definitely always saved it in there before. it creates the deployment, but doesn't touch prefect.yaml: Successfully built 23a6a1b41984

prefect._experimental._launchers

has moved to

prefect._internal.launchers

. The old import path will be removed in a future release. ? Would you like to configure schedules for this deployment? [y/n] (y): y ? What type of schedule would you like to use? [Use arrows to move; enter to select] ┏━━━━┳━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓ ┃ ┃ Schedule Type ┃ Description ┃ ┡━━━━╇━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩ │ │ Interval │ Allows you to set flow runs to be executed at fixed time intervals. │ │ > │ Cron │ Allows you to define recurring flow runs based on a specified pattern using cron syntax. │ │ │ RRule │ Allows you to define recurring flow runs using RFC 2445 recurrence rules. │ └────┴───────────────┴──────────────────────────────────────────────────────────────────────────────────────────┘ ? Cron string (0 0 * * *): 0 8 * * * ? Timezone (UTC): America/Regina ? Would you like to activate this schedule? [y/n] (y): y ? Would you like to add another schedule? [y/n] (n): n ╭────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮ │ Deployment 'accounting_files_2/default' successfully created with id '5bbccb5e-d2e4-42fb-8aa7-99ae4526db6d'.

thanks! helpful context, will take a look

@Nate don't know if I'm reading this correctly, but I think a regression may have snuck in when there was a large-scale migration done on the CLI. It seems to only prompt if you want to save the configuration now if the prefect_file (default prefect.yaml) doesn't exist: (prefect/cli/deploy/_core.py, line 423): if is_interactive() and not prefect_file.exists(): if confirm( ( "Would you like to save configuration for this deployment for faster" " deployments in the future?" ), console=console, ): deploy_config_before_templating.update({"schedules": _schedules}) _save_deployment_to_prefect_file( deploy_config_before_templating, build_steps=build_steps or None, push_steps=push_steps or None, pull_steps=pull_steps or None, triggers=trigger_specs or None, sla=sla_specs or None, prefect_file=prefect_file, ) console.print( ( f"\n[green]Deployment configuration saved to {prefect_file}![/]" " You can now deploy using this deployment configuration" " with:\n\n\t[blue]$ prefect deploy -n" f" {deploy_config['name']}[/]\n\nYou can also make changes to" " this deployment configuration by making changes to the" " YAML file." ), ) The method it calls (_save_deployment_to_prefect_file) already seems to check if the file exists, and gracefully handles creating or editing it, so I'm not sure if that exists() call is necessary? "def _save_deployment_to_prefect_file( deployment: dict[str, Any], build_steps: list[dict[str, Any]] | None = None, push_steps: list[dict[str, Any]] | None = None, pull_steps: list[dict[str, Any]] | None = None, triggers: list[dict[str, Any]] | None = None, sla: list[dict[str, Any]] | None = None, prefect_file: Path = Path("prefect.yaml"), ): """ Save a deployment configuration to the

prefect.yaml

file in the current directory. Will create a prefect.yaml file if one does not already exist. Args: - deployment: a dictionary containing a deployment configuration """ deployment = _format_deployment_for_saving_to_prefect_file(deployment) current_directory_name = os.path.basename(os.getcwd()) if not prefect_file.exists(): if triggers: deployment["triggers"] = triggers if sla: deployment["sla"] = sla create_default_prefect_yaml( ".", current_directory_name, contents={ "deployments": [deployment], "build": build_steps, "push": push_steps, "pull": pull_steps, }, ) create_default_ignore_file(".") else: # use ruamel.yaml to preserve comments ryaml = YAML() ... proceeds to read the existing config, merge with new config, then write the new file "

thanks for the thoughts! i think this an edge case more likely when using

prefect init

that arose from an overly broad heuristic to stop extra prompts. will get this reviewed and update here!