Scrape: include libretro + OpenVGDB covers alongside IGDB

app/covers.py

@@ -466,6 +466,33 @@ def _ovgdb_name_list():
     return _ovgdb_names
 
 
+_ovgdb_index = None
+
+
+def _load_ovgdb_index():
+    """Load {key: {"title", "url"}} written by app/openvgdb.py, or {}.
+
+    Looks in the runtime covers dir first, then the shipped ``data/covers``
+    copy (so a fresh install still shows readable titles without re-running
+    the full OpenVGDB scrape).
+    """
+    global _ovgdb_index
+    if _ovgdb_index is not None:
+        return _ovgdb_index
+    _ovgdb_index = {}
+    for path in (
+        config.COVERS_DIR / "openvgdb-index.json",
+        config.LEGACY_COVERS_DIR / "openvgdb-index.json",
+    ):
+        if path.exists():
+            try:
+                _ovgdb_index = json.loads(path.read_text(encoding="utf-8"))
+                break
+            except Exception:  # noqa: BLE001
+                _ovgdb_index = {}
+    return _ovgdb_index
+
+
 def fetch_openvgdb_local(name):
     """Return a bundled OpenVGDB cover image for `name`, or None."""
     names = _ovgdb_name_list()
@@ -1028,6 +1055,45 @@ def _search_libretro(query, limit):
     return results
 
 
+def _search_openvgdb(query, limit):
+    """Return [{title, source, thumb_url, image_url}] from the offline OpenVGDB
+    bundle, using the display titles stored in openvgdb-index.json."""
+    names = _ovgdb_name_list()
+    if not names:
+        return []
+    index = _load_ovgdb_index()
+    db = _norm(query)
+    if not db:
+        return []
+    scored = []
+    for key in names:
+        if key == db:
+            score = 5
+        elif len(key) >= 4 and key.startswith(db):
+            score = 4
+        elif len(key) >= 4 and db in key:
+            score = 3
+        elif len(db) >= 4 and len(key) >= 4 and key in db:
+            score = 2
+        else:
+            continue
+        info = index.get(key) or {}
+        title = info.get("title") or key
+        scored.append((score, title, key))
+    scored.sort(key=lambda t: (-t[0], t[1]))
+    return [
+        {
+            "id": None,
+            "title": title,
+            "source": "openvgdb",
+            "thumb_url": "",
+            "image_url": "",
+            "_key": key,
+        }
+        for _score, title, key in scored[:limit]
+    ]
+
+
 # --------------------------------------------------------------------------- #
 # Game metadata (details sidebar)
 # --------------------------------------------------------------------------- #
@@ -1411,29 +1477,46 @@ def movie_metadata_from_wikipedia(page_title, timeout=10):
 
 
 def search_metadata_candidates(system, query, limit=20, timeout=10):
-    """Search candidates the user can link an item to."""
+    """Search candidates the user can link an item to.
+
+    Games: IGDB (rich metadata) first, then libretro box art and OpenVGDB
+    (GameFAQs) box art so covers can be scraped even without an IGDB key.
+    """
     if system == "videos":
         return search_movies(query, limit, timeout)
-    results = search_games_igdb(query, limit, timeout)
-    if results:
-        return results
-    # No IGDB results (or no key): fall back to libretro catalog names so the
-    # user can still link to the correct entry (title-only).
-    results = _search_libretro(query, limit)
-    for r in results:
-        r["source"] = "libretro"
-        r["id"] = None
-    return results
+    results = []
+    seen = set()
+
+    def add(r):
+        key = (r.get("source"), _norm(r.get("title") or ""))
+        if key[1] and key not in seen:
+            seen.add(key)
+            results.append(r)
+
+    for r in search_games_igdb(query, limit, timeout):
+        add(r)
+    for r in _search_libretro(query, limit):
+        r.setdefault("id", None)
+        add(r)
+    for r in _search_openvgdb(query, limit):
+        add(r)
+    return results[:limit]
 
 
 def link_metadata(system, candidate, timeout=10):
     """Fetch full metadata for a search candidate the user picked."""
     if candidate.get("source") == "wikipedia":
         return movie_metadata_from_wikipedia(candidate.get("title") or "", timeout)
-    if candidate.get("source") == "libretro":
-        # Catalog-name link only: no metadata backend, just the correct title.
+    if candidate.get("source") in ("libretro", "openvgdb"):
+        # Cover repositories have no metadata backend; try IGDB by title so the
+        # user still gets synopsis/genre/etc. when a key is configured.
+        title = candidate.get("title") or ""
+        if title:
+            meta = get_game_metadata_igdb(title, timeout)
+            if meta:
+                return meta
         return {
-            "title": candidate.get("title") or "",
+            "title": title,
             "release_date": "",
             "overview": "",
             "genres": [],

app/gui.py

@@ -824,8 +824,17 @@ class ScrapeSearchWorker(QObject):
     def run(self):
         found = covers.search_metadata_candidates(self.system, self.query, limit=20, timeout=10)
         for r in found:
-            url = r.get("thumb_url")
-            r["thumb_bytes"] = covers.download_image(url, timeout=8) if url else None
+            src = r.get("source")
+            if src == "openvgdb":
+                r["thumb_bytes"] = covers.fetch_openvgdb_local(r.get("title") or "")
+            elif src == "libretro":
+                data = covers.fetch_libretro_local(r.get("title") or "")
+                r["thumb_bytes"] = data if data else (
+                    covers.download_image(r.get("thumb_url"), timeout=8) if r.get("thumb_url") else None
+                )
+            else:
+                url = r.get("thumb_url")
+                r["thumb_bytes"] = covers.download_image(url, timeout=8) if url else None
         self.results.emit(found)
         self.finished.emit()
 
@@ -849,6 +858,17 @@ class ScrapeFetchWorker(QObject):
 
 def _scrape_cover(system, candidate, rid, timeout=20):
     """Download the cover for a metadata candidate; returns bytes or None."""
+    src = candidate.get("source")
+    title = candidate.get("title") or ""
+    # Cover repositories ship offline bundles — read them directly (no network).
+    if src == "openvgdb" and title:
+        data = covers.fetch_openvgdb_local(title)
+        if data:
+            return data
+    if src == "libretro" and title:
+        data = covers.fetch_libretro_local(title)
+        if data:
+            return data
     url = candidate.get("image_url")
     if url:
         data = covers.download_image(url, timeout)
@@ -858,7 +878,6 @@ def _scrape_cover(system, candidate, rid, timeout=20):
             if candidate.get("source") == "igdb":
                 return covers.compose_game_cover(data)
             return data
-    title = candidate.get("title") or ""
     if title:
         path = covers.fetch_cover(title, rid, system, timeout=timeout)
         if path is not None:
@@ -955,6 +974,8 @@ class ScrapeDialog(QDialog):
                 icon = QIcon(pm)
             year = r.get("year") or (r.get("release_date") or "")[:4]
             label = f"{r['title']} ({year})" if year else r["title"]
+            if r.get("source") not in (None, "igdb", "tmdb"):
+                label += f"  [{r['source']}]"
             item = QListWidgetItem(icon, label)
             if r.get("overview"):
                 item.setToolTip(r["overview"])

app/openvgdb.py

@@ -17,6 +17,7 @@ tags removed). The GUI reads this bundle with no network access.
 """
 import argparse
 import concurrent.futures
+import json
 import logging
 import os
 import re
@@ -94,6 +95,19 @@ def build_index(sqlite_path):
     return best
 
 
+def write_index(index, path=None):
+    """Persist {key: {"title", "url"}} so the GUI can show readable titles when
+    searching the offline bundle (see covers._search_openvgdb)."""
+    path = Path(path) if path else (config.COVERS_DIR / "openvgdb-index.json")
+    data = {k: {"title": t, "url": u} for k, (t, u, _r) in index.items()}
+    try:
+        path.parent.mkdir(parents=True, exist_ok=True)
+        path.write_text(json.dumps(data, ensure_ascii=False), encoding="utf-8")
+    except OSError as exc:  # noqa: BLE001
+        log.warning("could not write index %s: %s", path, exc)
+    return path
+
+
 def _existing_keys():
     """Titles already present in the bundle (for resume)."""
     if not BUNDLE.exists():
@@ -152,6 +166,7 @@ def scrape(limit=None, workers=8, rate=3.0):
 
     index = build_index(sqlite_path)
     log.info("indexed %d unique PSP covers", len(index))
+    write_index(index)
 
     existing = _existing_keys()
     todo = [(k, v) for k, v in sorted(index.items()) if k not in existing]