Merge branch 'main' into issue-1693-filter-span

Author: chethanuk

.github/ISSUE_TEMPLATE/new-release.md

@@ -70,7 +70,8 @@ This document defines the process for releasing llm-d-router.
      A release branch should already exist. In this case, check out the existing branch:
 
      ```shell
-     git checkout release-${BRANCH_VERSION} ${REMOTE}/release-${BRANCH_VERSION}
+     git fetch ${REMOTE}
+     git reset --hard ${REMOTE}/release-${BRANCH_VERSION}
      ```
 
 1. By default, `LATENCY_PREDICTOR_TAG` in the `Makefile` resolves from the router release tag (via `BUILD_REF`). If the latency predictor tag does **not** align with the router version, update the default value of `LATENCY_PREDICTOR_TAG` in the `Makefile` to match your exported `${LATENCY_PREDICTOR_TAG}`.

.github/workflows/ci-dev.yaml

@@ -14,6 +14,7 @@ permissions:
 
 jobs:
   set-params:
+    if: github.event.repository.fork != true
     runs-on: ubuntu-latest
     outputs:
       epp_name: ${{ steps.version.outputs.epp_name }}
@@ -33,6 +34,7 @@ jobs:
           echo "tag=${GITHUB_REF_NAME}" >> "$GITHUB_OUTPUT"
 
   build-and-push:
+    if: github.event.repository.fork != true
     needs: set-params
     uses: ./.github/workflows/ci-build-images.yaml
     with:

.github/workflows/ci-lint.yaml

@@ -78,7 +78,7 @@ jobs:
           echo "build=$HOME/.cache/llm-d-gobuildcache" >> "$GITHUB_OUTPUT"
 
       - name: Cache Go modules and build cache
-        uses: actions/cache@v5
+        uses: actions/cache@v6
         with:
           path: |
             ${{ steps.go-cache.outputs.mod }}

.github/workflows/ci-pr-checks.yaml

@@ -80,7 +80,7 @@ jobs:
           echo "build=$HOME/.cache/llm-d-gobuildcache" >> "$GITHUB_OUTPUT"
 
       - name: Cache Go modules and build cache
-        uses: actions/cache@v5
+        uses: actions/cache@v6
         with:
           path: |
             ${{ steps.go-cache.outputs.mod }}
@@ -93,7 +93,7 @@ jobs:
       # components as new and exits 0.
       - name: Restore main branch coverage baseline
         if: github.event_name == 'pull_request'
-        uses: actions/cache/restore@v5
+        uses: actions/cache/restore@v6
         with:
           path: coverage/baseline
           key: coverage-main
@@ -179,7 +179,7 @@ jobs:
 
       - name: Save coverage baseline (main branch only)
         if: github.ref == 'refs/heads/main' && github.event_name == 'push'
-        uses: actions/cache/save@v5
+        uses: actions/cache/save@v6
         with:
           path: coverage/baseline
           key: coverage-main

.github/workflows/ci-release.yaml

@@ -73,7 +73,7 @@ jobs:
 
       - name: Build Artifacts
         run: |
-          make artifacts
+          make artifacts BUNDLE_VERSION=${{ needs.set-params.outputs.tag }}
 
       - name: Upload Release Assets
         env:

.github/workflows/md-link-check.yml

@@ -18,14 +18,30 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
+  
+  check-for-markdown-files:
+    runs-on: ubuntu-latest
+    outputs:
+      src: ${{ steps.filter.outputs.src }}
+    steps:
+      - name: Checkout source
+        uses: actions/checkout@v7
+      - uses: dorny/paths-filter@v4
+        id: filter
+        with:
+          filters: |
+            src:
+              - '**/*.md'
+
   lychee:
     name: Check Markdown Links
+    needs: check-for-markdown-files
     # pull_request_target fires even when GITHUB_TOKEN creates the PR (e.g. the
     # release-notes bot). pull_request does not. We register pull_request_target
     # so bot-created PRs get a "skipped" conclusion that satisfies branch
     # protection, but we skip all real work: pull_request already covers human
     # PRs, and running twice would cause the concurrency group to cancel one run.
-    if: github.event_name != 'pull_request_target'
+    if: github.event_name != 'pull_request_target' && ${{ needs.check-changes.outputs.src == 'true' }}
     runs-on: ubuntu-latest
 
     steps:

.github/workflows/release-notes-assemble.yaml

@@ -7,6 +7,8 @@ on:
 
 jobs:
   assemble:
+    # RC tags are pre-release; only assemble notes for final releases.
+    if: ${{ !contains(github.ref_name, '-rc') }}
     runs-on: ubuntu-latest
     permissions:
       contents: read
@@ -45,6 +47,9 @@ jobs:
         run: |
           set -euo pipefail
 
+          # ── Phase 1: read fragments from the tagged commit ──────────
+          # The checkout step landed us on the tagged commit so that
+          # fragments merged into main *after* the tag are excluded.
           shopt -s nullglob
           fragments=( release-notes.d/unreleased/*.md )
           if [[ ${#fragments[@]} -eq 0 ]]; then
@@ -54,15 +59,9 @@ jobs:
 
           tmp=$(mktemp)
           for f in "${fragments[@]}"; do
-            # Extract frontmatter values: strip "key: " prefix, take first match.
-            # Plain field-split breaks on URLs because they contain colons.
             pr=$(sed -n 's/^pr: *//p' "$f" | head -1)
             url=$(sed -n 's/^url: *//p' "$f" | head -1)
             date=$(sed -n 's/^date: *//p' "$f" | head -1)
-            # Print everything after the second `---` line verbatim — once
-            # we're past the closing frontmatter marker, a body line that
-            # happens to be `---` (Markdown horizontal rule) must be kept,
-            # not counted as a third delimiter.
             note=$(awk 'p {print; next} /^---$/ {if (++c == 2) p = 1}' "$f" | sed '/^[[:space:]]*$/d' | tr '\n' ' ' | sed 's/[[:space:]]*$//')
             if [[ -z "$pr" || -z "$url" || -z "$date" || -z "$note" ]]; then
               echo "Malformed fragment: $f" >&2
@@ -81,17 +80,30 @@ jobs:
           done < "$tmp"
           printf '\n' >> "$new_section"
 
+          # ── Phase 2: branch from main so the PR contains only
+          #    release-notes changes, not unrelated diffs between
+          #    the tagged commit and main HEAD ─────────────────────
+          git fetch origin main
+          BRANCH="release-notes/assemble-${TAG}"
+          git config user.name "${APP_SLUG}[bot]"
+          git config user.email "${APP_USER_ID}+${APP_SLUG}[bot]@users.noreply.github.com"
+          git checkout -b "$BRANCH" origin/main
+
+          # ── Phase 3: apply release-notes changes on top of main ─
           touch RELEASE-NOTES.md
           cat "$new_section" RELEASE-NOTES.md > RELEASE-NOTES.md.new
           mv RELEASE-NOTES.md.new RELEASE-NOTES.md
+          git add RELEASE-NOTES.md
 
-          git rm -- release-notes.d/unreleased/*.md
+          # Remove only the fragments captured from the tagged commit,
+          # not the current main glob which may include post-tag fragments.
+          # The -f guard skips fragments already absent from main.
+          for f in "${fragments[@]}"; do
+            if [[ -f "$f" ]]; then
+              git rm -- "$f"
+            fi
+          done
 
-          BRANCH="release-notes/assemble-${TAG}"
-          git config user.name "${APP_SLUG}[bot]"
-          git config user.email "${APP_USER_ID}+${APP_SLUG}[bot]@users.noreply.github.com"
-          git checkout -b "$BRANCH"
-          git add RELEASE-NOTES.md
           git commit -s -m "docs: assemble release notes for ${TAG}"
           # Force-push so a re-run (re-tag, manual workflow re-run) overwrites
           # the prior branch rather than failing non-fast-forward.

.lychee.toml

@@ -1,7 +1,9 @@
 # Ignore transient failures on gnu.org (it sometimes refuses connections)
 exclude = [
   "^https://www.gnu.org/software/make/?$",
-  "^https://www.envoyproxy.io/"
+  "^https://www.envoyproxy.io/",
+  "^https://gateway-api-inference-extension.sigs.k8s.io/guides/serving-multiple-inference-pools-latest/",
+  "^https://github.com/kubernetes-sigs/gateway-api-inference-extension/blob/main/site-src/guides/metrics-and-observability.md"
 ]
 
 # Per-request timeout in seconds. Sized for slow upstream docs sites

AGENTS.md

@@ -30,6 +30,12 @@ llm-d Router. Go service that routes inference requests to model-serving pods vi
 - The PR addresses that issue and nothing else: no renames, reformatting, refactors, new abstractions, or pattern changes beyond what the issue requires.
 - Unrelated improvements belong in their own issue and PR, not folded into this PR. If you spot dead code or unrelated bugs in passing, mention them; don't fix them.
 - Self-check on the way out: if the change grew larger than expected or the fix feels hacky, rewrite the clean version before opening the PR.
+- Verify the code passs `make presubmit` locally before submitting a PR.
+- Always use the project's `.github/PULL_REQUEST_TEMPLATE.md`.
+  - Document user (not developer) facing changes in the ```release-note``` block. The  `release-notes.d/unreleased/*`
+    file is automatically generated from the block's content - do not create the file directly. 
+  - If you include a test plan section, mark passing tests with [x] so it is clear which ran and passed.
+    List only new tests - indicate functionality verified, not the test names.
 
 ## Code style
 

Makefile

@@ -362,8 +362,11 @@ helm-push-standalone: ## Package and push the llm-d-router-standalone Helm chart
 
 ##@ Release
 
+BUNDLE_VERSION ?= main-dev
+export BUNDLE_VERSION
+
 .PHONY: artifacts
-artifacts: yq check-kustomize ## Generate release artifacts (CRD manifests)
+artifacts: generate yq check-kustomize ## Generate release artifacts (CRD manifests).
 	if [ -d artifacts ]; then rm -rf artifacts; fi
 	mkdir -p artifacts
 	kubectl kustomize config/crd > artifacts/manifests_all.yaml