ci(whatsnew): generate release highlights via GitHub Models

Activate the What's New highlights step that condenses each release's Features into <=8 short bullets and injects a  block the app reads (utils/whatsNew.parseWhatsNew), with a deterministic fallback. Runs as a needs: job inside the release-please workflows rather than on a standalone release: published trigger, because release-please creates the release with GITHUB_TOKEN and GitHub never starts new workflow runs from token-generated events -- a standalone trigger would never fire. Shared as a reusable workflow_call so the stable and beta channels stay in sync. Best-effort: continue-on-error + fallback mean it can never break a release. Requires GitHub Models enabled for the org; until then the fallback is used.
2026-06-30 16:43:54 +02:00
parent 25bf7bb523
commit 5aeb6905ac
3 changed files with 56 additions and 23 deletions
@@ -1,76 +0,0 @@
-# What's New highlights — GitHub Models release step
-#
-# DRAFT — move this into `.github/workflows/` on the repo that publishes the
-# canonical releases the app reads (PicPeak/picpeak). It is kept under docs/
-# here so it does NOT auto-run on the fork.
-#
-# What it does: when a release is published (release-please), it condenses that
-# release's "### Features" into <=8 short bullets via GitHub Models (free tier,
-# `models: read`) and injects a `<!-- whatsnew -->` block at the top of the
-# release notes. The picpeak app reads that block (utils/whatsNew.parseWhatsNew),
-# and falls back to the raw Features list for releases without it — so this is
-# purely a quality upgrade, never a hard dependency.
-#
-# Validated end-to-end on a fork (extract -> openai/gpt-4o-mini -> inject into
-# real release notes; app parseWhatsNew() reads the block back). The one thing
-# that can't be checked from a fork: GitHub Models must be enabled for the
-# PicPeak ORG (owner setting). Failure is isolated anyway by `continue-on-error`
-# + the deterministic fallback below, so it can never break a release.
-
-name: What's New highlights
-
-on:
-  release:
-    types: [published]
-
-permissions:
-  contents: write   # to edit the release body
-  models: read      # GitHub Models (free tier)
-
-jobs:
-  highlights:
-    runs-on: ubuntu-latest
-    steps:
-      - name: Extract Features from the published release
-        id: feat
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          TAG: ${{ github.event.release.tag_name }}
-        run: |
-          BODY=$(gh release view "$TAG" --json body -q .body)
-          FEATURES=$(printf '%s\n' "$BODY" | awk '/^#{2,4} +Features/{f=1;next} /^#{1,4} +\S/{f=0} f')
-          { echo "features<<EOF"; printf '%s\n' "$FEATURES"; echo EOF; } >> "$GITHUB_OUTPUT"
-
-      - name: Summarize with GitHub Models
-        if: ${{ steps.feat.outputs.features != '' }}
-        id: ai
-        uses: actions/ai-inference@v1
-        with:
-          model: openai/gpt-4o-mini   # catalog id (verified present); openai/gpt-4.1-mini or openai/gpt-5-nano also work
-          system-prompt: >
-            You write release highlights for the admins of a self-hosted
-            photo-gallery + CRM app. Given raw changelog "Features" lines, output
-            AT MOST 8 markdown bullets, each 3-4 words, user-facing, no scopes,
-            no jargon, no issue numbers. One bullet per distinct user-visible
-            feature. Output ONLY "- " bullets, nothing else.
-          prompt: ${{ steps.feat.outputs.features }}
-
-      - name: Inject the What's New block
-        if: ${{ steps.feat.outputs.features != '' }}
-        continue-on-error: true   # never let highlights break a release
-        env:
-          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-          TAG: ${{ github.event.release.tag_name }}
-          AI: ${{ steps.ai.outputs.response }}
-          FEATURES: ${{ steps.feat.outputs.features }}
-        run: |
-          BULLETS="$AI"
-          # Deterministic fallback if the model returned nothing.
-          if [ -z "$BULLETS" ]; then
-            BULLETS=$(printf '%s\n' "$FEATURES" | head -8 \
-              | sed -E 's/^\* \*\*[^:]+:\*\* */- /; s/ \(\[[^]]*\]\([^)]*\)\)//g')
-          fi
-          BODY=$(gh release view "$TAG" --json body -q .body)
-          # Idempotent: strip any prior block before re-injecting.
-          BODY=$(printf '%s' "$BODY" | perl -0pe 's/<!--\s*whatsnew\s*-->.*?<!--\s*\/whatsnew\s*-->\n*//is')
-          gh release edit "$TAG" --notes "$(printf '<!-- whatsnew -->\n%s\n<!-- /whatsnew -->\n\n%s' "$BULLETS" "$BODY")"