Improve generated script

It no longer tries to find the "release notes" since they are quite free form. Instead, an LLM can do that but after the script pulls the content.

Author: NickLarsenNZ

scripts/gather-release-notes.sh

@@ -85,15 +85,28 @@ gh api -X GET search/issues --paginate -f q="$QUERY" --jq '.items[]' \
     | jq -s 'unique_by(.html_url)' > "$ITEMS_FILE"
 
 if [ "$OUTPUT" = "json" ]; then
-    cat "$ITEMS_FILE"
+    # Enrich each item with its comments and a trimmed set of fields, so the
+    # release-note text (which authors write free-form, in the body OR a
+    # comment) is available in one place for a human/agent to read and
+    # interpret. Deliberately NO regex extraction here - see the note below.
+    jq -c '.[]' "$ITEMS_FILE" | while IFS= read -r item; do
+        full=$(echo "$item" | jq -r '.repository_url | sub(".*/repos/";"")')
+        number=$(echo "$item" | jq -r '.number')
+        comments=$(gh api "repos/$full/issues/$number/comments" --jq '[.[] | {author: .user.login, body}]' 2>/dev/null || echo '[]')
+        echo "$item" | jq -c --argjson comments "$comments" '{
+            repo: (.repository_url | sub(".*/repos/";"")),
+            number, title, html_url, state,
+            is_pr: (.pull_request != null),
+            merged: (.pull_request.merged_at != null),
+            labels: [.labels[].name],
+            action_required: (any(.labels[].name; . == "release-note/action-required")),
+            body: (.body // ""),
+            comments: $comments
+        }'
+    done | jq -s '.'
     exit 0
 fi
 
-# ------------------------------
-# Helper: is this item an action-required one?
-# ------------------------------
-is_action_required() { echo "$1" | jq -e 'any(.labels[].name; . == "release-note/action-required")' >/dev/null; }
-
 # ------------------------------
 # Counts (ground truth = number of items actually returned)
 # ------------------------------
@@ -139,48 +152,34 @@ echo "$OPEN_PRS"
 echo
 
 # ------------------------------
-# Items grouped by repository, with release-note snippet
+# Item inventory grouped by repository (metadata only).
+#
+# NOTE: we deliberately do NOT try to extract the release-note text here.
+# Authors write it free-form - under headings like `### Release notes`, an
+# inline `Release-Note:` label, a bare `# Release notes` in a *linked* comment,
+# or interleaved with reviewer instructions. A regex can't reliably tell a real
+# note from a template checklist line ("- [ ] Release note snippet added") or
+# reviewer chatter. Reading and interpreting that text is the skill's job:
+# run this script with `--json` to get each item's full body + comments, then
+# read them. This inventory is just the deterministic "what's in the release".
 # ------------------------------
-echo "## Items grouped by repository"
+echo "## Item inventory grouped by repository"
+echo "# For the release-note text of each item, run this script with --json and read the body + comments."
 
 REPOS=$(jq -r '[.[] | .repository_url | sub(".*/repos/stackabletech/";"")] | unique | .[]' "$ITEMS_FILE")
 
 while IFS= read -r repo; do
     [ -z "$repo" ] && continue
     echo
     echo "### $repo"
-    # Iterate items for this repo (compact JSON per line)
-    jq -c --arg R "$repo" '.[] | select((.repository_url | sub(".*/repos/stackabletech/";"")) == $R)' "$ITEMS_FILE" \
-    | while IFS= read -r item; do
-        number=$(echo "$item" | jq -r '.number')
-        title=$(echo "$item" | jq -r '.title')
-        url=$(echo "$item" | jq -r '.html_url')
-        labels=$(echo "$item" | jq -r '[.labels[].name] | join(", ")')
-        if echo "$item" | jq -e '.pull_request == null' >/dev/null; then
-            kind="issue"
-        elif echo "$item" | jq -e '.pull_request.merged_at != null' >/dev/null; then
-            kind="PR-merged"
-        else
-            kind="PR-$(echo "$item" | jq -r '.state')"
-        fi
-        marker=""
-        if is_action_required "$item"; then marker=" **ACTION-REQUIRED**"; fi
-
-        echo "- #$number [$kind]$marker $title"
-        echo "  $url"
-        echo "  labels: $labels"
-        # Extract the "Release note(s)" section from the body, if present.
-        snippet=$(echo "$item" | jq -r '.body // ""' | awk '
-            BEGIN { grab=0 }
-            /^#+[[:space:]]*[Rr]elease[[:space:]]*[Nn]ote/ { grab=1; next }
-            grab && /^#+[[:space:]]/ { exit }
-            grab { print }
-        ' | sed 's/^/    /' | sed '/^[[:space:]]*$/d')
-        if [ -n "$snippet" ]; then
-            echo "  release-note snippet:"
-            echo "$snippet"
-        else
-            echo "  release-note snippet: (none found in body)"
-        fi
-    done
+    jq -r --arg R "$repo" '
+        def kind: if .pull_request == null then "issue"
+                  elif .pull_request.merged_at != null then "PR-merged"
+                  else "PR-" + .state end;
+        def ar: if any(.labels[].name; . == "release-note/action-required")
+                then " **ACTION-REQUIRED**" else "" end;
+        [.[] | select((.repository_url | sub(".*/repos/stackabletech/";"")) == $R)]
+        | sort_by(.number)[]
+        | "- #\(.number) [\(kind)]\(ar) \(.title)\n  \(.html_url)\n  labels: \([.labels[].name] | join(", "))"
+    ' "$ITEMS_FILE"
 done <<< "$REPOS"