Nuke: Nushell Kubernetes Integration

---

Why?

Interacting with kubernetes looks too much like this:

# does not actually work
kubectl get po 
| detect columns --guess 
| update AGE {|po|
    $po.AGE 
    | str replace 's' 'sec' 
    | str replace 'm' 'min' 
    | str replace 'h' 'hour' 
    | str replace 'd' 'day' 
    | into datetime 
}
| sort-by AGE

I wish i could just kubectl get po | sort-by age and leverage the power of nushell. Don't you?

---

So What?

Nuke re-implements some of kubectl commands.
It talks directly with the kube-apiserver to retrieve structured objects and typed data, so we can run things like:

nuke get po -o wide | where node in (
  (nuke top no | sort-by memory -r | first 3).name
) # gets the pods running on the most overloaded nodes

• Nuke does not aim to reimplement all of kubectl. It focuses on the commands that wuould benefit from nushell's structured data.
• Nuke tries to mimick kubectl syntax to recreate a familiar environment. No need to learn a new tool.
• Nuke uses your kubeconfig as configuration. No additional setup is required.
• Nuke tries to adhere to kubectl semantics, integrating it with richer data.

Implemented Commands

Nuke Command	kubectl Equivalent
nuke get	kubectl get
nuke rollout status	kubectl rollout status
nuke http-get	kubectl get --raw
nuke api-resources	kubectl api-resources
nuke api-versions	kubectl api-versions
nuke top	kubectl top
nuke config	kubectl config

---

How Nuke Works

1. Reads your kubeconfig
2. Authenticates against the API server
3. Performs HTTP requests directly
4. Applies resource-specific formatter
5. Returns structured Nushell data

If no formatter is implemented, a default formatter is used.

Authentication

Nuke authenticates with the Kubernetes API server using the credentials defined in your kubeconfig, following a precedence model similar to kubectl.

Installation

Currently you need to build from source:

# Clone repo
git clone git@github.com:lassoColombo/nuke.git
cd nuke

# Build
cargo build

# Add plugin
plugin add target/debug/nu_plugin_nuke
plugin use nuke

# Verify installation:
nuke get po

---

Formatters

Commands that retrieve objects support three formats:

Format	Description
compact	minimal view (Default for lists).
wide	extended attributes (Default for single objects).
full	the complete object from the API

Nuke is currently under active development, so not all resources have a dedicated formatter yet.
When a specific formatter isn’t available, Nuke automatically falls back to the default formatter.

Decorator flags

nuke get supports decorator flags that add extra columns to the formatter output. They work with any formatter and any format (compact/wide):

Flag	Adds column
--show-labels	labels
--show-annotations	annotations
--show-owner	owner (controller from metadata.ownerReferences)
--show-finalizers	finalizers
--show-managed-fields	managed-fields (list of managers)

nuke get po --show-labels --show-owner

Looking up resources

nuke get accepts three forms for the resource argument:

nuke get po                              # short name / plural / kind (discovery-resolved)
nuke get all                             # a category (e.g. "all", "api-extensions")
nuke get metrics.k8s.io/v1beta1/pods     # fully-qualified group/version/plural
nuke get v1/pods                         # core group: version/plural

Nuke Http-Get

nuke http-get performs an authenticated GET against the kube API server (equivalent to kubectl get --raw). When the response is JSON it's parsed into structured Nushell data; otherwise the raw body is returned as a string.

Flag	Description
--params, -P	Query parameters as a record. Values may be strings or lists (lists produce repeated keys).
--headers, -H	Request headers as a record.
--raw, -r	Skip JSON parsing and return the response body as a plain string.
--context / --cluster / --user	Override kubeconfig selection.

# get pods
nuke http-get /api/v1/namespaces/<namespace>/pods

# get pods by label
nuke http-get /api/v1/namespaces/<namespace>/pods -P {
    labelSelector: 'my-label in (my-value-1, my-value-2)'
}

# get aggregated api discovery
nuke http-get /apis -H {
   Accept: "application/json;v=v2;g=apidiscovery.k8s.io;as=APIGroupDiscoveryList"
}

# fetch /metrics as plain text
nuke http-get /metrics --raw

---

Nuke Rollout Status

Tracks the rollout of a Deployment, DaemonSet, or StatefulSet and returns a structured record (name, kind, namespace, created, done, message, ready, desired, strategy).

nuke rollout status deployment my-app
nuke rollout status deployment my-app -n production
nuke rollout status deployment my-app --timeout 0   # don't wait, return current status

The default behavior waits up to 300 seconds for the rollout to complete; pass --timeout 0 for a one-shot read.

---

Kubeconfig

Nuke provides utilities to manage your kubectl configuration, and to help you switch context swiftly.

Context Switching

Context switching takes inspiration from kubectx and kubens:

nuke config switch-namespace monitoring # explicit switch
nuke config switch-context # interactive switch - triggers input list

Configuration Utilities

Nuke provides structured access to your kubeconfig data:

nuke config # returns the kubeconfig
nuke config get-path # get the path to the current detected kubeconfig
nuke config get-contexts # get all the contexts
nuke config get-contexts --current # get the current context
nuke config get-current-namespace # get the current namespace
nuke config get-clusters --current # get the current cluster
nuke config get-users --context k8s-001 # get the user of context k8s-001
nuke config get-clusters --context k8s-qa # get the cluster of context k8s-qa

---

Roadmap

• Improve coverage of built-in resource formatters
• Implement nuke describe command
• Implement --revision flag for rollout command
• Additional authentication methods
  • Exec plugins
• Watch functionality