What version are you on? The minimal reproduction is to just create a flow and deploy it?
What version are you on? The minimal reproduction is to just create a flow and deploy it?
R1.555.1
PAny specific actions? Imports?
PAnd what is your infra? k8s? Docker? Native?
DDid you ever do something to your global settings that might have deleted it ?
DI'm going through the code and assuming the migration ran, I don't see how this was possible
Rright now a local docker running
Rits a rather complicated flow, all python is checked and valid with tests and i can confirm they work.
RI can view all scripts etc on the frontend
TI've been setting instance-level settings using a
instance_settings.yaml that doesn't have any ducklake settingsDoh
Dthere we go 
thanks
thanksT@Diego just to be clear I do feel like the way this has broken is not an optimal customer/developer experience. I don't think upgrading versions should break in this way.
Dyes it's a bug
PCould it just be locking things? Could happen if you have poor connectivity.
But if it is complicated flow, I think it is the best to gradually reduce until it works
But if it is complicated flow, I think it is the best to gradually reduce until it works
T@Diego gotcha! happy to be a point of contact on the fix if that's helpful 
really appreciate how you + ruben have hopped in here so quickly
really appreciate how you + ruben have hopped in here so quicklyRill refactor it and start with each step until it breaks somewhere
PYeah. But you should really check dependency jobs. You can filter by that job kind in run page
PI just think you have something heavy or it is somehow stuck on a a dependency
Rits a local instance, no jobs running except the deployment
Pby heavy I mean you have hundreds of dependencies (including transitive) and your instance didn't finish fetching all necessarily metadata
Rthey all finished, is there a chance one may have not?
PIf it is a single flow and there is no relative imports, it would be a single dependency job. If that dependency job has finished then the flow should be viewable as deployed
MThis is how simple it could be to interact with wmill storage / ducklake
MBun
Bun 1.3 introduces zero-config frontend development, unified SQL API, built-in Redis client, security enhancements, package catalogs, async stack traces, VS Code test integration, and Node.js compatibility improvements.
MVery impressive from bun, when I was reading it I kept thinking that these can be made to interoperate with windmill by default
MResources β redis KV (types can be specified inline, defaulting to βstateβ _resource_type: βs3β)
Variables β secrets
Storage β s3
* Ducklake β sql
Variables β secrets
Storage β s3
* Ducklake β sql
RAgreed @metapontum , we will probably want a simplified syntax, maybe a one function call in our sdk to setup the s3 client
ZHave you thought about Bun to build/serve the raw apps, @rubenf? (Iβve actually been using raw apps to iframe to a subdomain with a cookie.) Was reading through some of the raw apps branch and was just curious?
R@Zach raw apps are purely static frontend rendering + windmill jobs for background runnables
RYouTubeWindmill
Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.
Kexcited for the ui polish, wanted to suggest some attention on this tiny papercut I run into a lot. Resizing these panes can be hard especially on high resolution/remote desktop.
Rat very least we could make it bigger on hover I suppose
Ki dont think it needs to be visibly bigger, but a few more pixels of "grip" would help a ton
Kthe area where my pointer turns into a resizer
Ki think this is extra bad with remote desktop that might be why few people notice
II have a git repo that i create and work with wmill scripts. I also have the windmill instance setup in a docker compose in said repo. I sync this repo between home and work machines. So home instance of windmill is not the same as the work instance. They should be configured the same and use the same workspace encryption key. I do a git pull, then wmill sync push. However files that were deleted keep showing back up, and im still having the issue where .sh scripts are not adding the input parameters to the yaml file. I have also tried git pull, wmill script generate-metadata, wmilll sync push with same results. Am I not doing something right?
REvery steps you've described should be easy to debug if something is going not as expected @IceCactus as it's very explicit. Also try with a more minimal example and see.
ISo what im doing should work? I dont want to waste time on doing something it wasn't designed to do.
Rthe sync will make target == source yes
Rincluding deleting files that are on target but not on source
IShould I git ignore the wmill lock file?
Rno
MHave never encountered that before but might want to try doing it on chrome
MCould be a issue with comet specifically
Iif I specify a resource as an input to a flow, how can I extract the properties of that resource to pass to a script? im currently using this: resource(flow_input.ftp_resource.replace('$res:', '')).host which works, but one would think you should be able to do flow_input.ftp_resource.host
MIs there a way to have on-site/off-site redundancy w/ community edition with shared db.
MIf not possible already using pgdog, I believe they are working on sharding
MWould not be surprised if such a feature is implemented soon
Rfor that you can just use standard postgresql WAL read replicas