feat(images): auto-select relevant article images and tidy detail header

backend/app/ingestion.py

@@ -4,8 +4,10 @@ from dataclasses import dataclass
 from datetime import datetime, timezone
 import hashlib
 import json
+import re
 import time
 from typing import Any
+from urllib.parse import unquote, urlparse
 
 import feedparser
 
@@ -67,6 +69,72 @@ def _parsed_get(parsed: object, key: str, default: object = None) -> object:
     return getattr(parsed, key, default)
 
 
+def _normalize_tokens(text: str) -> set[str]:
+    normalized = re.sub(r"[^a-z0-9]+", " ", text.lower())
+    return {token for token in normalized.split() if len(token) >= 4}
+
+
+def _rank_image_candidates(source_url: str, title: str, images: list[str]) -> list[dict[str, Any]]:
+    source_host = (urlparse(source_url).hostname or "").lower()
+    is_presseportal = "presseportal.de" in source_host
+    title_tokens = _normalize_tokens(title)
+    blocked_patterns = ("logo", "badge", "app-store", "google-play", "na-logo", "sprite", "icon", "favicon", "tracking", "pixel")
+
+    ranked: list[dict[str, Any]] = []
+    for url in images:
+        parsed = urlparse(url)
+        path = unquote(parsed.path.lower())
+        full = f"{parsed.netloc.lower()}{path}"
+        score = 0
+        reasons: list[str] = []
+
+        if any(token in full for token in blocked_patterns):
+            score -= 150
+            reasons.append("blocked-pattern")
+
+        if is_presseportal and "/thumbnail/story_big/" in path:
+            score += 120
+            reasons.append("presseportal-story-big")
+        elif is_presseportal and "/thumbnail/highlight/" in path:
+            score += 45
+            reasons.append("presseportal-highlight")
+        elif is_presseportal and "/thumbnail/liste/" in path:
+            score -= 40
+            reasons.append("presseportal-list")
+
+        if "crop=" in (parsed.query or "").lower():
+            score -= 10
+            reasons.append("cropped-preview")
+
+        path_tokens = _normalize_tokens(path.replace("-", " "))
+        overlap = len(title_tokens.intersection(path_tokens))
+        if overlap > 0:
+            score += min(30, overlap * 6)
+            reasons.append(f"title-match:{overlap}")
+
+        ranked.append({"url": url, "score": score, "reasons": reasons})
+
+    ranked.sort(key=lambda item: item["score"], reverse=True)
+    return ranked
+
+
+def _select_relevant_images(source_url: str, title: str, images: list[str], max_keep: int = 3) -> tuple[list[str], str | None, list[dict[str, Any]]]:
+    # dedupe incoming order first
+    deduped: list[str] = []
+    seen: set[str] = set()
+    for image in images:
+        if image and image not in seen:
+            seen.add(image)
+            deduped.append(image)
+
+    ranked = _rank_image_candidates(source_url, title, deduped)
+    kept = [item["url"] for item in ranked if item["score"] > 0][:max_keep]
+    if not kept and ranked:
+        kept = [ranked[0]["url"]]
+    primary = kept[0] if kept else None
+    return kept, primary, ranked
+
+
 def run_ingestion(feed_id: int | None = None) -> IngestionStats:
     run_id = create_run(RunCreate(run_type="ingestion", status="running", details="started"))
     feeds_processed = 0
@@ -167,6 +235,12 @@ def run_ingestion(feed_id: int | None = None) -> IngestionStats:
                 final_summary = extracted.summary or (summary[:1000] if summary else None)
                 final_content_raw = extracted.content_text or content_raw
                 final_canonical = extracted.canonical_url or entry.get("link")
+                selected_images, primary_image, ranked_images = _select_relevant_images(
+                    link,
+                    final_title,
+                    extracted.images,
+                    max_keep=3,
+                )
 
                 source_hash = _entry_hash(
                     entry,
@@ -188,6 +262,12 @@ def run_ingestion(feed_id: int | None = None) -> IngestionStats:
                 }
                 extraction_meta: dict[str, Any] = extracted_article_to_meta(extracted)
                 extraction_meta["fetched_from"] = link
+                extraction_meta["image_selection"] = {
+                    "primary": primary_image,
+                    "selected_count": len(selected_images),
+                    "total_candidates": len(extracted.images),
+                    "ranked": ranked_images,
+                }
                 article_id = upsert_article(
                     ArticleUpsert(
                         feed_id=int(feed["id"]),
@@ -201,7 +281,7 @@ def run_ingestion(feed_id: int | None = None) -> IngestionStats:
                         summary=final_summary,
                         content_raw=final_content_raw,
                         content_rewritten=None,
-                        image_urls_json=json.dumps(extracted.images, ensure_ascii=False) if extracted.images else None,
+                        image_urls_json=json.dumps(selected_images, ensure_ascii=False) if selected_images else None,
                         press_contact=extracted.press_contact,
                         source_name_snapshot=feed.get("source_name"),
                         source_terms_url_snapshot=feed.get("source_terms_url"),

backend/static/admin.css

@@ -179,6 +179,27 @@ button.secondary {
   background: #f8fafc;
 }
 
+.detail-grid {
+  display: grid;
+  grid-template-columns: repeat(auto-fit, minmax(220px, 1fr));
+  gap: 8px 12px;
+  margin-bottom: 10px;
+}
+
+.detail-item {
+  background: #f8fafc;
+  border: 1px solid #e2e8f0;
+  border-radius: 8px;
+  padding: 8px;
+  display: grid;
+  gap: 4px;
+}
+
+.detail-item .k {
+  font-size: 12px;
+  color: #64748b;
+}
+
 .thumb {
   width: 72px;
   height: 72px;

backend/templates/admin_article_detail.html

@@ -23,19 +23,19 @@
   <main class="container">
     <section class="card">
       <h2>{{ article.title }}</h2>
-      <p><strong>Status:</strong> <span class="badge">{{ article.status }}</span></p>
+      <div class="detail-grid">
-      <p><strong>Artikel-Datum:</strong> {{ article.published_at or "-" }}</p>
+        <div class="detail-item"><span class="k">Status</span><span><span class="badge">{{ article.status }}</span></span></div>
-      <p><strong>Alter:</strong> {{ article.days_old if article.days_old is not none else "-" }} Tage</p>
+        <div class="detail-item"><span class="k">Artikel-Datum</span><span>{{ article.published_at or "-" }}</span></div>
-      <p><strong>Relevanz:</strong> {{ article.relevance }}</p>
+        <div class="detail-item"><span class="k">Alter</span><span>{{ article.days_old if article.days_old is not none else "-" }} Tage</span></div>
-      <p><strong>Autor:</strong> {{ article.author or "-" }}</p>
+        <div class="detail-item"><span class="k">Relevanz</span><span>{{ article.relevance }}</span></div>
-      <p><strong>Feed:</strong> {{ feed.name if feed else "-" }}</p>
+        <div class="detail-item"><span class="k">Autor</span><span>{{ article.author or "-" }}</span></div>
-      <p><strong>Quelle Snapshot:</strong> {{ article.source_name_snapshot or "-" }}</p>
+        <div class="detail-item"><span class="k">Feed</span><span>{{ feed.name if feed else "-" }}</span></div>
-      <p><strong>Lizenz Snapshot:</strong> {{ article.source_license_name_snapshot or "-" }}</p>
+        <div class="detail-item"><span class="k">Quelle Snapshot</span><span>{{ article.source_name_snapshot or "-" }}</span></div>
-      <p><strong>Terms Snapshot:</strong> {{ article.source_terms_url_snapshot or "-" }}</p>
+        <div class="detail-item"><span class="k">Lizenz Snapshot</span><span>{{ article.source_license_name_snapshot or "-" }}</span></div>
+        <div class="detail-item"><span class="k">Terms Snapshot</span><span>{{ article.source_terms_url_snapshot or "-" }}</span></div>
+      </div>
       <p><strong>Quelle:</strong> <a href="{{ article.source_url }}" target="_blank" rel="noopener">{{ article.source_url }}</a></p>
-      {% if article.canonical_url %}
+      {% if article.canonical_url %}<p><strong>Canonical:</strong> <a href="{{ article.canonical_url }}" target="_blank" rel="noopener">{{ article.canonical_url }}</a></p>{% endif %}
-      <p><strong>Canonical:</strong> <a href="{{ article.canonical_url }}" target="_blank" rel="noopener">{{ article.canonical_url }}</a></p>
-      {% endif %}
       {% if article.summary %}
       <p><strong>Summary:</strong> {{ article.summary }}</p>
       {% endif %}

backend/tests/test_ingestion.py

@@ -85,6 +85,7 @@ class TestIngestion(unittest.TestCase):
         self.assertEqual(article["author"], "Autorin A")
         self.assertIn("Original Volltext", article["content_raw"] or "")
         self.assertIn("Pressekontakt", article["meta_json"] or "")
+        self.assertIsNotNone(article["image_urls_json"])
 
     @patch("backend.app.ingestion.extract_article")
     @patch("backend.app.ingestion.feedparser.parse")