ci: scaffold unified nightly and stable release workflow

[svvarom]

This replaces the separate nightly, RC, stable, and release-bundle workflows with one workflow for nightly and stable releases. Each run resolves an exact source commit and selected artifact set, optionally synchronizes NGC metadata, starts only the selected staging work, waits for the artifacts to reach their final locations, and reports the result. Stable releases also dispatch artifact registration, and full all releases create a GitHub Release. The legacy workflows, release-bundle metadata helper, and its tests are removed.

Linear: AIREINF-254 — Adapt Release Workflow by Newer Design

Stable and nightly behavior

Step	Stable release	Nightly pre-release	State
Plan and validate	Manual only. Requires an exact 40-character source SHA and MAJOR.MINOR.PATCH version.	Runs on the weekday schedule or manually. Uses an optional pinned SHA and generates nightly-YYYYMMDDHHMMSS.	Implemented
Select artifacts	Operator selects all, wheels, containers, helm, or a custom set of IDs.	Manual runs use the same selection model; scheduled runs select all.	Implemented
Synchronize NGC metadata	Optional manual input invokes the reusable metadata workflow.	Available for manual runs; scheduled runs skip it.	Implemented
Stage wheels	Dispatches the selected wheels with release cadence and version.	Dispatches the selected wheels with nightly cadence and timestamp.	Implemented
Stage containers	Dispatches the selected container IDs, source SHA, and release version.	Dispatches the selected IDs and source SHA using the generated nightly label.	Implemented
Stage Helm chart	Packages k8s/helm using the release version and publishes it to NGC.	Packages the same chart with a timestamped nightly version and publishes it to GHCR.	Implemented
Register release artifacts	Dispatches the selected wheel, container, and Helm IDs immediately after planning.	Skipped completely.	Implemented
Poll final locations	Waits for all selected artifacts to reach their public release locations.	Waits for all selected artifacts to reach their nightly locations.	Implemented
Create GitHub Release	Runs after polling succeeds for a full all release.	Skipped.	Implemented
Create deployment	Dispatches deployment creation after polling succeeds when Helm is selected.	Same behavior.	Implemented
Signal completed deployment	Separate downstream completion signal.	Same behavior.	Placeholder
Notify	Sends a start notification, a delayed-readiness alert after two hours, and a final success or failure notification.	Same behavior.	Implemented

Dry runs resolve and validate the release plan and package Helm when selected, but skip publishing, downstream release dispatches, final-location polling, GitHub Release creation, and deployment creation.

Artifact catalog

Type	IDs
Wheels	nemo-platform, nemo-platform-plugin
Containers	nmp-api, nmp-core, nmp-cpu-tasks, nmp-automodel-tasks, nmp-automodel-training, nmp-unsloth-training, auditor-tasks, safe-synthesizer-tasks
Helm	nemo-platform from k8s/helm

Dispatch contracts

Handoff	Runs for	Client payload
stage-wheels	Stable and nightly when wheels are selected	ref, cadence, release_label, nightly_timestamp, wheels
release	Stable and nightly when containers are selected	ref, version, containers
register-release-artifacts	Stable only	version, source_sha, source_run_url, wheel_ids, container_ids, helm_id
create-release-deployment	Stable and nightly after polling when Helm is selected	ref, cadence, release_label, helm_version

Helm staging runs directly in this workflow and does not use a repository dispatch. For custom stable releases, unselected ID arrays are empty and helm_id is null.

Before merge

Configure the SLACK_ALERTS_WEBHOOK and SLACK_RELEASE_WEBHOOK Actions secrets used by the notification jobs.

Summary by CodeRabbit

• New Features
  • Added a unified release.yaml workflow for nightly and stable releases, including artifact selection, validation, optional NGC metadata sync, Helm packaging/publishing, polling/verification, and Slack notifications (with dry-run).
  • Enforced an explicit required secret for NGC metadata synchronization.
• Removed
  • Removed nightly/RC/stable and release-bundle automation workflows, plus release bundle metadata generation and its unit tests.
  • Removed release asset manifest entries and NGC container description content for nmp-core.
• Documentation
  • Updated RELEASING.md to reflect the new release flow.
• Chores
  • Updated .gitignore for local Nektos Act files; improved kind pre-pulling and adjusted docker build targets.

[github-actions]

Suite	Lines Covered	Line Rate	Branch Rate
Unit Tests	23845/31009	76.9%	61.7%
Integration Tests	13816/29658	46.6%	19.6%

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 9d9d1e20-94aa-4b2f-89ce-5411cea028b0

📥 Commits

Reviewing files that changed from the base of the PR and between 6e3d135 and d611a5d.

📒 Files selected for processing (1)

• docker/base/Dockerfile.nmp-workspace

💤 Files with no reviewable changes (1)

• docker/base/Dockerfile.nmp-workspace

---

📝 Walkthrough

Walkthrough

The release process is consolidated into release.yaml, covering artifact selection, validation, staging, publication, polling, GitHub Releases, and Slack notifications. Legacy release automation is removed, while release documentation and local workflow support are updated.

Changes

Release pipeline

Layer / File(s)	Summary
Release planning and validation .github/workflows/release.yaml	Adds release inputs, artifact catalogs, selection checks, artifact validation, and wheel version resolution.
Artifact staging and publication .github/workflows/release.yaml, .github/workflows/ngc-metadata.yaml	Adds notifications, required NGC workflow secrets, optional metadata synchronization, artifact dispatches, and Helm publication.
Publication verification and completion .github/workflows/release.yaml	Adds registry polling, delay alerts, stable GitHub Release creation, deployment signaling, and final Slack notifications.
Legacy release automation removal .github/workflows/release-stable.yaml	Removes the previous stable release workflow.
Release procedure documentation RELEASING.md	Documents the unified workflow, inputs, catalog, destinations, notifications, secrets, and verification steps.

Build and local workflow support

Layer / File(s)	Summary
CPU build target selection docker-bake.hcl	Removes nmp-core-docker from the CPU build group.
Kind image preloading .github/actions/setup-kind-cluster/action.yaml	Adds the NMP API image to the images pre-pulled into Kind.
Nektos act file ignores .gitignore	Ignores .act-variables and .act-secrets.

Suggested reviewers: a2bondar, crookedstorm, maxdubrinsky, tylersbray

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the main change: replacing fragmented release paths with a unified nightly and stable release workflow.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch AIREINF-254-unified-release-workflow/smuley

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In @.github/workflows/release.yaml:
- Around line 702-705: Update the “Report packaged Helm chart” step to pass the
package output through the step’s env mapping, then reference that environment
variable in the shell command instead of directly expanding
steps.package.outputs.chart_package in run.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: f033b4c6-ba65-40c8-80f1-29d0efe65bf9

📥 Commits

Reviewing files that changed from the base of the PR and between 275c8cd and 399e17d.

📒 Files selected for processing (9)

• .github/scripts/write_release_bundle_metadata.py
• .github/workflows/ngc-metadata.yaml
• .github/workflows/release-bundle.yaml
• .github/workflows/release-nightly.yaml
• .github/workflows/release-rc.yaml
• .github/workflows/release-stable.yaml
• .github/workflows/release.yaml
• .gitignore
• tests/unit/release/test_write_release_bundle_metadata.py

💤 Files with no reviewable changes (6)

• .github/workflows/release-rc.yaml
• .github/workflows/release-nightly.yaml
• tests/unit/release/test_write_release_bundle_metadata.py
• .github/workflows/release-bundle.yaml
• .github/scripts/write_release_bundle_metadata.py
• .github/workflows/release-stable.yaml

[svvarom]

@crookedstorm can we also delete https://github.com/NVIDIA-NeMo/nemo-platform/blob/main/release/assets.yaml and delete/update https://github.com/NVIDIA-NeMo/nemo-platform/blob/main/RELEASING.md?

[coderabbitai]

🧹 Nitpick comments (1)

RELEASING.md (1)

141-153: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Missing "Next Steps" section.

Doc ends without cross-links to related docs (e.g. release.yaml, CONTRIBUTING). As per coding guidelines, "Include 'Next Steps' section at the end with cross-links to related documentation content."

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@RELEASING.md` around lines 141 - 153, Add a “Next Steps” section at the end
of RELEASING.md after the completed-release verification content, including
links to the related release workflow documentation (release.yaml) and
CONTRIBUTING guidance.

Source: Coding guidelines

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@RELEASING.md`:
- Around line 141-153: Add a “Next Steps” section at the end of RELEASING.md
after the completed-release verification content, including links to the related
release workflow documentation (release.yaml) and CONTRIBUTING guidance.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 38d174d9-1689-482f-8192-ec2dafd9f684

📥 Commits

Reviewing files that changed from the base of the PR and between 399e17d and 5c186f8.

📒 Files selected for processing (3)

• .github/workflows/release.yaml
• RELEASING.md
• release/assets.yaml

💤 Files with no reviewable changes (1)

• release/assets.yaml

🚧 Files skipped from review as they are similar to previous changes (1)

• .github/workflows/release.yaml