feat(helm): migrate chart publishing to GitHub Pages (#1257)

## Summary

- Replace the tag-triggered OCI/GitHub Container Registry Helm publish
workflow with Helm Chart Releaser on pushes to `main`.
- Publish charts from `deployments/helm` through GitHub Releases and
maintain the public Helm repository index on `gh-pages`.
- Document the public repository at `https://leptonai.github.io/gpud`
and migration from `oci://ghcr.io/leptonai/gpud` to `gpud/gpud`.
- Bump the Helm chart `version` and `appVersion` to `0.12.5`.
- Make `install.sh` accept both `0.12.5` and `v0.12.5`; it normalizes
archive downloads to the v-prefixed release-artifact form.

## Versioning

- Helm chart and container image tags use `0.12.5`.
- Source release artifacts use the v-prefixed form, `v0.12.5`.
- The Dockerfile and external-registry image build scripts remain
unchanged. The image is still independently selected with
`image.repository` and `image.tag`.

## Public chart usage

After the chart release workflow has run:

```bash
helm repo add gpud https://leptonai.github.io/gpud
helm repo update

helm upgrade --install my-gpud gpud/gpud \
  --version 0.12.5 \
  --create-namespace \
  --namespace gpud
```

## Validation

- `ct lint --debug --all --config ./.github/configs/ct-lint.yaml
--lint-conf ./.github/configs/lintconf.yaml --target-branch main`
- `helm lint deployments/helm/gpud`
- `bash -n install.sh`
- Installer normalization check for both `0.12.5` and `v0.12.5`
- `git diff --check`

Author: gyuho

.github/workflows/helm.yaml

@@ -1,71 +1,29 @@
-name: 'helm'
+name: Helm chart release
 
-on:
+"on":
   push:
-    tags:
-      - 'v*'
+    branches:
+      - main
 
 permissions:
-  id-token: write
-  packages: write
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true
-
+  contents: write
 
 jobs:
-  run:
+  release:
     runs-on: ubuntu-latest
     steps:
-      - name: "checkout"
-        uses: actions/checkout@v4
+      - uses: actions/checkout@v4
         with:
           fetch-depth: 0
 
-      - name: setup
-        run: |
-          wget https://github.com/helm/chart-testing/releases/download/v3.14.0/chart-testing_3.14.0_linux_amd64.tar.gz -O ct.tar.gz
-          tar -xzf ct.tar.gz
-          sudo mv ct /usr/local/bin/ct
-
-          wget https://get.helm.sh/helm-v3.19.2-linux-amd64.tar.gz -O helm.tar.gz
-          tar -xzf helm.tar.gz
-          sudo mv linux-amd64/helm /usr/local/bin/helm
-
-      - name: Install lint dependencies
-        run: |
-          python3 -m pip install --user --upgrade yamale yamllint
-          echo "$HOME/.local/bin" >> "$GITHUB_PATH"
-
-
-      - name: Run chart-testing
-        run: |
-          TARGET_BRANCH="${{ github.event.repository.default_branch }}"
-          if [ -z "$TARGET_BRANCH" ]; then
-            TARGET_BRANCH="main"
-          fi
-          if [ "${{ github.event_name }}" = "pull_request" ]; then
-            TARGET_BRANCH="${{ github.event.pull_request.base.ref }}"
-          fi
-          ct lint --debug --all --config ./.github/configs/ct-lint.yaml \
-            --lint-conf ./.github/configs/lintconf.yaml \
-            --target-branch "$TARGET_BRANCH"
-
-      - name: Run kubeconform
+      - name: Configure Git
         run: |
-          helm plugin install https://github.com/melmorabity/helm-kubeconform
-          # Validate the Helm chart manifests with kubeconform.
-          # Pass the chart path explicitly; the plugin errors if given the literal string "chart".
-          helm kubeconform deployments/helm/gpud --verbose --summary --strict --exit-on-error
+          git config user.name "$GITHUB_ACTOR"
+          git config user.email "$GITHUB_ACTOR@users.noreply.github.com"
 
-      - name: release
-        if: startsWith(github.ref, 'refs/tags/')
-        run: |
-          CHART_NAME="$(grep -e "^name:" deployments/helm/gpud/Chart.yaml | awk '{print $2}')"
-          CHART_VERSION="$(grep -e "^version:" deployments/helm/gpud/Chart.yaml | awk '{print $2}')"
-          CHART_REGISTRY="oci://ghcr.io/chart"
-          echo '${{ secrets.GITHUB_TOKEN }}' | helm registry login -u '${{ github.actor }}' --password-stdin 'ghcr.io'
-          helm package deployments/helm/gpud
-          echo "Pushing helm chart ${CHART_NAME}-${CHART_VERSION}.tgz to ${CHART_REGISTRY}"
-          helm push "${CHART_NAME}-${CHART_VERSION}.tgz" "${CHART_REGISTRY}"
+      - name: Release chart
+        uses: helm/chart-releaser-action@v1.6.0
+        with:
+          charts_dir: deployments/helm
+        env:
+          CR_TOKEN: ${{ secrets.GITHUB_TOKEN }}

deployments/helm/gpud/Chart.yaml

@@ -2,6 +2,6 @@ apiVersion: v2
 name: gpud
 description: GPUd Helm chart for Kubernetes
 type: application
-version: 0.12.2
-appVersion: "0.12.2"
+version: 0.12.5
+appVersion: "0.12.5"
 icon: https://assets.nvidiagrid.net/ngc/logos/Infrastructure.png

deployments/helm/gpud/README.md

@@ -11,10 +11,33 @@ GPUd is a lightweight, high-performance daemon that monitors GPU resources. This
 
 ## Installing the Chart
 
+Add the GPUd chart repository:
+
+```bash
+helm repo add gpud https://leptonai.github.io/gpud
+helm repo update
+```
+
+### Migrating from the OCI chart registry
+
+GitHub Container Registry no longer hosts GPUd charts. Replace an OCI install
+such as `oci://ghcr.io/leptonai/gpud` with the `gpud/gpud` chart after adding
+the repository above:
+
+```bash
+helm upgrade --install my-gpud gpud/gpud \
+  --version 0.12.5 \
+  --create-namespace \
+  --namespace gpud
+```
+
+The container image remains independently configurable through
+`image.repository` and `image.tag`.
+
 To install the chart with the release name `my-gpud`:
 
 ```bash
-helm install my-gpud <YOUR_REPO_NAME>/gpud \
+helm install my-gpud gpud/gpud \
   --create-namespace \
   --namespace gpud
 ```
@@ -24,7 +47,7 @@ helm install my-gpud <YOUR_REPO_NAME>/gpud \
 To install with a specific image tag and disable telemetry:
 
 ```bash
-helm install my-gpud <YOUR_REPO_NAME>/gpud \
+helm install my-gpud gpud/gpud \
   --create-namespace \
   --namespace gpud \
   --set image.tag="<MY_IMAGE_TAG>" \
@@ -34,7 +57,7 @@ helm install my-gpud <YOUR_REPO_NAME>/gpud \
 You can also provide a custom `values.yaml` file:
 
 ```bash
-helm install my-gpud <YOUR_REPO_NAME>/gpud \
+helm install my-gpud gpud/gpud \
   --namespace gpud -f my-values.yaml
 ```
 
@@ -168,7 +191,7 @@ Then install or upgrade with a values file like this (a full BYOK overlay):
 ```yaml
 image:
   repository: nvcr.io/nvstaging/dgx-cloud-lepton/gpud
-  tag: 0.12.2   # no "v" prefix from 0.12.2 onward
+  tag: 0.12.5   # no "v" prefix from 0.12.2 onward
 
 imagePullSecrets:
   - name: nvcr-staging-pull

install.sh

@@ -121,6 +121,8 @@ main() {
     # e.g., https://pkg.gpud.dev/unstable_latest.txt
     APP_VERSION=$(curl -fsSL https://pkg.gpud.dev/"$TRACK"_latest.txt)
   fi
+  # Git release artifacts are v-prefixed; accept either input spelling.
+  APP_VERSION="v${APP_VERSION#v}"
 
   if ! type lsb_release >/dev/null 2>&1; then
     # Try /etc/os-release first