[FEATURE] Kubernetes deployment + CRD

[yairyairyair]

Describe the feature you'd like Would like flowise to have better kubernetes deployment, currently its just a docker container. Would be easier if its still a docker container however when deploying a flow it will create a CRD that deploys its own container, maybe something like this:

apiVersion: 'flowiseai.com/v1alpha1' kind: 'LLMApp' spec: { flow: SomeFlowObjectOrSpecHere } status:{ address:{ url: 'https:// the deployed url' } }

That way even if the whole flowise container is crashing the app will still work since different container, giving more "Production ready" features, making flowise move closer towards prod-ready external apps instead of just internal apps.

Ideas:

• this feature can use knative serving for deploying a single app like most AI deployment frameworks do, for example: kserve, seldon, replicate, tensorflow etc... optional.

[HenryHengZJ]

if you have it setup successfully with kubernetes, would be awesome if you can share the steps!

[yairyairyair]

dont have it currently, its a feature that i want. what i do now is just knative service with flowiseai/flowise docker container i am talking about the main container "spawning" new containers that serve only a single flow each. can be done in kubernetes easily via a controller or operator with a CRD

[Padarn]

There is a helm chart here: https://github.com/cowboysysop/charts/tree/master/charts/flowise (I haven't used it yet). Would a PR adding a helm chart and instructions be something you'd want to have?