merged ln metadata into manga mover

Introduce a new background service that periodically re-checks chapters
whose volume could not be resolved at move time.

- Add KavitaVolumeCoverUpdater.py to resolve null volumes via MangaDex,
  update ComicInfo.xml in-archive, and swap in MangaBaka volume covers
- Wire updater into main.py entry point with UPDATER_ENABLED env flag
- Add UPDATER_ENABLED env var to docker-compose.prod.yml
- Update CronSchedule.py to schedule updater runs

.env.example

@@ -0,0 +1,22 @@
+# Shared
+KAVITA_URL=http://192.168.1.100:5000
+KAVITA_API_KEY=your-api-key-here
+LANGUAGE=en
+TZ=Europe/Berlin
+
+# Manga container (manga-mover-and-metadata-collector)
+HOST_SUWAYOMI_PATH=/path/to/suwayomi/downloads
+HOST_KAVITA_PATH=/path/to/kavita/library
+HOST_MANGA_CONFIG_PATH=/path/to/manga-config
+MANGA_WEB_PORT=8080
+SETTLE_SECONDS=600
+DELETE_SOURCE=true
+UPDATER_ENABLED=true
+UPDATER_SCHEDULE=0 19 * * 1,4
+COVER_CACHE_PATH=/config/covers
+
+# Light-novel container (kavita-lightnovel-metadata-fetcher)
+HOST_LN_CONFIG_PATH=/path/to/ln-config
+LN_WEB_PORT=8081
+LN_LIBRARY_IDS=3,5
+

.gitea/workflows/build.yml

@@ -5,6 +5,11 @@ on:
     branches:
       - master
 
+env:
+  REGISTRY: gitea.johannesbot.de/johannesbot
+  MANGA_IMAGE: manga-mover-and-metadata-collector
+  LN_IMAGE: kavita-lightnovel-metadata-fetcher
+
 jobs:
   build:
     runs-on: ubuntu-latest
@@ -17,11 +22,16 @@ jobs:
           echo "${{ secrets.REGISTRY_PASSWORD }}" | \
           docker login https://gitea.johannesbot.de -u ${{ secrets.REGISTRY_USER }} --password-stdin
 
-      - name: Build Image
-        run: docker build -t gitea.johannesbot.de/johannesbot/manga-mover-and-metadata-collector:latest .
+      - name: Build Manga Image
+        run: docker build --build-arg APP=manga -t ${{ env.REGISTRY }}/${{ env.MANGA_IMAGE }}:latest .
 
-      - name: Push Image
-        run: docker push gitea.johannesbot.de/johannesbot/manga-mover-and-metadata-collector:latest
+      - name: Build LN Image
+        run: docker build --build-arg APP=ln -t ${{ env.REGISTRY }}/${{ env.LN_IMAGE }}:latest .
+
+      - name: Push Images
+        run: |
+          docker push ${{ env.REGISTRY }}/${{ env.MANGA_IMAGE }}:latest
+          docker push ${{ env.REGISTRY }}/${{ env.LN_IMAGE }}:latest
 
   deploy:
     needs: build
@@ -37,7 +47,7 @@ jobs:
           username: ${{ secrets.SSH_USER }}
           password: ${{ secrets.SSH_PASSWORD }}
           port: ${{ secrets.SSH_PORT || 22 }}
-          script: mkdir -p /home/${{ secrets.SSH_USER }}/manga-mover-and-metadata-collector
+          script: mkdir -p /home/${{ secrets.SSH_USER }}/kavita-metadata-collector
 
       - name: Copy docker-compose via SCP
         uses: appleboy/scp-action@v0.1.7
@@ -47,7 +57,7 @@ jobs:
           password: ${{ secrets.SSH_PASSWORD }}
           port: ${{ secrets.SSH_PORT || 22 }}
           source: "docker-compose.prod.yml"
-          target: "/home/${{ secrets.SSH_USER }}/manga-mover-and-metadata-collector"
+          target: "/home/${{ secrets.SSH_USER }}/kavita-metadata-collector"
 
       - name: Deploy via SSH
         uses: appleboy/ssh-action@v1.0.3
@@ -57,7 +67,7 @@ jobs:
           password: ${{ secrets.SSH_PASSWORD }}
           port: ${{ secrets.SSH_PORT || 22 }}
           script: |
-            cd /home/${{ secrets.SSH_USER }}/manga-mover-and-metadata-collector
+            cd /home/${{ secrets.SSH_USER }}/kavita-metadata-collector
             mv docker-compose.prod.yml docker-compose.yml
             echo "${{ secrets.REGISTRY_PASSWORD }}" | sudo docker login https://gitea.johannesbot.de -u ${{ secrets.REGISTRY_USER }} --password-stdin
             sudo docker compose pull

.gitea/workflows/release.yml

@@ -0,0 +1,38 @@
+name: Build Release
+
+on:
+  push:
+    tags:
+      - 'v*'
+
+env:
+  REGISTRY: gitea.johannesbot.de/johannesbot
+  MANGA_IMAGE: manga-mover-and-metadata-collector
+  LN_IMAGE: kavita-lightnovel-metadata-fetcher
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Login to Gitea Registry
+        run: |
+          echo "${{ secrets.REGISTRY_PASSWORD }}" | \
+          docker login https://gitea.johannesbot.de -u ${{ secrets.REGISTRY_USER }} --password-stdin
+
+      - name: Extract Tag
+        id: tag
+        run: echo "VERSION=${GITHUB_REF#refs/tags/v}" >> "$GITHUB_OUTPUT"
+
+      - name: Build Manga Image
+        run: docker build --build-arg APP=manga -t ${{ env.REGISTRY }}/${{ env.MANGA_IMAGE }}:${{ steps.tag.outputs.VERSION }} .
+
+      - name: Build LN Image
+        run: docker build --build-arg APP=ln -t ${{ env.REGISTRY }}/${{ env.LN_IMAGE }}:${{ steps.tag.outputs.VERSION }} .
+
+      - name: Push Images
+        run: |
+          docker push ${{ env.REGISTRY }}/${{ env.MANGA_IMAGE }}:${{ steps.tag.outputs.VERSION }}
+          docker push ${{ env.REGISTRY }}/${{ env.LN_IMAGE }}:${{ steps.tag.outputs.VERSION }}

Dockerfile

@@ -1,8 +1,18 @@
+# One Dockerfile, two images: the build arg APP selects the entry point.
+#
+#   docker build --build-arg APP=manga -t .../manga-mover-and-metadata-collector .
+#   docker build --build-arg APP=ln    -t .../kavita-lightnovel-metadata-fetcher .
+#
+# Both variants share src/; the variant-specific code lives in
+# src/manga/ resp. src/ln/ and is selected by the entry point.
+
 FROM python:3.12-slim
 
+ARG APP=manga
+
 WORKDIR /app
 
-# System deps for Pillow (image dimensions); kept minimal.
+# System deps for Pillow (image dimensions, manga variant); kept minimal.
 RUN apt-get update \
  && apt-get install -y --no-install-recommends \
         libjpeg62-turbo \
@@ -12,14 +22,16 @@ COPY requirements.txt .
 RUN pip install --no-cache-dir -r requirements.txt
 
 COPY src/          /app/src/
-COPY main.py  /app/main.py
+COPY main_manga.py main_ln.py /app/
 
 ENV PYTHONUNBUFFERED=1 \
-    PYTHONDONTWRITEBYTECODE=1
+    PYTHONDONTWRITEBYTECODE=1 \
+    APP_VARIANT=${APP}
 
-# Mount points used by main.py defaults
-VOLUME ["/mnt/suwayomi", "/mnt/kavita", "/config"]
+# /config is used by both variants; the manga variant additionally mounts
+# /mnt/suwayomi and /mnt/kavita (see docker-compose.prod.yml).
+VOLUME ["/config"]
 
 EXPOSE 8080
 
-CMD ["python", "/app/main.py"]
+CMD python /app/main_${APP_VARIANT}.py

docker-compose.prod.yml

@@ -1,5 +1,8 @@
 services:
-  manga-mover:
+  # ------------------------------------------------------------------
+  # Manga: Suwayomi -> Kavita mover + metadata enrichment
+  # ------------------------------------------------------------------
+  manga-mover-and-metadata-collector:
     image: gitea.johannesbot.de/johannesbot/manga-mover-and-metadata-collector:latest
     container_name: manga-mover-and-metadata-collector
     restart: unless-stopped
@@ -9,11 +12,39 @@ services:
       LANGUAGE:       "${LANGUAGE:-en}"
       SETTLE_SECONDS: "${SETTLE_SECONDS:-600}"
       DELETE_SOURCE:  "${DELETE_SOURCE:-true}"
-      MATCH_PATH:     "${MATCH_PATH:-/config/matches.json}"
-      WEB_PORT:       "${WEB_PORT:-8080}"
+      MATCH_PATH:     "/config/matches.json"
+      # Volume/cover back-fill updater
+      UPDATER_ENABLED:  "${UPDATER_ENABLED:-true}"
+      # Cron expression: "0 19 * * 1,4" = 19:00 every Monday and Thursday
+      # (local time, see TZ)
+      UPDATER_SCHEDULE: "${UPDATER_SCHEDULE:-0 19 * * 1,4}"
+      UPDATER_LOG:      "/config/volume_updater.log"
+      # Persistent cover cache (empty = temp dir, deleted on container stop)
+      COVER_CACHE_PATH: "${COVER_CACHE_PATH:-/config/covers}"
+      # Timezone for the cron schedule — without this 19:00 means 19:00 UTC
+      TZ:               "${TZ:-Europe/Berlin}"
     ports:
-      - "${WEB_PORT:-8080}:${WEB_PORT:-8080}"
+      - "${MANGA_WEB_PORT:-8080}:8080"
     volumes:
       - "${HOST_SUWAYOMI_PATH}:/mnt/suwayomi"
       - "${HOST_KAVITA_PATH}:/mnt/kavita"
-      - "${HOST_CONFIG_PATH}:/config"
+      - "${HOST_MANGA_CONFIG_PATH}:/config"
+
+  # ------------------------------------------------------------------
+  # Light novels: Kavita metadata fetcher (HTTP only, no file mover)
+  # ------------------------------------------------------------------
+  kavita-lightnovel-metadata-fetcher:
+    image: gitea.johannesbot.de/johannesbot/kavita-lightnovel-metadata-fetcher:latest
+    container_name: kavita-lightnovel-metadata-fetcher
+    restart: unless-stopped
+    environment:
+      KAVITA_URL:     "${KAVITA_URL}"
+      KAVITA_API_KEY: "${KAVITA_API_KEY}"
+      LIBRARY_IDS:    "${LN_LIBRARY_IDS}"
+      LANGUAGE:       "${LANGUAGE:-en}"
+      MATCH_PATH:     "/config/matches.json"
+      TZ:             "${TZ:-Europe/Berlin}"
+    ports:
+      - "${LN_WEB_PORT:-8081}:8080"
+    volumes:
+      - "${HOST_LN_CONFIG_PATH}:/config"

main_ln.py

@@ -0,0 +1,129 @@
+"""
+main_ln.py
+==========
+
+Container entry point for the **light-novel** variant (Kavita metadata
+fetcher).  The manga variant has its own entry point (main_manga.py);
+both share the modules in src/ and add their variant-specific code from
+src/ln/ resp. src/manga/.
+
+Reads configuration from environment variables, starts the orchestrator
+and exposes the Flask WebApp on WEB_HOST:WEB_PORT.  Everything happens
+through HTTP — there is no folder watcher and no file mover (Kavita is
+the source of truth for the library content; this service only writes
+metadata back to it).
+
+Environment variables
+---------------------
+  Required:
+    KAVITA_URL          base URL of the Kavita server, e.g. http://kavita:5000
+    KAVITA_API_KEY      Kavita API key (Settings -> User -> API key)
+
+  Optional:
+    LIBRARY_IDS         comma-separated default library ids (e.g. "3,5").
+                        Empty = user picks in the WebUI each time.
+    LANGUAGE            default "en"
+    REQUEST_TIMEOUT     default 30
+    MATCH_PATH          default /config/matches.json
+    WEB_PORT            default 8080
+    WEB_HOST            default 0.0.0.0
+"""
+
+from __future__ import annotations
+
+import os
+import sys
+from pathlib import Path
+
+try:
+    from dotenv import load_dotenv
+    load_dotenv()
+except ImportError:
+    pass
+
+# Shared code in src/, LN-specific code in src/ln/.  Modules are imported
+# by their plain names so src-internal imports resolve to the same module
+# objects (a `src.X` import would load everything twice).
+_BASE = Path(__file__).resolve().parent
+sys.path.insert(0, str(_BASE / "src"))
+sys.path.insert(0, str(_BASE / "src" / "ln"))
+
+from MatchesCache import MatchesCache                       # noqa: E402
+from LightNovelOrchestrator import LightNovelOrchestrator   # noqa: E402
+from MatchesWebApp import MatchesWebApp                     # noqa: E402
+
+
+def _env_str(name: str, default: "str | None" = None,
+             required: bool = False) -> "str | None":
+    value = os.environ.get(name, default)
+    if required and not value:
+        print(f"[main] missing required env var: {name}", flush=True)
+        sys.exit(2)
+    return value
+
+
+def _env_int(name: str, default: int) -> int:
+    raw = os.environ.get(name)
+    if raw is None or raw == "":
+        return default
+    try:
+        return int(raw)
+    except ValueError:
+        print(f"[main] {name}={raw!r} is not a valid integer; "
+              f"falling back to {default}", flush=True)
+        return default
+
+
+def _env_int_list(name: str) -> list[int]:
+    raw = os.environ.get(name) or ""
+    out: list[int] = []
+    for part in raw.split(","):
+        part = part.strip()
+        if not part:
+            continue
+        try:
+            out.append(int(part))
+        except ValueError:
+            print(f"[main] {name}: ignoring non-integer value {part!r}",
+                  flush=True)
+    return out
+
+
+def main() -> int:
+    kavita_url      = _env_str("KAVITA_URL",     required=True)
+    kavita_api_key  = _env_str("KAVITA_API_KEY", required=True)
+    language        = _env_str("LANGUAGE", "en") or "en"
+    request_timeout = _env_int("REQUEST_TIMEOUT", 30)
+    match_path      = _env_str("MATCH_PATH", "/config/matches.json")
+    web_host        = _env_str("WEB_HOST", "0.0.0.0") or "0.0.0.0"
+    web_port        = _env_int("WEB_PORT", 8080)
+    library_ids     = _env_int_list("LIBRARY_IDS")
+
+    print(f"[main] kavita url    = {kavita_url}",     flush=True)
+    print(f"[main] language      = {language}",       flush=True)
+    print(f"[main] match path    = {match_path}",     flush=True)
+    print(f"[main] libraries     = {library_ids or '(picked in WebUI)'}",
+          flush=True)
+    print(f"[main] web           = {web_host}:{web_port}", flush=True)
+
+    cache = MatchesCache(match_path)
+    orchestrator = LightNovelOrchestrator(
+        kavita_url=kavita_url,
+        kavita_api_key=kavita_api_key,
+        matches_cache=cache,
+        language=language,
+        request_timeout=request_timeout,
+    )
+
+    app = MatchesWebApp(
+        cache, orchestrator=orchestrator,
+        default_library_ids=library_ids,
+        host=web_host, port=web_port,
+    )
+    app.start()
+    app.wait()
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

main_manga.py

@@ -1,10 +1,11 @@
 """
-main.py
-=======
+main_manga.py
+=============
 
-Container entry point.  Watches the mounted Suwayomi download directory
-and, after a quiet period, triggers SuwayomiMover (which also runs the
-Kavita person sync for every processed series).
+Container entry point for the **manga** variant (Suwayomi -> Kavita mover
+plus metadata enrichment).  The light-novel variant has its own entry
+point (main_ln.py); both share the modules in src/ and add their
+variant-specific code from src/manga/ resp. src/ln/.
 
 Mount points (Docker)
 ---------------------
@@ -15,7 +16,7 @@ Environment variables
 ---------------------
   Required:
     KAVITA_URL          base URL of the Kavita server, e.g. http://kavita:5000
-    KAVITA_API_KEY      Kavita API key (Settings → User → API key)
+    KAVITA_API_KEY      Kavita API key (Settings -> User -> API key)
 
   Optional:
     SUWAYOMI_PATH       default /mnt/suwayomi
@@ -27,22 +28,39 @@ Environment variables
     MATCH_PATH          default /config/matches.json
     WEB_PORT            default 8080  (Flask web UI for matches.json)
     WEB_HOST            default 0.0.0.0
+    UPDATER_ENABLED     default true  (volume/cover back-fill cron)
+    UPDATER_SCHEDULE    cron expression for the updater scans,
+                        default "0 19 * * 1,4" = 19:00 every Mon + Thu
+                        (local time — set TZ inside the container!)
+    UPDATER_LOG         default /config/volume_updater.log
+    COVER_CACHE_PATH    directory for the persistent cover cache;
+                        empty (default) = temporary cache, deleted on exit
 """
 
 from __future__ import annotations
 
 import os
-import signal
 import sys
 from pathlib import Path
 
-# Make src/ importable when running as `python main.py`.
-sys.path.insert(0, str(Path(__file__).resolve().parent / "src"))
+try:
+    from dotenv import load_dotenv
+    load_dotenv()
+except ImportError:
+    pass
 
-from src.SuwayomiMover import SuwayomiMover                       # noqa: E402
-from src.SuwayomiFolderWatcher import SuwayomiFolderWatcher       # noqa: E402
-from src.MatchesCache import MatchesCache                          # noqa: E402
-from src.MatchesWebApp import MatchesWebApp                        # noqa: E402
+# Shared code in src/, manga-specific code in src/manga/.  Modules are
+# imported by their plain names so src-internal imports resolve to the
+# same module objects (a `src.X` import would load everything twice).
+_BASE = Path(__file__).resolve().parent
+sys.path.insert(0, str(_BASE / "src"))
+sys.path.insert(0, str(_BASE / "src" / "manga"))
+
+from SuwayomiMover import SuwayomiMover                        # noqa: E402
+from SuwayomiFolderWatcher import SuwayomiFolderWatcher        # noqa: E402,F401
+from MatchesCache import MatchesCache                          # noqa: E402
+from MatchesWebApp import MatchesWebApp                        # noqa: E402
+from KavitaVolumeCoverUpdater import KavitaVolumeCoverUpdater  # noqa: E402
 
 
 def _env_str(name: str, default: "str | None" = None,
@@ -76,8 +94,8 @@ def _env_bool(name: str, default: bool) -> bool:
 def main() -> int:
     suwayomi_path   = _env_str("SUWAYOMI_PATH", "/mnt/suwayomi")
     kavita_path     = _env_str("KAVITA_PATH",   "/mnt/kavita")
-    kavita_url      = _env_str("KAVITA_URL",     required=True)
-    kavita_api_key  = _env_str("KAVITA_API_KEY", required=True)
+    kavita_url      = _env_str("KAVITA_URL",     "http://kavita:5000")
+    kavita_api_key  = _env_str("KAVITA_API_KEY",  "")
     language        = _env_str("LANGUAGE", "en") or "en"
     settle_seconds  = _env_int("SETTLE_SECONDS",  600)
     request_timeout = _env_int("REQUEST_TIMEOUT",  30)
@@ -85,6 +103,10 @@ def main() -> int:
     match_path      = _env_str("MATCH_PATH", "/config/matches.json")
     web_host        = _env_str("WEB_HOST", "0.0.0.0") or "0.0.0.0"
     web_port        = _env_int("WEB_PORT", 8080)
+    updater_enabled  = _env_bool("UPDATER_ENABLED", True)
+    updater_schedule = _env_str("UPDATER_SCHEDULE", "0 19 * * 1,4")
+    updater_log      = _env_str("UPDATER_LOG", "/config/volume_updater.log")
+    cover_cache_path = _env_str("COVER_CACHE_PATH", "") or None
 
     print(f"[main] suwayomi  = {suwayomi_path}",  flush=True)
     print(f"[main] kavita    = {kavita_path}",    flush=True)
@@ -105,6 +127,7 @@ def main() -> int:
         request_timeout=request_timeout,
         delete_source=delete_source,
         matches_cache=matches_cache,
+        cover_cache_dir=cover_cache_path,
     )
 
     # watcher = SuwayomiFolderWatcher(suwayomi_path, mover, settle_seconds=settle_seconds)
@@ -112,13 +135,24 @@ def main() -> int:
     web_app = MatchesWebApp(matches_cache, mover=mover, host=web_host, port=web_port)
     web_app.start()
 
-    # def shutdown(signum, _frame):
-    #     print(f"[main] received signal {signum}", flush=True)
-    #     watcher.stop()
-    #
-    # signal.signal(signal.SIGTERM, shutdown)
-    # signal.signal(signal.SIGINT,  shutdown)
-    #
+    if updater_enabled:
+        try:
+            updater = KavitaVolumeCoverUpdater(
+                kavita_path,
+                matches_cache=matches_cache,
+                language=language,
+                request_timeout=request_timeout,
+                log_path=updater_log,
+                schedule=updater_schedule,
+                cover_cache_dir=cover_cache_path,
+            )
+            updater.start()
+        except ValueError as exc:
+            # Invalid cron expression — keep the service up, just without
+            # the updater, and make the config error obvious in the logs.
+            print(f"[main] UPDATER_SCHEDULE invalid ({exc}); "
+                  f"volume/cover updater DISABLED", flush=True)
+
     # watcher.start()
     # watcher.wait()   # blocks until stop() is called via a signal
     web_app.wait()     # keep process alive while the watcher is disabled

requirements.txt

@@ -2,3 +2,4 @@ requests>=2.31
 Pillow>=10.0
 watchdog>=4.0
 Flask>=3.0
+python-dotenv>=1.0

src/AniListResolver.py

@@ -32,27 +32,35 @@ Dependencies
 from __future__ import annotations
 
 import datetime
-import difflib
 import time
 
 import requests
 
 from MediaResolver import MediaResolver
+from TextUtils import best_similarity
 
 
 # --------------------------------------------------------------------------
 # GraphQL query strings
 # --------------------------------------------------------------------------
-_SEARCH_MANGA = """
+# AniList models both manga and light novels as type MANGA; the format
+# clause decides which of the two a search returns.  The placeholder is
+# substituted at construction time (see `media_format`).
+_SEARCH_MANGA_TEMPLATE = """
 query ($search: String) {
   Page(page: 1, perPage: 5) {
-    media(search: $search, type: MANGA, format_not_in: [NOVEL]) {
+    media(search: $search, type: MANGA, __FORMAT_CLAUSE__) {
       id title { romaji english native } siteUrl
     }
   }
 }
 """
 
+_FORMAT_CLAUSES = {
+    "manga": "format_not_in: [NOVEL]",
+    "novel": "format_in: [NOVEL]",
+}
+
 _MANGA_STATS = """
 query ($id: Int) {
   Media(id: $id, type: MANGA) {
@@ -131,10 +139,24 @@ class AniListResolver(MediaResolver):
             cls._instance._initialized = False
         return cls._instance
 
-    def __init__(self, *, request_timeout: int = 30):
+    def __init__(self, *, request_timeout: int = 30,
+                 media_format: str = "manga"):
+        """
+        media_format : "manga" (excludes novels) or "novel" (novels only).
+                       Only the FIRST construction in the process sets it
+                       (singleton); construct the resolver with the correct
+                       format in the entry point / orchestrator.
+        """
         if self._initialized:
             return
 
+        if media_format not in _FORMAT_CLAUSES:
+            raise ValueError(f"media_format must be one of "
+                             f"{sorted(_FORMAT_CLAUSES)}, got {media_format!r}")
+        self.media_format = media_format
+        self._search_query = _SEARCH_MANGA_TEMPLATE.replace(
+            "__FORMAT_CLAUSE__", _FORMAT_CLAUSES[media_format])
+
         self.request_timeout = request_timeout
 
         self._session = requests.Session()
@@ -178,7 +200,7 @@ class AniListResolver(MediaResolver):
             return self._id_cache[key]
 
         try:
-            data = self._gql(_SEARCH_MANGA, {"search": title})
+            data = self._gql(self._search_query, {"search": title})
             results = ((data.get("data") or {})
                        .get("Page", {})
                        .get("media") or [])
@@ -469,18 +491,11 @@ class AniListResolver(MediaResolver):
 def _score_title(query: str, entry: dict) -> float:
     """Returns the best title-similarity score for an AniList media entry."""
     title_obj = entry.get("title") or {}
-    candidates = [
-        title_obj.get("romaji") or "",
-        title_obj.get("english") or "",
-        title_obj.get("native") or "",
-    ]
-    best = 0.0
-    q = query.lower()
-    for t in candidates:
-        if t:
-            ratio = difflib.SequenceMatcher(None, q, t.lower()).ratio()
-            best = max(best, ratio)
-    return best
+    return best_similarity(query, (
+        title_obj.get("romaji"),
+        title_obj.get("english"),
+        title_obj.get("native"),
+    ))
 
 
 # --------------------------------------------------------------------------

src/CoverCache.py

@@ -0,0 +1,148 @@
+"""
+cover_cache.py
+==============
+
+Disk-backed cache for downloaded cover images, keyed by URL.
+
+Why
+---
+The mover packs every chapter of a series individually, and each chapter
+needs a cover image.  Without caching, the same multi-megabyte cover is
+downloaded once per chapter (20-chapter volume = 20 identical downloads).
+This cache turns that into a single download per unique URL.
+
+Persistence
+-----------
+* ``cache_dir`` given     -> covers persist across runs in that directory.
+* ``cache_dir`` omitted   -> a temporary directory is used and removed
+                             automatically when the process exits.
+
+Files are stored as ``<sha256(url)[:32]><ext>``; the extension is derived
+from the URL / Content-Type at download time so it can be reused when
+writing the cover into a chapter folder.
+
+Thread safety: downloads are serialised per cache instance, so concurrent
+mover / updater threads never fetch the same URL twice.
+
+Dependencies
+------------
+    requests    ->  pip install requests
+"""
+
+from __future__ import annotations
+
+import atexit
+import hashlib
+import shutil
+import tempfile
+import threading
+from pathlib import Path
+
+import requests
+
+
+_IMAGE_EXTS = {".jpg", ".jpeg", ".png", ".gif", ".webp", ".bmp", ".avif"}
+
+
+def _guess_extension(url: str, content_type: str) -> str:
+    """Derives an image file extension from a URL or HTTP Content-Type."""
+    url_ext = Path(url.split("?")[0]).suffix.lower()
+    if url_ext in _IMAGE_EXTS:
+        return url_ext
+    ct = (content_type or "").lower()
+    if "png"  in ct: return ".png"
+    if "webp" in ct: return ".webp"
+    if "gif"  in ct: return ".gif"
+    return ".jpg"
+
+
+class CoverCache:
+    """
+    URL-keyed image cache on disk.
+
+    Parameters
+    ----------
+    cache_dir       : Directory for cached covers.  None -> temporary
+                      directory, deleted automatically at process exit.
+    session         : Optional shared requests.Session for downloads.
+    request_timeout : HTTP timeout in seconds.
+    """
+
+    def __init__(self, cache_dir=None, *,
+                 session: "requests.Session | None" = None,
+                 request_timeout: int = 30):
+        self._persistent = cache_dir is not None
+        if self._persistent:
+            self._dir = Path(cache_dir)
+            self._dir.mkdir(parents=True, exist_ok=True)
+        else:
+            self._dir = Path(tempfile.mkdtemp(prefix="cover_cache_"))
+            atexit.register(self.close)
+
+        self._session = session or requests.Session()
+        self._session.headers.setdefault("User-Agent", "CoverCache/1.0")
+        self._timeout = request_timeout
+        self._lock = threading.Lock()
+
+    # ------------------------------------------------------------------
+    # Public API
+    # ------------------------------------------------------------------
+    def get(self, url: str) -> "tuple[bytes, str] | None":
+        """
+        Returns ``(image_bytes, extension)`` for the URL — from cache when
+        present, downloading (and caching) otherwise.  Returns None when
+        the URL is empty or the download fails.
+        """
+        if not url:
+            return None
+
+        with self._lock:
+            cached = self._find_cached(url)
+            if cached is not None:
+                try:
+                    return cached.read_bytes(), cached.suffix
+                except OSError:
+                    pass  # unreadable cache file -> re-download
+
+            return self._download(url)
+
+    def clear(self) -> None:
+        """Removes all cached covers (the directory itself is kept)."""
+        with self._lock:
+            for f in self._dir.glob("*"):
+                if f.is_file():
+                    f.unlink(missing_ok=True)
+
+    def close(self) -> None:
+        """Deletes the cache directory when it is non-persistent."""
+        if not self._persistent:
+            shutil.rmtree(self._dir, ignore_errors=True)
+
+    # ------------------------------------------------------------------
+    # Internal
+    # ------------------------------------------------------------------
+    @staticmethod
+    def _key(url: str) -> str:
+        return hashlib.sha256(url.encode("utf-8")).hexdigest()[:32]
+
+    def _find_cached(self, url: str) -> "Path | None":
+        matches = list(self._dir.glob(self._key(url) + ".*"))
+        return matches[0] if matches else None
+
+    def _download(self, url: str) -> "tuple[bytes, str] | None":
+        try:
+            resp = self._session.get(url, timeout=self._timeout)
+            resp.raise_for_status()
+        except requests.RequestException:
+            return None
+
+        ext = _guess_extension(url, resp.headers.get("Content-Type", ""))
+
+        target = self._dir / f"{self._key(url)}{ext}"
+        try:
+            tmp = target.with_suffix(target.suffix + ".tmp")
+            tmp.write_bytes(resp.content)
+            tmp.replace(target)
+        except OSError:
+            pass  # cache write failure is non-fatal — still return the bytes
+        return resp.content, ext

src/CronSchedule.py

@@ -0,0 +1,159 @@
+"""
+cron_schedule.py
+================
+
+Minimal cron-expression parser — no external dependency.
+
+Supports the classic 5-field syntax::
+
+    ┌──────── minute        (0-59)
+    │ ┌────── hour          (0-23)
+    │ │ ┌──── day of month  (1-31)
+    │ │ │ ┌── month         (1-12 or jan-dec)
+    │ │ │ │ ┌ day of week   (0-7 or sun-sat; 0 and 7 = Sunday)
+    │ │ │ │ │
+    0 19 * * 1,4    ->  19:00 every Monday and Thursday
+
+Field syntax: ``*``, single values, ranges (``a-b``), steps (``*/n``,
+``a-b/n``) and comma lists.  Month / weekday names (``jan``, ``mon``, …)
+are accepted case-insensitively.
+
+As in Vixie cron, when *both* day-of-month and day-of-week are restricted
+the job runs when **either** matches.
+
+Times are evaluated against the local system clock (``datetime.now()``) —
+in Docker set the ``TZ`` environment variable so "19:00" means local time.
+"""
+
+from __future__ import annotations
+
+from datetime import datetime, timedelta
+
+
+_MONTH_NAMES = {"jan": 1, "feb": 2, "mar": 3, "apr": 4, "may": 5, "jun": 6,
+                "jul": 7, "aug": 8, "sep": 9, "oct": 10, "nov": 11, "dec": 12}
+_DAY_NAMES   = {"sun": 0, "mon": 1, "tue": 2, "wed": 3, "thu": 4,
+                "fri": 5, "sat": 6}
+
+
+def _parse_value(token: str, lo: int, hi: int,
+                 names: "dict[str, int] | None") -> int:
+    token = token.strip().lower()
+    if names and token in names:
+        return names[token]
+    try:
+        value = int(token)
+    except ValueError:
+        raise ValueError(f"invalid cron value {token!r}") from None
+    if not (lo <= value <= hi):
+        raise ValueError(f"cron value {value} out of range {lo}-{hi}")
+    return value
+
+
+def _parse_field(field: str, lo: int, hi: int,
+                 names: "dict[str, int] | None" = None) -> "set[int]":
+    """Parses one cron field into the set of matching integer values."""
+    result: set[int] = set()
+    for part in field.split(","):
+        part = part.strip()
+        if not part:
+            raise ValueError(f"empty element in cron field {field!r}")
+
+        step = 1
+        if "/" in part:
+            part, step_text = part.split("/", 1)
+            try:
+                step = int(step_text)
+            except ValueError:
+                raise ValueError(f"invalid cron step {step_text!r}") from None
+            if step < 1:
+                raise ValueError(f"cron step must be >= 1, got {step}")
+
+        if part == "*":
+            start, end = lo, hi
+        elif "-" in part:
+            a, b = part.split("-", 1)
+            start = _parse_value(a, lo, hi, names)
+            end   = _parse_value(b, lo, hi, names)
+            if end < start:
+                raise ValueError(f"inverted cron range {part!r}")
+        else:
+            start = end = _parse_value(part, lo, hi, names)
+
+        result.update(range(start, end + 1, step))
+    return result
+
+
+class CronSchedule:
+    """
+    Parsed 5-field cron expression with ``next_after()`` evaluation.
+
+    Usage::
+
+        cron = CronSchedule("0 19 * * mon,thu")
+        run_at = cron.next_after(datetime.now())
+    """
+
+    def __init__(self, expression: str):
+        self.expression = expression.strip()
+        fields = self.expression.split()
+        if len(fields) != 5:
+            raise ValueError(
+                f"cron expression needs 5 fields "
+                f"(minute hour dom month dow), got {len(fields)}: "
+                f"{expression!r}")
+
+        minute, hour, dom, month, dow = fields
+        self._minutes = _parse_field(minute, 0, 59)
+        self._hours   = _parse_field(hour,   0, 23)
+        self._dom     = _parse_field(dom,    1, 31)
+        self._months  = _parse_field(month,  1, 12, _MONTH_NAMES)
+        dow_values    = _parse_field(dow,    0, 7,  _DAY_NAMES)
+        # 7 is an alias for Sunday (= 0)
+        self._dow = {0 if v == 7 else v for v in dow_values}
+
+        # Vixie-cron rule: dom/dow are OR-combined when both are restricted.
+        self._dom_restricted = dom != "*"
+        self._dow_restricted = dow != "*"
+
+    def __repr__(self) -> str:
+        return f"CronSchedule({self.expression!r})"
+
+    # ------------------------------------------------------------------
+    def _day_matches(self, day: "datetime.date") -> bool:
+        if day.month not in self._months:
+            return False
+        dom_ok = day.day in self._dom
+        # Python: Monday=0 … Sunday=6  ->  cron: Sunday=0 … Saturday=6
+        dow_ok = ((day.weekday() + 1) % 7) in self._dow
+        if self._dom_restricted and self._dow_restricted:
+            return dom_ok or dow_ok
+        if self._dom_restricted:
+            return dom_ok
+        if self._dow_restricted:
+            return dow_ok
+        return True
+
+    def next_after(self, dt: datetime) -> datetime:
+        """
+        Returns the first matching time strictly after ``dt``
+        (second/microsecond precision is dropped).
+        """
+        cand = (dt + timedelta(minutes=1)).replace(second=0, microsecond=0)
+        hours   = sorted(self._hours)
+        minutes = sorted(self._minutes)
+
+        # Walk day by day (covers rare dom/month combos like Feb 29).
+        for _ in range(366 * 5):
+            if self._day_matches(cand.date()):
+                for h in hours:
+                    if h < cand.hour:
+                        continue
+                    for m in minutes:
+                        if h == cand.hour and m < cand.minute:
+                            continue
+                        return cand.replace(hour=h, minute=m)
+            cand = (cand + timedelta(days=1)).replace(hour=0, minute=0)
+
+        raise ValueError(
+            f"cron {self.expression!r}: no occurrence within 5 years")

src/KavitaClient.py

@@ -0,0 +1,272 @@
+"""
+kavita_client.py
+================
+
+Thin HTTP client for the Kavita server REST API (v0.9.x).
+
+Authenticates via the ``x-api-key`` header.  All series / library /
+collection / metadata reads and writes used by the light-novel updater
+go through this single client so request shaping (paging, content types,
+timeouts, retries) is consistent.
+
+The class is intentionally state-light: no caching layer, just one
+``requests.Session``.  Higher-level diff / update logic lives in
+KavitaSeriesUpdater, KavitaPersonUpdater and RelationshipSync.
+"""
+
+from __future__ import annotations
+
+import base64
+from typing import Iterable
+
+import requests
+
+
+class KavitaClient:
+    def __init__(self, base_url: str, api_key: str, *,
+                 request_timeout: int = 30):
+        self._base = base_url.rstrip("/")
+        self._timeout = request_timeout
+
+        # API session: sends + receives JSON.
+        self._session = requests.Session()
+        self._session.headers.update({
+            "x-api-key":    api_key,
+            "Accept":       "application/json",
+            "Content-Type": "application/json",
+        })
+
+        # Plain session for downloading external images (covers).  Must NOT
+        # carry the API headers — some CDNs refuse to return image bytes
+        # when the client sends Accept: application/json.
+        self._image_session = requests.Session()
+        self._image_session.headers.update({
+            "User-Agent": "KavitaLightNovelUpdater/1.0",
+        })
+
+    # ------------------------------------------------------------------
+    # Libraries
+    # ------------------------------------------------------------------
+    def list_libraries(self) -> list[dict]:
+        """Returns all libraries the authenticated user can access."""
+        r = self._session.get(f"{self._base}/api/Library/libraries",
+                              timeout=self._timeout)
+        r.raise_for_status()
+        return r.json() or []
+
+    # ------------------------------------------------------------------
+    # Series
+    # ------------------------------------------------------------------
+    def list_series_in_library(self, library_id: int, *,
+                               page_size: int = 200) -> list[dict]:
+        """
+        Returns all SeriesDto entries in the given library.
+
+        Uses POST /api/Series/all-v2 with a FilterV2 that scopes by
+        library id.  Pages through until an empty page is returned.
+        """
+        results: list[dict] = []
+        page = 1
+        while True:
+            body = {
+                "statements": [
+                    {
+                        "comparison": 0,     # Equal
+                        "field": 19,         # Libraries field id (Kavita v0.9.x)
+                        "value": str(library_id),
+                    }
+                ],
+                "combination": 1,           # And
+                "sortOptions": {"isAscending": True, "sortField": 1},
+                "limitTo": 0,
+            }
+            r = self._session.post(
+                f"{self._base}/api/Series/all-v2",
+                params={"PageNumber": page, "PageSize": page_size},
+                json=body, timeout=self._timeout)
+            r.raise_for_status()
+            chunk = r.json() or []
+            if not chunk:
+                break
+            results.extend(chunk)
+            if len(chunk) < page_size:
+                break
+            page += 1
+        return results
+
+    def get_series(self, series_id: int) -> dict:
+        """Returns the SeriesDto for the given series id."""
+        r = self._session.get(f"{self._base}/api/Series/{series_id}",
+                              timeout=self._timeout)
+        r.raise_for_status()
+        return r.json() or {}
+
+    def update_series(self, series: dict) -> None:
+        """Updates the Series-level data (name, sortName, malId, …)."""
+        r = self._session.post(f"{self._base}/api/Series/update",
+                               json=series, timeout=self._timeout)
+        r.raise_for_status()
+
+    # ------------------------------------------------------------------
+    # Series metadata
+    # ------------------------------------------------------------------
+    def get_series_metadata(self, series_id: int) -> dict:
+        """Returns the SeriesMetadataDto for a series."""
+        r = self._session.get(
+            f"{self._base}/api/Series/metadata",
+            params={"seriesId": series_id}, timeout=self._timeout)
+        r.raise_for_status()
+        return r.json() or {}
+
+    def update_series_metadata(self, metadata: dict) -> None:
+        """
+        Writes a SeriesMetadataDto back to Kavita.
+
+        Kavita expects the payload wrapped: {seriesMetadata: {...}}.
+        """
+        r = self._session.post(
+            f"{self._base}/api/Series/metadata",
+            json={"seriesMetadata": metadata},
+            timeout=self._timeout)
+        r.raise_for_status()
+
+    # ------------------------------------------------------------------
+    # Related series
+    # ------------------------------------------------------------------
+    def get_related(self, series_id: int) -> dict:
+        """Returns all related series grouped by relation type."""
+        r = self._session.get(
+            f"{self._base}/api/Series/all-related",
+            params={"seriesId": series_id}, timeout=self._timeout)
+        r.raise_for_status()
+        return r.json() or {}
+
+    def update_related(self, payload: dict) -> None:
+        """
+        Sets the related-series relationships for a series.
+
+        Payload shape (UpdateRelatedSeriesDto):
+            {seriesId, prequels, sequels, sideStories, spinOffs,
+             adaptations, characters, contains, others,
+             alternativeSettings, alternativeVersions, doujinshis,
+             editions, annuals}
+        Each *_ids list contains target series ids (ints).
+        """
+        r = self._session.post(
+            f"{self._base}/api/Series/update-related",
+            json=payload, timeout=self._timeout)
+        r.raise_for_status()
+
+    # ------------------------------------------------------------------
+    # Collections
+    # ------------------------------------------------------------------
+    def list_collections(self) -> list[dict]:
+        """Returns all collection tags visible to the authenticated user."""
+        r = self._session.get(
+            f"{self._base}/api/Collection",
+            params={"ownedOnly": "false", "sortByLastModified": "false"},
+            timeout=self._timeout)
+        r.raise_for_status()
+        return r.json() or []
+
+    def add_series_to_collection(self, *, collection_id: int,
+                                 title: str,
+                                 series_ids: Iterable[int]) -> dict:
+        """
+        Adds (or creates) a collection and attaches series to it.
+
+        Pass collection_id=0 to create a new collection named `title`.
+        For an existing collection set collection_id to its id (title is
+        still required by the API but acts as no-op when the id matches).
+        """
+        body = {
+            "collectionTagId":    int(collection_id),
+            "collectionTagTitle": title,
+            "seriesIds":          [int(s) for s in series_ids],
+        }
+        r = self._session.post(
+            f"{self._base}/api/Collection/update-for-series",
+            json=body, timeout=self._timeout)
+        r.raise_for_status()
+        try:
+            return r.json() or {}
+        except ValueError:
+            return {}
+
+    # ------------------------------------------------------------------
+    # Persons
+    # ------------------------------------------------------------------
+    def search_persons(self, name: str) -> list[dict]:
+        """Returns PersonDto entries matching `name` (Kavita's own search)."""
+        r = self._session.get(
+            f"{self._base}/api/Person/search",
+            params={"queryString": name}, timeout=self._timeout)
+        r.raise_for_status()
+        return r.json() or []
+
+    def update_person(self, payload: dict) -> None:
+        """Writes a person record (malId, aniListId, description, …)."""
+        r = self._session.post(f"{self._base}/api/Person/update",
+                               json=payload, timeout=self._timeout)
+        r.raise_for_status()
+
+    # ------------------------------------------------------------------
+    # Cover uploads
+    # ------------------------------------------------------------------
+    def upload_series_cover(self, series_id: int, image_url: str, *,
+                            lock: bool = False) -> None:
+        """Downloads an external image and uploads it as the series cover."""
+        self._upload_cover("/api/Upload/series", series_id, image_url, lock)
+
+    def upload_person_cover(self, person_id: int, image_url: str, *,
+                            lock: bool = False) -> None:
+        """Downloads an external image and uploads it as a person cover."""
+        self._upload_cover("/api/Upload/person", person_id, image_url, lock)
+
+    def _upload_cover(self, endpoint: str, entity_id: int,
+                      image_url: str, lock: bool) -> None:
+        """
+        Shared cover-upload path.  Kavita's upload endpoints accept a raw
+        base64 blob (no ``data:`` prefix) in the ``url`` field — a data
+        URI or the two-step upload-by-url flow are rejected with HTTP 400
+        (verified against Kavita 0.9.0.2).
+        """
+        img = self._image_session.get(image_url, timeout=self._timeout)
+        img.raise_for_status()
+        b64 = base64.b64encode(img.content).decode()
+        r = self._session.post(
+            f"{self._base}{endpoint}",
+            json={"id": entity_id, "url": b64, "lockCover": lock},
+            timeout=self._timeout)
+        if r.status_code >= 400:
+            # Include the body excerpt — Kavita's upload errors carry the
+            # actual reason there, not in the status line.
+            raise requests.HTTPError(
+                f"{endpoint} HTTP {r.status_code}: {_short_body(r)}",
+                response=r)
+
+    # ------------------------------------------------------------------
+    # Generic GET helper (used by callers that need a response object)
+    # ------------------------------------------------------------------
+    def get(self, path: str, params: "dict | None" = None) -> requests.Response:
+        return self._session.get(f"{self._base}{path}",
+                                 params=params, timeout=self._timeout)
+
+    def post(self, path: str, *,
+             json: "dict | list | None" = None,
+             params: "dict | None" = None) -> requests.Response:
+        return self._session.post(f"{self._base}{path}",
+                                  json=json, params=params,
+                                  timeout=self._timeout)
+
+
+def _short_body(resp: requests.Response, limit: int = 400) -> str:
+    """Returns the response body trimmed to `limit` chars for error messages."""
+    try:
+        text = resp.text or ""
+    except Exception:
+        return "<unreadable response body>"
+    text = text.strip().replace("\n", " ").replace("\r", " ")
+    if len(text) > limit:
+        text = text[:limit] + "…"
+    return text or "<empty body>"

src/KavitaPersonUpdater.py

@@ -15,46 +15,22 @@ the updater:
      an 'about' text (requires an extra Jikan request per character; only
      performed when update_descriptions=True).
 
-Kavita API version
-------------------
+All HTTP traffic to Kavita goes through the shared :class:`KavitaClient`
+(`/api/Person/search`, `/api/Person/update`, `/api/Upload/person`).
+
 Tested against Kavita 0.9.0.2.
-
-Authentication
---------------
-Uses the `x-api-key` header (API key from Kavita user settings).
-No JWT login is required.
-
-Relevant endpoints (Kavita 0.9.0.2)
--------------------------------------
-  GET  /api/Person/search       find persons by name / alias
-  POST /api/Person/update       write metadata (malId, description, …)
-  POST /api/Upload/person       set cover image (base64 data URI)
-  POST /api/Upload/upload-by-url  download an external URL to temp storage
-                                  (used as an alternative upload path)
-
-Cover upload flow
------------------
-The image is downloaded locally, base64-encoded, and sent as a data URI
-to POST /api/Upload/person.  This is more reliable than the
-upload-by-url → upload/person two-step because it avoids Kavita's temp
-file handling (which had known issues in 0.8.x – 0.9.x, GitHub #3900).
-
-Dependencies
-------------
-    requests    ->  pip install requests
 """
 
 from __future__ import annotations
 
-import base64
 import datetime
-import difflib
-import re
 
 import requests
 
+from KavitaClient import KavitaClient
 from MALResolver import MALResolver
 from AniListResolver import AniListResolver
+from TextUtils import best_similarity, paragraphs_to_html, person_name_with_id
 
 
 class KavitaPersonUpdater:
@@ -63,41 +39,22 @@ class KavitaPersonUpdater:
 
     Parameters
     ----------
-    kavita_base_url : Base URL of the Kavita server, e.g. "http://192.168.2.2:5000"
-    api_key         : Kavita API key (Settings → User → API key)
+    client          : Shared KavitaClient (session, auth, cover uploads)
     mal_resolver    : Shared MALResolver singleton (created automatically if omitted)
-    request_timeout : HTTP timeout in seconds for both Kavita and image requests
+    al_resolver     : Shared AniListResolver singleton (created automatically if omitted)
     min_name_score  : Minimum difflib similarity ratio (0–1) required to accept a
                       Kavita person as a match for a MAL name.  Default 0.80.
     """
 
-    def __init__(self, kavita_base_url: str, api_key: str, *,
+    def __init__(self, client: KavitaClient, *,
                  mal_resolver: "MALResolver | None" = None,
                  al_resolver: "AniListResolver | None" = None,
-                 request_timeout: int = 30,
                  min_name_score: float = 0.80):
-        self._base = kavita_base_url.rstrip("/")
-        self._timeout = request_timeout
+        self._client = client
         self._min_score = min_name_score
         self._mal = mal_resolver or MALResolver()
         self._al  = al_resolver  or AniListResolver()
 
-        # Session used for Kavita API calls.
-        self._session = requests.Session()
-        self._session.headers.update({
-            "x-api-key": api_key,
-            "Content-Type": "application/json",
-            "Accept": "application/json",
-        })
-
-        # Plain session used to download external images (MAL CDN etc.).
-        # Must NOT carry the Kavita API headers — Accept: application/json
-        # would prevent MAL CDN from returning the image bytes.
-        self._image_session = requests.Session()
-        self._image_session.headers.update({
-            "User-Agent": "KavitaPersonUpdater/1.0",
-        })
-
         # Cache: normalised name -> list of PersonDto dicts (best matches first)
         self._person_search_cache: dict[str, list[dict]] = {}
 
@@ -195,11 +152,28 @@ class KavitaPersonUpdater:
             if not name and not raw_name:
                 continue
 
-            # Search by the cleaned (XML-safe) name first; if Kavita stores
+            if kind == "character":
+                # Characters are stored under their disambiguated name
+                # ("Rem (MAL 118737)") — see person_name_with_id.  The
+                # series metadata write creates the person under exactly
+                # this name, so only that form is searched.
+                search_names = [person_name_with_id(
+                    name, mal_id=entry.get("mal_id"),
+                    al_id=entry.get("al_id"))]
+            else:
+                # Staff: cleaned (XML-safe) name first; if Kavita stores
                 # the legacy comma form, retry with the raw MAL name.
-            matches = self._find_kavita_person(name) if name else []
-            if not matches and raw_name and raw_name != name:
-                matches = self._find_kavita_person(raw_name)
+                search_names = [name]
+                if raw_name and raw_name != name:
+                    search_names.append(raw_name)
+
+            matches: list[dict] = []
+            for search_name in search_names:
+                if not search_name:
+                    continue
+                matches = self._find_kavita_person(search_name)
+                if matches:
+                    break
 
             if not matches:
                 result["not_found"] += 1
@@ -230,29 +204,17 @@ class KavitaPersonUpdater:
             return self._person_search_cache[key]
 
         try:
-            resp = self._session.get(
-                f"{self._base}/api/Person/search",
-                params={"queryString": name},
-                timeout=self._timeout,
-            )
-            resp.raise_for_status()
-            persons: list[dict] = resp.json() or []
+            persons = self._client.search_persons(name)
         except requests.RequestException:
             self._person_search_cache[key] = []
             return []
 
-        def score(p: dict) -> float:
-            candidates = [p.get("name") or ""]
-            candidates += [a for a in (p.get("aliases") or []) if a]
-            best = 0.0
-            q = key
-            for c in candidates:
-                r = difflib.SequenceMatcher(None, q, c.lower()).ratio()
-                best = max(best, r)
-            return best
-
-        ranked = sorted(persons, key=score, reverse=True)
-        filtered = [p for p in ranked if score(p) >= self._min_score]
+        scored = []
+        for p in persons:
+            candidates = [p.get("name")] + list(p.get("aliases") or [])
+            scored.append((best_similarity(key, candidates), p))
+        scored.sort(key=lambda pair: pair[0], reverse=True)
+        filtered = [p for score, p in scored if score >= self._min_score]
         self._person_search_cache[key] = filtered
         return filtered
 
@@ -289,6 +251,20 @@ class KavitaPersonUpdater:
 
         current_mal_id: int = person.get("malId") or 0
         current_al_id:  int = person.get("aniListId") or 0
+
+        # Collision guard: the Kavita person is already linked to a
+        # *different* tracker entity — same display name, different
+        # character/person.  Never overwrite; first writer wins.
+        if ((mal_id and current_mal_id and current_mal_id != mal_id)
+                or (al_id and current_al_id and current_al_id != al_id)):
+            if errors is not None:
+                errors.append(
+                    f"conflict: '{person_name}' (#{person_id}) is linked to "
+                    f"malId={current_mal_id or '-'}/aniListId={current_al_id or '-'} "
+                    f"but this entry has malId={mal_id or '-'}/aniListId={al_id or '-'} "
+                    f"— skipped")
+            return False
+
         needs_mal_id = bool(mal_id and current_mal_id != mal_id)
         needs_al_id  = bool(al_id  and current_al_id  != al_id)
 
@@ -323,12 +299,7 @@ class KavitaPersonUpdater:
                 "aniListId": al_id if needs_al_id  else (current_al_id  or None),
             }
             try:
-                resp = self._session.post(
-                    f"{self._base}/api/Person/update",
-                    json=payload,
-                    timeout=self._timeout,
-                )
-                resp.raise_for_status()
+                self._client.update_person(payload)
                 changed = True
             except requests.RequestException as e:
                 if errors is not None:
@@ -350,88 +321,21 @@ class KavitaPersonUpdater:
                 and bool(person.get("coverImage"))
             )
             if image_url and not already_uploaded:
-                if self._upload_cover(person_id, image_url,
-                                      person_name=person_name,
-                                      errors=errors):
+                try:
+                    self._client.upload_person_cover(person_id, image_url)
                     changed = True
+                except requests.RequestException as e:
+                    if errors is not None:
+                        errors.append(
+                            f"cover upload failed for #{person_id} "
+                            f"'{person_name}' ({image_url}): {e}")
 
         return changed
 
-    # ------------------------------------------------------------------
-    # Internal: cover upload
-    # ------------------------------------------------------------------
-    def _upload_cover(self, person_id: int, image_url: str,
-                      lock: bool = False, *,
-                      person_name: str = "",
-                      errors: "list | None" = None) -> bool:
-        """
-        Uploads a cover image to a Kavita person.
-
-        The image is downloaded with the plain (header-less) image session
-        and posted to `POST /api/Upload/person` as a raw base64 string in
-        the `url` field.
-
-        Notes on protocol quirks discovered against Kavita 0.9.0.2:
-          - The two-step `upload-by-url` -> `Upload/person` flow returns
-            "Unable to save cover image to Person" (HTTP 400).
-          - A `data:image/jpeg;base64,...` data URI is rejected with the
-            same error.
-          - Only the raw base64 blob (no prefix) is accepted.
-        """
-        label = (f"#{person_id} '{person_name}'"
-                 if person_name else f"#{person_id}")
-
-        # 1) Download the image with a clean session — the Kavita session's
-        #    `Accept: application/json` header makes some CDNs refuse to
-        #    return image bytes.
-        try:
-            img_resp = self._image_session.get(image_url,
-                                               timeout=self._timeout)
-            img_resp.raise_for_status()
-        except requests.RequestException as e:
-            if errors is not None:
-                errors.append(
-                    f"image download failed for {label} ({image_url}): {e}")
-            return False
-
-        b64 = base64.b64encode(img_resp.content).decode()
-
-        # 2) POST the raw base64 blob.
-        try:
-            resp = self._session.post(
-                f"{self._base}/api/Upload/person",
-                json={"id": person_id, "url": b64, "lockCover": lock},
-                timeout=self._timeout,
-            )
-            if resp.status_code >= 400:
-                if errors is not None:
-                    errors.append(
-                        f"Upload/person HTTP {resp.status_code} for {label}: "
-                        f"{_short_body(resp)}")
-                return False
-            return True
-        except requests.RequestException as e:
-            if errors is not None:
-                errors.append(
-                    f"Upload/person failed for {label}: {e}")
-            return False
-
 
 # --------------------------------------------------------------------------
 # Module helpers: description builders
 # --------------------------------------------------------------------------
-def _plain_to_html(text: str) -> str:
-    """Converts plain text with paragraph breaks to compact HTML (no raw \\n)."""
-    if not text:
-        return ""
-    parts: list[str] = []
-    for para in re.split(r"\n{2,}", text.strip()):
-        para = para.strip()
-        if para:
-            parts.append(f"<p>{para.replace(chr(10), '<br>')}</p>")
-    return "".join(parts)
-
-
 def _format_birthday(birthday: str) -> str:
     """Converts an ISO 8601 birthday string to "D Month YYYY"."""
     if not birthday:
@@ -457,7 +361,7 @@ def _build_character_description(details: dict) -> str:
         parts.append(f'<p><a href="{url}" target="_blank">Favorites: {favorites:,}</a></p>')
     about = (details.get("about") or "").strip()
     if about:
-        parts.append(_plain_to_html(about))
+        parts.append(paragraphs_to_html(about))
     return "<br>".join(parts)
 
 
@@ -501,33 +405,19 @@ def _build_person_description(details: dict) -> str:
         parts.append(f'<table>{"".join(rows)}</table>')
     about = (details.get("about") or "").strip()
     if about:
-        parts.append(_plain_to_html(about))
+        parts.append(paragraphs_to_html(about))
     return "<br>".join(parts)
 
 
-# --------------------------------------------------------------------------
-# Module helper
-# --------------------------------------------------------------------------
-def _short_body(resp: requests.Response, limit: int = 400) -> str:
-    """Returns the response body trimmed to `limit` chars for error logging."""
-    try:
-        text = resp.text or ""
-    except Exception:
-        return "<unreadable response body>"
-    text = text.strip().replace("\n", " ").replace("\r", " ")
-    if len(text) > limit:
-        text = text[:limit] + "…"
-    return text or "<empty body>"
-
-
 # --------------------------------------------------------------------------
 # Usage example
 # --------------------------------------------------------------------------
 if __name__ == "__main__":
-    KAVITA_URL = "http://192.168.2.2:5000"
-    KAVITA_KEY = "Sq4a3hcV171dn3gzCl0K4eN7hZNk4sOA"
+    import os
 
-    updater = KavitaPersonUpdater(KAVITA_URL, KAVITA_KEY)
+    client = KavitaClient(os.environ["KAVITA_URL"],
+                          os.environ["KAVITA_API_KEY"])
+    updater = KavitaPersonUpdater(client)
 
     mal = MALResolver()
     mal_id = mal.find_mal_id("よふかしのうた")

src/MALResolver.py

@@ -30,12 +30,12 @@ Dependencies
 from __future__ import annotations
 
 import datetime
-import difflib
 import time
 
 import requests
 
 from MediaResolver import MediaResolver
+from TextUtils import best_similarity
 
 
 class MALResolver(MediaResolver):
@@ -57,12 +57,21 @@ class MALResolver(MediaResolver):
             cls._instance._initialized = False
         return cls._instance
 
-    def __init__(self, *, request_timeout: int = 30):
+    def __init__(self, *, request_timeout: int = 30,
+                 search_type: str = "manga"):
+        """
+        search_type : Jikan `type` filter for title searches — "manga" for
+                      the manga container, "lightnovel" for the LN container.
+                      Only the FIRST construction in the process sets it
+                      (singleton); construct the resolver with the correct
+                      type in the entry point / orchestrator.
+        """
         if self._initialized:
             return
 
         self.JIKAN_BASE = "https://api.jikan.moe/v4"
         self.request_timeout = request_timeout
+        self.search_type = search_type
 
         self._session = requests.Session()
         self._session.headers.setdefault("User-Agent", "MALResolver/1.0")
@@ -106,7 +115,7 @@ class MALResolver(MediaResolver):
 
         try:
             data = self._get(f"{self.JIKAN_BASE}/manga",
-                             {"q": title, "limit": 5, "type": "manga"})
+                             {"q": title, "limit": 5, "type": self.search_type})
             results = data.get("data") or []
         except requests.RequestException:
             return None
@@ -404,19 +413,12 @@ def _clean_mal_name(name: str) -> str:
 def _score_title(query: str, entry: dict) -> float:
     """Returns the best title-similarity score for a Jikan manga entry."""
     candidates = [
-        entry.get("title") or "",
-        entry.get("title_english") or "",
-        entry.get("title_japanese") or "",
+        entry.get("title"),
+        entry.get("title_english"),
+        entry.get("title_japanese"),
     ]
-    for alt in (entry.get("titles") or []):
-        candidates.append(alt.get("title") or "")
-    best = 0.0
-    q = query.lower()
-    for t in candidates:
-        if t:
-            ratio = difflib.SequenceMatcher(None, q, t.lower()).ratio()
-            best = max(best, ratio)
-    return best
+    candidates += [alt.get("title") for alt in (entry.get("titles") or [])]
+    return best_similarity(query, candidates)
 
 
 # --------------------------------------------------------------------------

src/MangaBakaRateLimit.py

@@ -0,0 +1,92 @@
+"""
+mangabaka_rate_limit.py
+=======================
+
+Process-wide rate limiter for the MangaBaka API.
+
+Apply via:
+
+    from MangaBakaRateLimit import apply_to_session
+    apply_to_session(session)
+
+This mounts a custom ``requests.adapters.HTTPAdapter`` on the given
+``requests.Session`` for the ``api.mangabaka.dev`` host.  Every request
+going through that adapter is:
+
+  * throttled so that no two requests are dispatched within
+    ``_MIN_INTERVAL`` seconds of one another, and
+  * retried on HTTP 429, honouring the ``Retry-After`` header when
+    present, otherwise exponential backoff capped at ``_MAX_BACKOFF``.
+
+Throttle state is module-global, so even if several sessions exist in
+the same process they share one budget — important because they all hit
+the same upstream IP-based limit.
+"""
+
+from __future__ import annotations
+
+import threading
+import time
+
+from requests.adapters import HTTPAdapter
+
+
+# Tune these if MangaBaka tightens or loosens limits.
+_MIN_INTERVAL = 1.1      # seconds between consecutive requests
+_MAX_RETRIES  = 6        # retries on 429 before giving up
+_MAX_BACKOFF  = 60.0     # cap on per-attempt backoff sleep
+
+
+# --- shared throttle state --------------------------------------------------
+_state_lock        = threading.Lock()
+_last_request_time = 0.0
+
+
+def _wait_for_slot() -> None:
+    """Block until the next request slot is available, then reserve it."""
+    global _last_request_time
+    while True:
+        with _state_lock:
+            now  = time.monotonic()
+            wait = _MIN_INTERVAL - (now - _last_request_time)
+            if wait <= 0:
+                _last_request_time = now
+                return
+        time.sleep(wait)
+
+
+class _MangaBakaRateLimitAdapter(HTTPAdapter):
+    def send(self, request, **kwargs):
+        response = None
+        for attempt in range(_MAX_RETRIES + 1):
+            _wait_for_slot()
+            response = super().send(request, **kwargs)
+            if response.status_code != 429:
+                return response
+
+            retry_after = response.headers.get("Retry-After")
+            try:
+                wait = (float(retry_after) if retry_after
+                        else min(_MAX_BACKOFF, 2.0 * (2 ** attempt)))
+            except ValueError:
+                wait = min(_MAX_BACKOFF, 2.0 * (2 ** attempt))
+
+            print(f"[MangaBaka] 429 — backing off {wait:.1f}s "
+                  f"(attempt {attempt + 1}/{_MAX_RETRIES})",
+                  flush=True)
+            response.close()
+            time.sleep(wait)
+
+        # Retries exhausted — let the caller deal with the last 429.
+        return response
+
+
+def apply_to_session(session) -> None:
+    """
+    Mount the rate-limit adapter on ``session`` so every MangaBaka call
+    is automatically throttled. Safe to call multiple times (later mounts
+    just replace the earlier adapter for the same prefix).
+    """
+    adapter = _MangaBakaRateLimitAdapter()
+    session.mount("https://api.mangabaka.dev/", adapter)
+    session.mount("http://api.mangabaka.dev/",  adapter)

src/MangaBakaWorksResolver.py

@@ -2,7 +2,7 @@
 mangabaka_works_resolver.py
 ===========================
 
-Fetches volume-level (work) data from the MangaBaka API.
+Fetches volume-level (work) data and volume cover images from the MangaBaka API.
 
 Each "work" is a physical tankobon volume and may carry:
   - volume number
@@ -11,10 +11,16 @@ Each "work" is a physical tankobon volume and may carry:
   - release date
   - cover image  (raw / default / small variants)
 
-Only works that have a usable cover are kept in the cache.
-Works without a cover are discarded at fetch time.
-If no volume is assigned for a chapter, callers fall back to the
-default series cover from the series object itself.
+Cover resolution order (per volume)
+------------------------------------
+1. GET /v1/series/{id}/images  — covers that exist independently of a work
+   (some series have covers but no works).  English edition preferred;
+   original language used when no English cover is available.
+2. GET /v1/series/{id}/works   — physical tankobon data including covers.
+   Fallback when /images returns nothing for the requested volume.
+
+If no volume cover is found at all, callers fall back to the series-level
+default cover from the series object itself.
 
 Dependencies
 ------------
@@ -26,10 +32,75 @@ from __future__ import annotations
 import requests
 
 
+# --------------------------------------------------------------------------
+# Generic image-block URL picker (shared by /images and /works responses)
+# --------------------------------------------------------------------------
+def _pick_image_url(image) -> "str | None":
+    """
+    Returns the best URL from a MangaBaka image block.
+
+    Handles the common ``{raw, x150, x250, x350}`` structure used by both
+    the ``cover`` field on series/work objects and the ``image`` field on
+    ``/images`` endpoint items::
+
+        {
+          "raw":  {"url": "...", "size": ..., "height": ..., "width": ...},
+          "x150": {"x1": "...", "x2": "...", "x3": "..."},
+          "x250": {...},
+          "x350": {...}
+        }
+
+    Preference: raw original > x350@x3 > x250@x3 > x150@x3 > … (falling
+    through to lower densities and sizes as needed).
+    """
+    if not image:
+        return None
+    if isinstance(image, str):
+        return image
+    if not isinstance(image, dict):
+        return None
+
+    # 1) Raw / unscaled image
+    raw = image.get("raw")
+    if isinstance(raw, dict):
+        url = raw.get("url")
+        if isinstance(url, str) and url:
+            return url
+    elif isinstance(raw, str) and raw:
+        return raw
+
+    # 2) Size-keyed CDN variants, largest first, highest density first
+    for size_key in ("x350", "x250", "x150"):
+        variant = image.get(size_key)
+        if isinstance(variant, dict):
+            for density in ("x3", "x2", "x1"):
+                url = variant.get(density)
+                if isinstance(url, str) and url:
+                    return url
+        elif isinstance(variant, str) and variant:
+            return variant
+
+    # 3) Last-ditch: any HTTP URL anywhere in the structure
+    for val in image.values():
+        if isinstance(val, str) and val.startswith("http"):
+            return val
+        if isinstance(val, dict):
+            for sub_val in val.values():
+                if isinstance(sub_val, str) and sub_val.startswith("http"):
+                    return sub_val
+    return None
+
+
 class MangaBakaWorksResolver:
     """
-    Fetches and caches MangaBaka volume (work) data for a series.
-    Only works that have a cover image are retained in the cache.
+    Fetches and caches MangaBaka volume (work) data and cover images.
+
+    Cover lookup order per volume
+    ------------------------------
+    1. ``/v1/series/{id}/images`` — edition covers (English > original).
+    2. ``/v1/series/{id}/works``  — physical tankobon covers.
+
+    Only works that carry a cover image are retained in the works cache.
     """
 
     def __init__(self, api_base_url: str = "https://api.mangabaka.dev/v1",
@@ -42,30 +113,24 @@ class MangaBakaWorksResolver:
 
         # Cache: series_id (str) -> list of work dicts (only those with covers)
         self._cache: dict[str, list[dict]] = {}
+        # Cache: series_id (str) -> {norm_vol (str): url (str)}
+        self._images_cache: dict[str, dict[str, str]] = {}
 
     # ------------------------------------------------------------------
     # Public API
     # ------------------------------------------------------------------
-    def get_works(self, series_id: str) -> list[dict]:
+    def _fetch_all_pages(self, endpoint: str) -> list[dict]:
         """
-        Returns volume-level works for a series, filtered to those that have
-        a usable cover image.  Results are cached per series.
-
-        Pages through the API (limit=50) until the response returns an empty
-        page, collecting all works before applying the cover filter.
+        Pages through a MangaBaka list endpoint (limit=50 per page) and
+        returns all collected `data` items.  Network errors end the
+        pagination early; items fetched so far are returned.
         """
-        if not series_id:
-            return []
-
-        if series_id in self._cache:
-            return self._cache[series_id]
-
-        all_works: list[dict] = []
+        items: list[dict] = []
         page = 1
         try:
             while True:
                 resp = self._session.get(
-                    f"{self.api_base_url}/series/{series_id}/works",
+                    f"{self.api_base_url}/series/{endpoint}",
                     params={"limit": 50, "page": page},
                     timeout=self.request_timeout,
                 )
@@ -73,16 +138,34 @@ class MangaBakaWorksResolver:
                 page_data = resp.json().get("data") or []
                 if not page_data:
                     break
-                all_works.extend(page_data)
+                items.extend(page_data)
                 if len(page_data) < 50:
                     break
                 page += 1
         except requests.RequestException:
-            if not all_works:
+            pass
+        return items
+
+    def get_works(self, series_id: str) -> list[dict]:
+        """
+        Returns volume-level works for a series, filtered to those that have
+        a usable cover image.
+
+        Non-empty results are cached per series; empty results are not, so
+        works added on MangaBaka later become visible without restarting
+        the (long-running) process.
+        """
+        if not series_id:
             return []
 
+        if series_id in self._cache:
+            return self._cache[series_id]
+
+        all_works = self._fetch_all_pages(f"{series_id}/works")
+
         # Discard works that carry no usable cover
         works_with_cover = [w for w in all_works if w.get("images")]
+        if works_with_cover:
             self._cache[series_id] = works_with_cover
         return works_with_cover
 
@@ -101,12 +184,85 @@ class MangaBakaWorksResolver:
                 return work
         return None
 
-    def get_cover_for_volume(self, series_id: str, volume) -> "str | None":
-        """Returns the cover URL for a specific volume, or None if not found."""
-        work = self.get_work_for_volume(series_id, volume)
-        if not work:
+    def get_volume_covers(self, series_id: str) -> "dict[str, str]":
+        """
+        Fetches all volume-type cover images for a series from
+        ``/v1/series/{id}/images`` and returns a
+        ``{normalised_volume_str: url}`` mapping.
+
+        English-edition covers are preferred; the first available language
+        is used as fallback when no English cover exists for a volume.
+        Results are cached per series.
+        """
+        if not series_id:
+            return {}
+
+        if series_id in self._images_cache:
+            return self._images_cache[series_id]
+
+        raw_items = self._fetch_all_pages(f"{series_id}/images")
+
+        # Group by normalised volume index; collect all languages per volume.
+        by_volume: dict[str, dict[str, str]] = {}  # norm_vol -> {lang: url}
+        for item in raw_items:
+            if item.get("type") != "volume":
+                continue
+            idx = item.get("index_numeric")
+            if idx is None:
+                continue
+            norm = _norm_vol(idx)
+            lang = (item.get("language") or "").lower() or "unknown"
+            url  = _pick_image_url(item.get("image"))
+            if not url:
+                continue
+            if norm not in by_volume:
+                by_volume[norm] = {}
+            # First entry per language wins (API order reflects quality/rank).
+            if lang not in by_volume[norm]:
+                by_volume[norm][lang] = url
+
+        # Pick best language per volume: English first, then first available.
+        result: dict[str, str] = {}
+        for norm, lang_map in by_volume.items():
+            url = lang_map.get("en") or next(iter(lang_map.values()), None)
+            if url:
+                result[norm] = url
+
+        # Empty results are not cached — covers added on MangaBaka later
+        # become visible without restarting the long-running process.
+        if result:
+            self._images_cache[series_id] = result
+        return result
+
+    def get_cover_for_volume_from_images(self, series_id: str,
+                                         volume) -> "str | None":
+        """
+        Returns the cover URL for a specific volume from the /images endpoint,
+        or None if not available.
+        """
+        covers = self.get_volume_covers(series_id)
+        if not covers:
             return None
-        return self._pick_cover_url(work.get("images")[0].get("image"))
+        return covers.get(_norm_vol(volume))
+
+    def get_cover_for_volume(self, series_id: str, volume) -> "str | None":
+        """
+        Returns the best cover URL for a specific volume.
+
+        Tries the ``/images`` endpoint first (covers that exist even when no
+        physical work has been catalogued), then falls back to the ``/works``
+        endpoint.  Returns None if neither source has a cover for the volume.
+        """
+        # 1. /images endpoint (covers without works)
+        url = self.get_cover_for_volume_from_images(series_id, volume)
+        if url:
+            return url
+
+        # 2. /works endpoint fallback
+        work = self.get_work_for_volume(series_id, volume)
+        if not work or not work.get("images"):
+            return None
+        return _pick_image_url(work["images"][0].get("image"))
 
     def get_page_counts(self, series_id: str) -> "dict[str, int]":
         """
@@ -125,59 +281,9 @@ class MangaBakaWorksResolver:
         return result
 
     def clear_cache(self) -> None:
-        """Clears the internal works cache."""
+        """Clears both the works cache and the images cover cache."""
         self._cache.clear()
-
-    # ------------------------------------------------------------------
-    # Helpers
-    # ------------------------------------------------------------------
-    @staticmethod
-    def _pick_cover_url(cover) -> "str | None":
-        """
-        Selects the best cover URL from a MangaBaka cover object.
-
-        Real API shape:
-            "raw":  {"url": "...", "size": ..., "height": ..., "width": ...}
-            "x150": {"x1": "...", "x2": "...", "x3": "..."}
-            "x250": {...}
-            "x350": {...}
-
-        Order: raw original > x350@x3 > x250@x3 > x150@x3 ...
-        """
-        if not cover:
-            return None
-        if isinstance(cover, str):
-            return cover
-        if not isinstance(cover, dict):
-            return None
-
-        raw = cover.get("raw")
-        if isinstance(raw, dict):
-            url = raw.get("url")
-            if isinstance(url, str) and url:
-                return url
-        elif isinstance(raw, str) and raw:
-            return raw
-
-        for size_key in ("x350", "x250", "x150"):
-            variant = cover.get(size_key)
-            if isinstance(variant, dict):
-                for density in ("x3", "x2", "x1"):
-                    url = variant.get(density)
-                    if isinstance(url, str) and url:
-                        return url
-            elif isinstance(variant, str) and variant:
-                return variant
-
-        # Last-ditch: any HTTP URL anywhere in the structure
-        for val in cover.values():
-            if isinstance(val, str) and val.startswith("http"):
-                return val
-            if isinstance(val, dict):
-                for sub_val in val.values():
-                    if isinstance(sub_val, str) and sub_val.startswith("http"):
-                        return sub_val
-        return None
+        self._images_cache.clear()
 
 
 # --------------------------------------------------------------------------

src/MatchesWebApp.py

@@ -1,355 +0,0 @@
-"""
-matches_web_app.py
-==================
-
-Flask web UI for inspecting and editing the matches.json file produced by
-MatchesCache.
-
-Routes
-------
-GET  /                       HTML table view (one row per cached match)
-GET  /api/matches            JSON dump of the full cache
-POST /api/matches            Upsert / rename an entry
-                             body: {originalTitle?, title, mangabakaId,
-                                    mangabakaName, imageUrl, firstMatchTime?}
-POST /api/matches/delete     Remove an entry      body: {title}
-POST /api/build              Trigger a full re-scan via SuwayomiMover.build_matches_only
-                             (only available if a mover is wired in)
-
-The Title cell is rendered as a link to MangaBaka's search page, restricted
-to the manga / manhwa / manhua types.
-"""
-
-from __future__ import annotations
-
-import threading
-from urllib.parse import quote_plus
-
-from flask import Flask, jsonify, request, Response
-
-from MatchesCache import MatchesCache
-
-
-_INDEX_HTML = """<!doctype html>
-<html lang="en">
-<head>
-  <meta charset="utf-8">
-  <title>MangaBaka matches</title>
-  <style>
-    body  { font-family: system-ui, sans-serif; margin: 1.5rem; background: #111; color: #eee; }
-    h1    { margin: 0 0 1rem; font-size: 1.4rem; }
-    .bar  { display: flex; gap: .5rem; align-items: center; margin-bottom: 1rem; flex-wrap: wrap; }
-    .bar input[type=search] { padding: .3rem .5rem; min-width: 18rem; background:#222; color:#eee; border:1px solid #444; }
-    button { padding: .35rem .7rem; cursor: pointer; background:#2a2a2a; color:#eee; border:1px solid #555; }
-    button.primary { background:#2563eb; border-color:#2563eb; color:white; }
-    button.danger  { background:#7f1d1d; border-color:#7f1d1d; color:white; }
-    table { border-collapse: collapse; width: 100%; }
-    th, td { border: 1px solid #333; padding: .4rem .6rem; vertical-align: top; }
-    th { background: #1d1d1d; text-align: left; position: sticky; top: 0; }
-    tr:nth-child(even) td { background: #161616; }
-    td.image img { max-width: 90px; max-height: 130px; display:block; }
-    td input { width: 100%; padding: .25rem; background:#222; color:#eee; border:1px solid #444; }
-    td.title a { color: #60a5fa; text-decoration: none; }
-    td.title a:hover { text-decoration: underline; }
-    td.actions { white-space: nowrap; }
-    .status { margin-left: .5rem; color:#9ca3af; font-size: .9rem; }
-    .dirty td { background: #1f2937 !important; }
-  </style>
-</head>
-<body>
-<h1>MangaBaka matches</h1>
-<div class="bar">
-  <input id="filter" type="search" placeholder="Filter by title…">
-  <button id="reload">Reload</button>
-  <button id="build"  class="primary">Build all (rescan)</button>
-  <span class="status" id="status"></span>
-</div>
-
-<table>
-  <thead>
-    <tr>
-      <th>Title</th>
-      <th>mangabakaId</th>
-      <th>mangabakaName</th>
-      <th>firstMatchTime</th>
-      <th>Image</th>
-      <th></th>
-    </tr>
-  </thead>
-  <tbody id="rows"></tbody>
-</table>
-
-<script>
-const TYPES = "&type=manhwa&type=manhua&type=manga";
-
-function fmtTime(unix) {
-  if (!unix) return "";
-  const d = new Date(unix * 1000);
-  return d.toLocaleString();
-}
-
-function searchUrl(title) {
-  return "https://mangabaka.org/search?q=" + encodeURIComponent(title) + TYPES;
-}
-
-function setStatus(msg) { document.getElementById("status").textContent = msg; }
-
-function makeRow(title, e) {
-  const tr = document.createElement("tr");
-  tr.dataset.originalTitle = title;
-
-  const titleTd = document.createElement("td");
-  titleTd.className = "title";
-  const titleLink = document.createElement("a");
-  titleLink.href = searchUrl(title);
-  titleLink.target = "_blank";
-  titleLink.rel = "noopener";
-  titleLink.textContent = title;
-  const titleInput = document.createElement("input");
-  titleInput.value = title;
-  titleInput.style.marginTop = ".25rem";
-  titleInput.addEventListener("input", () => {
-    titleLink.textContent = titleInput.value;
-    titleLink.href = searchUrl(titleInput.value);
-    tr.classList.add("dirty");
-  });
-  titleTd.append(titleLink, titleInput);
-  tr.appendChild(titleTd);
-
-  function field(value) {
-    const td = document.createElement("td");
-    const inp = document.createElement("input");
-    inp.value = value || "";
-    inp.addEventListener("input", () => tr.classList.add("dirty"));
-    td.appendChild(inp);
-    return [td, inp];
-  }
-
-  const [idTd,   idInp]   = field(e.mangabakaId);
-  const [nameTd, nameInp] = field(e.mangabakaName);
-  const [urlTd,  urlInp]  = field(e.imageUrl);
-  tr.appendChild(idTd);
-  tr.appendChild(nameTd);
-
-  const timeTd = document.createElement("td");
-  timeTd.textContent = fmtTime(e.firstMatchTime);
-  tr.appendChild(timeTd);
-
-  const imgTd = document.createElement("td");
-  imgTd.className = "image";
-  const img = document.createElement("img");
-  img.src = e.imageUrl || "";
-  img.alt = "";
-  img.loading = "lazy";
-  urlInp.addEventListener("input", () => { img.src = urlInp.value; });
-  imgTd.append(img, urlInp);
-  tr.appendChild(imgTd);
-
-  const actTd = document.createElement("td");
-  actTd.className = "actions";
-  const save = document.createElement("button");
-  save.textContent = "Save";
-  save.className = "primary";
-  save.addEventListener("click", async () => {
-    save.disabled = true;
-    setStatus("Saving " + titleInput.value + "…");
-    const body = {
-      originalTitle: tr.dataset.originalTitle,
-      title:         titleInput.value,
-      mangabakaId:   idInp.value,
-      mangabakaName: nameInp.value,
-      imageUrl:      urlInp.value,
-    };
-    try {
-      const r = await fetch("/api/matches", {
-        method: "POST",
-        headers: { "Content-Type": "application/json" },
-        body: JSON.stringify(body),
-      });
-      if (!r.ok) throw new Error(await r.text());
-      tr.dataset.originalTitle = titleInput.value;
-      tr.classList.remove("dirty");
-      setStatus("Saved " + titleInput.value);
-    } catch (err) {
-      setStatus("Save failed: " + err.message);
-    } finally {
-      save.disabled = false;
-    }
-  });
-  const del = document.createElement("button");
-  del.textContent = "Delete";
-  del.className = "danger";
-  del.style.marginLeft = ".25rem";
-  del.addEventListener("click", async () => {
-    if (!confirm("Delete " + tr.dataset.originalTitle + "?")) return;
-    setStatus("Deleting " + tr.dataset.originalTitle + "…");
-    try {
-      const r = await fetch("/api/matches/delete", {
-        method: "POST",
-        headers: { "Content-Type": "application/json" },
-        body: JSON.stringify({ title: tr.dataset.originalTitle }),
-      });
-      if (!r.ok) throw new Error(await r.text());
-      tr.remove();
-      setStatus("Deleted");
-    } catch (err) {
-      setStatus("Delete failed: " + err.message);
-    }
-  });
-  actTd.append(save, del);
-  tr.appendChild(actTd);
-  return tr;
-}
-
-async function load() {
-  setStatus("Loading…");
-  const tbody = document.getElementById("rows");
-  tbody.innerHTML = "";
-  try {
-    const r = await fetch("/api/matches");
-    const data = await r.json();
-    const matches = data.matches || {};
-    const titles = Object.keys(matches).sort((a,b)=>a.localeCompare(b));
-    for (const t of titles) tbody.appendChild(makeRow(t, matches[t]));
-    setStatus(titles.length + " entries");
-    applyFilter();
-  } catch (err) {
-    setStatus("Load failed: " + err.message);
-  }
-}
-
-function applyFilter() {
-  const q = document.getElementById("filter").value.toLowerCase();
-  for (const tr of document.querySelectorAll("#rows tr")) {
-    const t = tr.dataset.originalTitle.toLowerCase();
-    tr.style.display = t.includes(q) ? "" : "none";
-  }
-}
-
-document.getElementById("filter").addEventListener("input", applyFilter);
-document.getElementById("reload").addEventListener("click", load);
-document.getElementById("build").addEventListener("click", async () => {
-  if (!confirm("Run full scan? This may take several minutes.")) return;
-  setStatus("Building… (running on the server)");
-  try {
-    const r = await fetch("/api/build", { method: "POST" });
-    if (!r.ok) throw new Error(await r.text());
-    setStatus("Build finished");
-    load();
-  } catch (err) {
-    setStatus("Build failed: " + err.message);
-  }
-});
-
-load();
-</script>
-</body>
-</html>
-"""
-
-
-class MatchesWebApp:
-    """
-    Flask app exposing the MatchesCache. `mover` is optional — if provided,
-    POST /api/build triggers SuwayomiMover.build_matches_only() on a worker
-    thread.
-    """
-
-    def __init__(self, cache: MatchesCache, *,
-                 mover=None,
-                 host: str = "0.0.0.0",
-                 port: int = 8080):
-        self._cache = cache
-        self._mover = mover
-        self._host = host
-        self._port = port
-        self._build_lock = threading.Lock()
-        self._app = Flask(__name__)
-        self._thread: "threading.Thread | None" = None
-        self._register_routes()
-
-    @property
-    def app(self) -> Flask:
-        return self._app
-
-    def start(self) -> threading.Thread:
-        """
-        Starts the Flask server on a background thread and returns it.
-
-        The thread is non-daemon so the process stays alive even when the
-        caller does not explicitly join() — important when this is the
-        only foreground task (e.g. watcher disabled for testing).
-        """
-        if self._thread is not None and self._thread.is_alive():
-            return self._thread
-        self._thread = threading.Thread(
-            target=self._app.run,
-            kwargs={"host": self._host, "port": self._port,
-                    "debug": False, "use_reloader": False,
-                    "threaded": True},
-            name="MatchesWebApp",
-            daemon=False,
-        )
-        self._thread.start()
-        print(f"[MatchesWebApp] listening on {self._host}:{self._port}",
-              flush=True)
-        return self._thread
-
-    def wait(self) -> None:
-        """Blocks until the Flask thread exits (or returns immediately if not started)."""
-        if self._thread is not None:
-            self._thread.join()
-
-    # ------------------------------------------------------------------
-    # Routes
-    # ------------------------------------------------------------------
-    def _register_routes(self) -> None:
-        app = self._app
-        cache = self._cache
-
-        @app.get("/")
-        def index() -> Response:
-            return Response(_INDEX_HTML, mimetype="text/html; charset=utf-8")
-
-        @app.get("/api/matches")
-        def api_list():
-            return jsonify(cache.all())
-
-        @app.post("/api/matches")
-        def api_upsert():
-            body = request.get_json(silent=True) or {}
-            title = (body.get("title") or "").strip()
-            if not title:
-                return Response("title is required", status=400)
-            original = (body.get("originalTitle") or "").strip() or title
-            if original != title:
-                cache.rename(original, title)
-            entry = cache.upsert(
-                title,
-                mangabaka_id=body.get("mangabakaId"),
-                mangabaka_name=body.get("mangabakaName"),
-                image_url=body.get("imageUrl"),
-                first_match_time=body.get("firstMatchTime"),
-            )
-            return jsonify({"title": title, "entry": entry})
-
-        @app.post("/api/matches/delete")
-        def api_delete():
-            body = request.get_json(silent=True) or {}
-            title = (body.get("title") or "").strip()
-            if not title:
-                return Response("title is required", status=400)
-            removed = cache.remove(title)
-            return jsonify({"removed": removed, "title": title})
-
-        @app.post("/api/build")
-        def api_build():
-            if self._mover is None:
-                return Response("no mover configured", status=503)
-            if not self._build_lock.acquire(blocking=False):
-                return Response("build already running", status=409)
-            try:
-                result = self._mover.build_matches_only()
-            finally:
-                self._build_lock.release()
-            return jsonify(result)

src/TextUtils.py

@@ -0,0 +1,72 @@
+"""
+text_utils.py
+=============
+
+Small text helpers shared across modules:
+
+* ``paragraphs_to_html`` — converts plain text with blank-line paragraph
+  breaks into compact HTML (used for Kavita summary / description fields,
+  which must not contain raw newlines).
+* ``best_similarity`` — best difflib ratio between a query string and a
+  list of candidate strings (used for title / person-name matching).
+"""
+
+from __future__ import annotations
+
+import difflib
+import re
+from typing import Iterable
+
+
+def paragraphs_to_html(text: str) -> str:
+    """Converts plain text with paragraph breaks to compact HTML (no raw \\n)."""
+    if not text:
+        return ""
+    parts: list[str] = []
+    for para in re.split(r"\n{2,}", text.strip()):
+        para = para.strip()
+        if para:
+            parts.append(f"<p>{para.replace(chr(10), '<br>')}</p>")
+    return "".join(parts)
+
+
+def best_similarity(query: str, candidates: Iterable[str]) -> float:
+    """
+    Returns the best case-insensitive difflib similarity ratio (0..1)
+    between `query` and any non-empty candidate.
+    """
+    q = (query or "").lower()
+    best = 0.0
+    for candidate in candidates:
+        if candidate:
+            ratio = difflib.SequenceMatcher(
+                None, q, str(candidate).lower()).ratio()
+            best = max(best, ratio)
+    return best
+
+
+def person_name_with_id(name: str, *,
+                        mal_id: "int | None" = None,
+                        al_id:  "int | None" = None) -> str:
+    """
+    Disambiguates a character name with its tracker id: "Rem (MAL 118737)".
+
+    Kavita Person records are global and keyed by name only, so two
+    different characters who share a name would collapse into one record.
+    Suffixing the tracker *character* id keeps them apart while still
+    sharing the record across the manga and light-novel version of the
+    same series (MAL/AniList character ids are per character, not per
+    medium).  MAL is preferred; AniList ids get an "AL" marker so the two
+    id spaces cannot collide.  Without any id the name is returned as-is.
+
+    The format must stay in sync with the manga project so both tools
+    address the same Kavita person records.
+    """
+    name = (name or "").strip()
+    if not name:
+        return name
+    if mal_id:
+        return f"{name} (MAL {mal_id})"
+    if al_id:
+        return f"{name} (AL {al_id})"
+    return name

src/ln/KavitaSeriesUpdater.py

@@ -0,0 +1,313 @@
+"""
+kavita_series_updater.py
+========================
+
+Diff-based update of a single Kavita series record from a
+LightNovelMetadataBuilder output dict.
+
+Behaviour
+---------
+* Locked fields in Kavita (``*Locked`` flags) are never touched, no matter
+  what MangaBaka returns.
+* Scalar fields (summary, releaseYear, ageRating, publicationStatus,
+  language, score, sortName, localizedName) are overwritten when the
+  newly-built value differs from the value currently stored in Kavita.
+* List fields (genres, tags, characters, writers, coverArtists,
+  publishers, imprints) are diff-merged: a name appearing in the new
+  set but not in the current one is added (id=0 so Kavita creates the
+  record); a name that is in Kavita but no longer in the new set is
+  dropped.  Comparison is case-insensitive on the ``name`` field.
+* Web links are stored as a comma-separated string in Kavita; this
+  updater treats them as a set and re-joins on write.
+* Series-level cover image (URL different from last time) is re-uploaded
+  whenever ``coverImageLocked`` is False.  The MangaBaka cover URL is
+  stamped onto matches.json as ``imageUrl`` so a subsequent run can skip
+  the upload when nothing changed.
+
+Returns a small diff report ({field: 'changed'/'skipped'/'locked'}) per
+series so the WebApp can surface what happened.
+"""
+
+from __future__ import annotations
+
+from typing import Iterable
+
+from KavitaClient import KavitaClient
+
+
+# Maps Kavita "list" fields on SeriesMetadataDto to (lock_flag, item_key).
+# `item_key` is the dict key Kavita uses for the display name on each item:
+# GenreTagDto / TagDto use "title", PersonDto uses "name".
+_LIST_FIELDS: list[tuple[str, str, str]] = [
+    ("genres",       "genresLocked",      "title"),
+    ("tags",         "tagsLocked",        "title"),
+    ("characters",   "characterLocked",   "name"),
+    ("writers",      "writerLocked",      "name"),
+    ("coverArtists", "coverArtistLocked", "name"),
+    ("publishers",   "publisherLocked",   "name"),
+    ("imprints",     "imprintLocked",     "name"),
+]
+
+
+def _norm(name: str) -> str:
+    return (name or "").strip().lower()
+
+
+def _merge_list(
+    current: list[dict],
+    new_names: Iterable[str],
+    item_key: str,
+) -> "tuple[list[dict], bool]":
+    """
+    Diff-merges a Kavita list field with the canonical name list from
+    MangaBaka.  Returns (merged_list, changed_flag).
+
+    `item_key` is the dict key Kavita uses for the display name on each
+    item ("title" for GenreTagDto/TagDto, "name" for PersonDto).
+
+    * Items in `current` whose display value appears in `new_names` are
+      kept verbatim so existing ids and ancillary fields survive.
+    * New names (no matching entry in `current`) are appended with
+      ``{"id": 0, <item_key>: <name>}`` — Kavita creates the record on save.
+    * Items in `current` whose display value is *not* in `new_names` are
+      dropped.
+    """
+    new_set = [n for n in new_names if n and n.strip()]
+    new_index = {_norm(n): n.strip() for n in new_set}
+
+    merged: list[dict] = []
+    kept_keys: set[str] = set()
+    for item in (current or []):
+        key = _norm(item.get(item_key))
+        if key in new_index:
+            merged.append(item)
+            kept_keys.add(key)
+
+    added = False
+    for key, display in new_index.items():
+        if key not in kept_keys:
+            merged.append({"id": 0, item_key: display})
+            added = True
+
+    removed = len(current or []) != len(kept_keys)
+    return merged, added or removed
+
+
+def _parse_web_links(value) -> list[str]:
+    if not value:
+        return []
+    if isinstance(value, list):
+        return [str(v).strip() for v in value if v]
+    return [p.strip() for p in str(value).split(",") if p.strip()]
+
+
+def _merge_web_links(current_str, new_links: list[str]) -> "tuple[str, bool]":
+    current = _parse_web_links(current_str)
+    new_norm = [l for l in new_links if l]
+    if not new_norm:
+        return ",".join(current), False
+
+    # Mirror MangaBaka's set: keep order from new_norm, then anything from
+    # current that's still in new_norm (already covered above).  Anything
+    # in current that's not in new_norm is dropped.
+    new_set = set(new_norm)
+    merged = list(new_norm)
+    changed = sorted(new_set) != sorted(set(current))
+    return ",".join(merged), changed
+
+
+class KavitaSeriesUpdater:
+    def __init__(self, client: KavitaClient):
+        self._client = client
+
+    # ------------------------------------------------------------------
+    # Public
+    # ------------------------------------------------------------------
+    def update_series(self, series_id: int, built: dict, *,
+                      previous_cover_url: "str | None" = None) -> dict:
+        """
+        Applies the diff between Kavita's current state for `series_id`
+        and the freshly-built MangaBaka dict.  Returns a per-field diff
+        report.
+        """
+        series   = self._client.get_series(series_id)
+        metadata = self._client.get_series_metadata(series_id)
+        report: dict = {}
+
+        meta_changed = self._diff_metadata(metadata, built, report)
+        if meta_changed:
+            self._client.update_series_metadata(metadata)
+
+        series_changed = self._diff_series(series, built, report)
+        if series_changed:
+            self._client.update_series(series)
+
+        # Cover: only re-upload when not locked AND URL actually changed.
+        new_cover = built.get("coverUrl")
+        if (new_cover
+                and not series.get("coverImageLocked")
+                and new_cover != previous_cover_url):
+            try:
+                self._client.upload_series_cover(series_id, new_cover)
+                report["coverImage"] = "changed"
+            except Exception as exc:
+                report["coverImage"] = f"error: {exc}"
+        elif series.get("coverImageLocked"):
+            report["coverImage"] = "locked"
+        else:
+            report["coverImage"] = "skipped"
+
+        return report
+
+    # ------------------------------------------------------------------
+    # Internal: SeriesMetadataDto
+    # ------------------------------------------------------------------
+    def _diff_metadata(self, metadata: dict, built: dict,
+                       report: dict) -> bool:
+        changed = False
+
+        # ----- Scalars ------------------------------------------------
+        # (built_key, metadata_key, locked_key, transform, skip_when_zero)
+        # `skip_when_zero` covers fields where 0 means "no data" rather
+        # than a real value (releaseYear, ageRating).  publicationStatus 0
+        # is a valid "Ongoing" status — never skip it.
+        scalar_map = [
+            ("summary",            "summary",            "summaryLocked",            None, False),
+            ("releaseYear",        "releaseYear",        "releaseYearLocked",        int,  True),
+            ("ageRating",          "ageRating",          "ageRatingLocked",          int,  True),
+            ("publicationStatus",  "publicationStatus",  "publicationStatusLocked",  int,  False),
+            ("language",           "language",           "languageLocked",           None, False),
+        ]
+        for built_key, meta_key, locked_key, transform, skip_zero in scalar_map:
+            new_val = built.get(built_key)
+            if new_val is None or new_val == "":
+                report[meta_key] = "skipped"
+                continue
+            if transform is not None:
+                try:
+                    new_val = transform(new_val)
+                except (TypeError, ValueError):
+                    report[meta_key] = "skipped"
+                    continue
+            if skip_zero and new_val == 0:
+                report[meta_key] = "skipped"
+                continue
+            if metadata.get(locked_key):
+                report[meta_key] = "locked"
+                continue
+            if metadata.get(meta_key) != new_val:
+                metadata[meta_key] = new_val
+                changed = True
+                report[meta_key] = "changed"
+            else:
+                report[meta_key] = "unchanged"
+
+        # ----- Web links (single comma-separated string) ---------------
+        # SeriesMetadataDto has no dedicated lock for webLinks — always update.
+        web_str, web_changed = _merge_web_links(
+            metadata.get("webLinks"), built.get("webLinks") or [])
+        if web_changed:
+            metadata["webLinks"] = web_str
+            changed = True
+            report["webLinks"] = "changed"
+        else:
+            report["webLinks"] = "unchanged"
+
+        # ----- List fields --------------------------------------------
+        list_map = {
+            "genres":       built.get("genres"),
+            "tags":         built.get("tags"),
+            "characters":   built.get("characters"),
+            "writers":      built.get("writers"),
+            "coverArtists": built.get("coverArtists"),
+            "publishers":   built.get("publishers"),
+            "imprints":     [built["imprint"]] if built.get("imprint") else [],
+        }
+        for meta_key, locked_key, item_key in _LIST_FIELDS:
+            new_names = list_map.get(meta_key) or []
+            if metadata.get(locked_key):
+                report[meta_key] = "locked"
+                continue
+            if not new_names and not (metadata.get(meta_key) or []):
+                report[meta_key] = "unchanged"
+                continue
+            merged, list_changed = _merge_list(
+                metadata.get(meta_key) or [], new_names, item_key)
+            if list_changed:
+                metadata[meta_key] = merged
+                changed = True
+                report[meta_key] = "changed"
+            else:
+                report[meta_key] = "unchanged"
+
+        return changed
+
+    # ------------------------------------------------------------------
+    # Internal: SeriesDto (sortName, userRating, tracker ids)
+    # ------------------------------------------------------------------
+    def _diff_series(self, series: dict, built: dict, report: dict) -> bool:
+        changed = False
+
+        # sortName / localizedName
+        if not series.get("sortNameLocked"):
+            new_sort = built.get("sortName") or ""
+            if new_sort and series.get("sortName") != new_sort:
+                series["sortName"] = new_sort
+                changed = True
+                report["sortName"] = "changed"
+            else:
+                report["sortName"] = "unchanged"
+        else:
+            report["sortName"] = "locked"
+
+        if not series.get("localizedNameLocked"):
+            new_loc = built.get("localizedName") or ""
+            if new_loc and series.get("localizedName") != new_loc:
+                series["localizedName"] = new_loc
+                changed = True
+                report["localizedName"] = "changed"
+            else:
+                report["localizedName"] = "unchanged"
+        else:
+            report["localizedName"] = "locked"
+
+        # Tracker ids — Kavita exposes malId, aniListId, mangaBakaId
+        for built_key, series_key in (
+            ("malId",       "malId"),
+            ("anilistId",   "aniListId"),
+            ("mangabakaId", "mangaBakaId"),
+        ):
+            new_val = built.get(built_key)
+            if new_val in (None, "", 0):
+                continue
+            try:
+                new_int = int(new_val)
+            except (TypeError, ValueError):
+                continue
+            if int(series.get(series_key) or 0) != new_int:
+                series[series_key] = new_int
+                changed = True
+                report[series_key] = "changed"
+
+        # userRating from MangaBaka (0..5)
+        new_score = built.get("score")
+        if new_score is not None:
+            try:
+                new_score = float(new_score)
+            except (TypeError, ValueError):
+                new_score = None
+        if new_score is not None:
+            current_score = series.get("userRating")
+            try:
+                current_score = float(current_score) if current_score is not None else None
+            except (TypeError, ValueError):
+                current_score = None
+            if current_score != new_score:
+                series["userRating"] = new_score
+                series["hasUserRated"] = True
+                changed = True
+                report["userRating"] = "changed"
+            else:
+                report["userRating"] = "unchanged"
+
+        return changed

src/ln/LightNovelMetadataBuilder.py

@@ -0,0 +1,571 @@
+"""
+light_novel_metadata_builder.py
+===============================
+
+Fetches series-level metadata for a light novel from MangaBaka, enriches
+it with MyAnimeList / AniList tracker statistics and character data, and
+returns a structured dict ready to be diffed against Kavita's
+SeriesMetadataDto.
+
+Differences vs. the manga project's ComicInfoBuilder:
+  - No chapter / page handling — Kavita reads volumes from the files.
+  - No XML output — produces a plain dict.
+  - No MangaDex resolver — light novels don't have a chapter→volume
+    mapping problem.
+  - MangaBaka search type is fixed to ``novel`` so only light/web novels
+    are returned.
+"""
+
+from __future__ import annotations
+
+import re
+
+import requests
+
+from MangaBakaRateLimit import apply_to_session as _apply_mangabaka_rate_limit
+from MALResolver import MALResolver
+from AniListResolver import AniListResolver
+from MatchesCache import MatchesCache
+from TextUtils import paragraphs_to_html, person_name_with_id
+
+
+# MangaBaka series type for the search endpoint.
+_SEARCH_TYPES = ["novel"]
+
+# MangaBaka content_rating  ->  Kavita AgeRating enum
+# Kavita AgeRating values (from openapi.json):
+#   0=Unknown, 3=Everyone, 8=Teen, 10=Mature17Plus, 13=AdultsOnly
+_AGE_RATING_MAP = {
+    "safe":          3,    # Everyone
+    "suggestive":    8,    # Teen
+    "erotica":       10,   # Mature17Plus
+    "pornographic":  13,   # AdultsOnly
+}
+
+# MangaBaka status  ->  Kavita PublicationStatus enum
+# Kavita PublicationStatus (from openapi.json):
+#   0=OnGoing, 1=Hiatus, 2=Completed, 3=Cancelled, 4=Ended
+_PUB_STATUS_MAP = {
+    "ongoing":   0,
+    "hiatus":    1,
+    "completed": 2,
+    "cancelled": 3,
+    "ended":     4,
+}
+
+# External-tracker URL templates used to enrich the web-links list.
+_TRACKER_URL_TEMPLATES = {
+    "anilist":          "https://anilist.co/manga/{id}",
+    "myanimelist":      "https://myanimelist.net/manga/{id}",
+    "mal":              "https://myanimelist.net/manga/{id}",
+    "mangaupdates":     "https://www.mangaupdates.com/series.html?id={id}",
+    "kitsu":            "https://kitsu.app/manga/{id}",
+    "animenewsnetwork": "https://www.animenewsnetwork.com/encyclopedia/manga.php?id={id}",
+    "ann":              "https://www.animenewsnetwork.com/encyclopedia/manga.php?id={id}",
+    "animeplanet":      "https://www.anime-planet.com/manga/{id}",
+    "shikimori":        "https://shikimori.one/mangas/{id}",
+    "bookwalker":       "https://bookwalker.jp/{id}",
+}
+
+_MD_ESCAPE_RE = re.compile(r'\\([\\`*_{}\[\]()\#+\-.!|~])')
+
+
+# --------------------------------------------------------------------------
+# Helpers
+# --------------------------------------------------------------------------
+def _normalise_key(key) -> str:
+    return re.sub(r"[^a-z0-9]", "", str(key).lower())
+
+
+def _format_term(value: str) -> str:
+    return str(value).replace("_", " ").strip().title() if value else ""
+
+
+def _md_to_html(text: str) -> str:
+    """Converts the subset of Markdown produced by MangaBaka to compact HTML."""
+    if not text:
+        return ""
+    text = _MD_ESCAPE_RE.sub(r'\1', text)
+    text = re.sub(
+        r'\[([^\]]+)\]\(([^)]+)\)',
+        lambda m: f'<a href="{m.group(2)}">{m.group(1)}</a>',
+        text,
+    )
+    text = re.sub(r'\*\*(.+?)\*\*', r'<strong>\1</strong>', text, flags=re.DOTALL)
+    text = re.sub(r'\*(.+?)\*',     r'<em>\1</em>',         text, flags=re.DOTALL)
+    return paragraphs_to_html(text)
+
+
+def pick_cover_url(cover) -> "str | None":
+    """Selects the best cover URL from a MangaBaka cover object."""
+    if not cover:
+        return None
+    if isinstance(cover, str):
+        return cover
+    if not isinstance(cover, dict):
+        return None
+
+    raw = cover.get("raw")
+    if isinstance(raw, dict):
+        url = raw.get("url")
+        if isinstance(url, str) and url:
+            return url
+    elif isinstance(raw, str) and raw:
+        return raw
+
+    for size_key in ("x350", "x250", "x150"):
+        variant = cover.get(size_key)
+        if isinstance(variant, dict):
+            for density in ("x3", "x2", "x1"):
+                url = variant.get(density)
+                if isinstance(url, str) and url:
+                    return url
+        elif isinstance(variant, str) and variant:
+            return variant
+
+    for val in cover.values():
+        if isinstance(val, str) and val.startswith("http"):
+            return val
+        if isinstance(val, dict):
+            for sub in val.values():
+                if isinstance(sub, str) and sub.startswith("http"):
+                    return sub
+    return None
+
+
+def pick_thumbnail_url(cover) -> "str | None":
+    """Picks a small cover variant suitable for a UI thumbnail."""
+    if not cover:
+        return None
+    if isinstance(cover, str):
+        return cover
+    if not isinstance(cover, dict):
+        return None
+    for size_key in ("x150", "x250", "x350"):
+        variant = cover.get(size_key)
+        if isinstance(variant, dict):
+            for density in ("x2", "x1", "x3"):
+                url = variant.get(density)
+                if isinstance(url, str) and url:
+                    return url
+        elif isinstance(variant, str) and variant:
+            return variant
+    return pick_cover_url(cover)
+
+
+def _id_from_source(md: dict, *names: str) -> "int | None":
+    target = {_normalise_key(n) for n in names}
+    for raw_key, info in (md.get("source") or {}).items():
+        if _normalise_key(raw_key) in target and isinstance(info, dict):
+            mid = info.get("id")
+            if mid is not None:
+                try:
+                    return int(mid)
+                except (TypeError, ValueError):
+                    pass
+    return None
+
+
+# --------------------------------------------------------------------------
+# Builder
+# --------------------------------------------------------------------------
+class LightNovelMetadataBuilder:
+    """
+    Resolves a light-novel series on MangaBaka and produces a structured
+    metadata dict ready to be merged into Kavita.
+    """
+
+    def __init__(self, *,
+                 api_base_url: str = "https://api.mangabaka.dev/v1",
+                 language: str = "en",
+                 request_timeout: int = 30,
+                 session: "requests.Session | None" = None,
+                 mal_resolver: "MALResolver | None" = None,
+                 al_resolver: "AniListResolver | None" = None,
+                 matches_cache: "MatchesCache | None" = None):
+        self.api_base_url = api_base_url.rstrip("/")
+        self.language = language
+        self.request_timeout = request_timeout
+
+        self._session = session or requests.Session()
+        self._session.headers.setdefault("User-Agent",
+                                         "LightNovelMetadataBuilder/1.0")
+        _apply_mangabaka_rate_limit(self._session)
+
+        self._mal = mal_resolver or MALResolver(
+            request_timeout=request_timeout, search_type="lightnovel")
+        self._al  = al_resolver  or AniListResolver(
+            request_timeout=request_timeout, media_format="novel")
+        self._matches_cache = matches_cache
+
+    # ------------------------------------------------------------------
+    # MangaBaka search / fetch
+    # ------------------------------------------------------------------
+    def search_series(self, title: str) -> "dict | None":
+        """Returns the top MangaBaka novel hit for `title`, or None."""
+        if not title or not title.strip():
+            return None
+        url = f"{self.api_base_url}/series/search"
+        try:
+            resp = self._session.get(
+                url, params={"q": title, "type": _SEARCH_TYPES,
+                             "page": 1, "limit": 1},
+                timeout=self.request_timeout)
+            resp.raise_for_status()
+        except requests.RequestException:
+            return None
+        data = resp.json().get("data") or []
+        return data[0] if data else None
+
+    def fetch_series(self, series_id) -> "dict | None":
+        """
+        Returns the full MangaBaka series dict for the given id, following
+        ``merged_with`` redirects.  A seen-set guards against merge cycles.
+        """
+        if series_id is None or str(series_id).strip() == "":
+            return None
+        seen: set[str] = set()
+        current = series_id
+        while str(current) not in seen:
+            seen.add(str(current))
+            url = f"{self.api_base_url}/series/{current}"
+            resp = self._session.get(url, timeout=self.request_timeout)
+            resp.raise_for_status()
+            data = resp.json().get("data")
+            if data and data.get("state") == "merged" and data.get("merged_with"):
+                current = data["merged_with"]
+                continue
+            return data
+        return None
+
+    # ------------------------------------------------------------------
+    # Resolve title -> MangaBaka series (caches the match)
+    # ------------------------------------------------------------------
+    def resolve(self, title: str) -> "dict | None":
+        """
+        Returns the MangaBaka series for `title`.
+
+        Lookup order:
+          1. MatchesCache (uses stored mangabakaId, skips the search).
+          2. Fresh MangaBaka search — top hit. Result is persisted to the
+             cache so it survives a crash.
+        """
+        if self._matches_cache is not None:
+            cached = self._matches_cache.get(title)
+            if cached and cached.get("mangabakaId"):
+                try:
+                    series = self.fetch_series(cached["mangabakaId"])
+                    if series:
+                        return series
+                except Exception:
+                    pass
+
+        series = self.search_series(title)
+        if series and self._matches_cache is not None:
+            self._matches_cache.upsert(
+                title,
+                mangabaka_id=series.get("id"),
+                mangabaka_name=series.get("title") or "",
+                image_url=pick_thumbnail_url(series.get("cover")),
+            )
+        return series
+
+    # ------------------------------------------------------------------
+    # Main entry point
+    # ------------------------------------------------------------------
+    def build(self, *, title: str = "",
+              mangabaka_id=None) -> "dict | None":
+        """
+        Fetches and enriches metadata for one series, returning the
+        normalised dict described in the module docstring.
+
+        Pass either `title` (will resolve via cache/search) or
+        `mangabaka_id` (direct fetch).
+        """
+        if mangabaka_id is not None and str(mangabaka_id).strip():
+            md = self.fetch_series(mangabaka_id)
+        else:
+            md = self.resolve(title)
+        if not md:
+            return None
+        return self._assemble(md)
+
+    # ------------------------------------------------------------------
+    # Internal: assemble the result dict
+    # ------------------------------------------------------------------
+    def _assemble(self, md: dict) -> dict:
+        mal_id = _id_from_source(md, "myanimelist", "mal")
+        al_id  = _id_from_source(md, "anilist")
+
+        # Fall back to a title-based MAL lookup when the source map does
+        # not carry an id — Jikan is the only tracker that ships staff
+        # data we can use to enrich author / artist person records.
+        if mal_id is None:
+            mal_id = self._mal.find_mal_id(md.get("title") or "")
+
+        mal_stats = self._mal.get_stats(mal_id) if mal_id else None
+
+        characters_detailed = self._mal.get_characters_detailed(mal_id) if mal_id else []
+        if not characters_detailed and al_id:
+            characters_detailed = self._al.get_characters_detailed(al_id)
+
+        staff_detailed = self._mal.get_staff_detailed(mal_id) if mal_id else []
+        if not staff_detailed and al_id:
+            staff_detailed = self._al.get_staff_detailed(al_id)
+
+        # Character names for SeriesMetadata, disambiguated with the
+        # tracker character id ("Rem (MAL 118737)") because Kavita person
+        # records are global and keyed by name only.
+        character_names = [
+            person_name_with_id(c["name"],
+                                mal_id=c.get("mal_id"),
+                                al_id=c.get("al_id"))
+            for c in characters_detailed if c.get("name")
+        ]
+        # Writers come from MangaBaka first (authoritative for novels)
+        writers = list(md.get("authors") or [])
+        # Illustrators / artists -> CoverArtists (Kavita has no dedicated
+        # illustrator field, and Pencillers is the wrong semantic for
+        # text-only novels).
+        cover_artists = list(md.get("artists") or [])
+
+        # Publisher: prefer English licence, else original.  When both
+        # exist, the original publisher becomes the imprint.
+        english_pubs  = self._publishers_by_type(md, "English")
+        original_pubs = self._publishers_by_type(md, "Original")
+        publishers = english_pubs or original_pubs
+        imprint = original_pubs[0] if english_pubs and original_pubs else None
+
+        # Release year
+        release_year = None
+        try:
+            if md.get("year") is not None:
+                release_year = int(md["year"])
+        except (TypeError, ValueError):
+            pass
+
+        # Score: MangaBaka rating is 0..100  ->  Kavita userRating is 0..5
+        score = None
+        if md.get("rating") is not None:
+            try:
+                score = round(float(md["rating"]) / 20.0, 1)
+            except (TypeError, ValueError):
+                pass
+
+        # Tags / genres come back as snake_case slugs.
+        genres = [_format_term(g) for g in (md.get("genres") or []) if g]
+        tags   = [_format_term(t) for t in (md.get("tags")   or []) if t]
+
+        # Web links
+        web_links = self._collect_web_links(md)
+
+        # Summary HTML
+        summary = self._build_summary(md, mal_stats)
+
+        # Cover URL
+        cover_url = pick_cover_url(md.get("cover"))
+
+        # Title variants
+        all_alt = self._collect_all_alt_titles(md)
+
+        return {
+            "mangabakaId":     str(md.get("id") or ""),
+            "mangabakaTitle":  md.get("title") or "",
+            "originalName":    md.get("native_title") or "",
+            "localizedName":   md.get("romanized_title") or "",
+            "sortName":        self._sort_title(md),
+            "altTitles":       all_alt,
+            "summary":         summary,
+            "genres":          genres,
+            "tags":            tags,
+            "characters":      character_names,
+            "writers":         writers,
+            "coverArtists":    cover_artists,
+            "publishers":      publishers,
+            "imprint":         imprint,
+            "releaseYear":     release_year,
+            "ageRating":       _AGE_RATING_MAP.get(md.get("content_rating"), 0),
+            "publicationStatus": _PUB_STATUS_MAP.get(
+                (md.get("status") or "").lower(), 0),
+            "language":        self.language,
+            "webLinks":        web_links,
+            "score":           score,
+            "coverUrl":        cover_url,
+            "malId":           mal_id,
+            "anilistId":       al_id,
+            "relationships":   list(md.get("relationships_v2") or []),
+            "charactersDetailed": characters_detailed,
+            "staffDetailed":   staff_detailed,
+            "raw":             md,
+        }
+
+    # ------------------------------------------------------------------
+    # Internal helpers
+    # ------------------------------------------------------------------
+    @staticmethod
+    def _publishers_by_type(md: dict, ptype: str) -> list[str]:
+        return [p.get("name") for p in (md.get("publishers") or [])
+                if p.get("type") == ptype and p.get("name")]
+
+    def _sort_title(self, md: dict) -> str:
+        lang = self.language.lower()
+        alts = self._collect_alt_titles(md)
+        return alts.get(lang) or md.get("title") or ""
+
+    def _collect_alt_titles(self, md: dict) -> "dict[str, str]":
+        """Returns one best title per language code (en/de/jp/romaji)."""
+        titles = md.get("titles") or md.get("alt_titles") or []
+
+        def pick(language_codes: tuple, prefer_trait: "str | None" = None
+                 ) -> "str | None":
+            best_score = -1
+            best_title: "str | None" = None
+            for entry in titles:
+                if not isinstance(entry, dict):
+                    continue
+                lang = (entry.get("language") or entry.get("lang") or "").lower()
+                if lang not in language_codes:
+                    continue
+                title = entry.get("title")
+                if not title:
+                    continue
+                traits = entry.get("traits") or []
+                score = 0
+                if prefer_trait and prefer_trait in traits:
+                    score += 4
+                if "official" in traits:
+                    score += 2
+                if entry.get("is_primary"):
+                    score += 1
+                if score > best_score:
+                    best_score, best_title = score, title
+            return best_title
+
+        result: dict[str, str] = {}
+        kanji = pick(("ja",), prefer_trait="native") or md.get("native_title")
+        if kanji:
+            result["jp"] = kanji
+        romaji = pick(("ja-latn", "ja-romaji"))
+        if not romaji:
+            rt = md.get("romanized_title") or ""
+            if rt and all(ord(c) < 128 for c in rt):
+                romaji = rt
+        if romaji:
+            result["romaji"] = romaji
+        en = pick(("en",)) or md.get("title")
+        if en:
+            result["en"] = en
+        de = pick(("de",))
+        if de:
+            result["de"] = de
+        return result
+
+    @staticmethod
+    def _collect_all_alt_titles(md: dict) -> "dict[str, list[str]]":
+        _GROUPS = {
+            "en":        ("en",),
+            "de":        ("de",),
+            "ja":        ("ja",),
+            "ja-romaji": ("ja-latn", "ja-romaji"),
+            "ko":        ("ko",),
+            "ko-romaji": ("ko-latn", "ko-romaji"),
+            "zh":        ("zh", "zh-hk", "zh-tw", "zh-hans", "zh-hant"),
+            "zh-romaji": ("zh-latn",),
+        }
+        lang_to_group = {l: g for g, ls in _GROUPS.items() for l in ls}
+        result: dict[str, list[str]] = {}
+        seen:   dict[str, set]       = {}
+        for entry in (md.get("titles") or md.get("alt_titles") or []):
+            if not isinstance(entry, dict):
+                continue
+            lang  = (entry.get("language") or entry.get("lang") or "").lower()
+            group = lang_to_group.get(lang)
+            if not group:
+                continue
+            title = (entry.get("title") or "").strip()
+            if not title:
+                continue
+            result.setdefault(group, [])
+            seen.setdefault(group, set())
+            if title not in seen[group]:
+                result[group].append(title)
+                seen[group].add(title)
+        return result
+
+    def _collect_web_links(self, md: dict) -> list[str]:
+        links: list[str] = [l for l in (md.get("links") or []) if l]
+        for raw_key, info in (md.get("source") or {}).items():
+            template = _TRACKER_URL_TEMPLATES.get(_normalise_key(raw_key))
+            if not template or not isinstance(info, dict):
+                continue
+            source_id = info.get("id")
+            if source_id is not None:
+                links.append(template.format(id=source_id))
+        seen: set[str] = set()
+        unique: list[str] = []
+        for link in links:
+            if link not in seen:
+                seen.add(link)
+                unique.append(link)
+        return unique
+
+    def _build_summary(self, md: dict,
+                       mal_stats: "dict | None") -> str:
+        """Builds the HTML summary with stats table + description + alt titles."""
+        _TD = 'style="padding-right:1.5em"'
+        parts: list[str] = []
+
+        if mal_stats:
+            url     = mal_stats.get("url", "")
+            as_of   = mal_stats.get("as_of", "")
+            rows: list[str] = []
+            for label, key, fmt in (
+                ("Score",     "score",      "{}"),
+                ("Ranked",    "rank",       "#{}"),
+                ("Scored by", "scored_by",  "{:,} users"),
+                ("Popularity","popularity", "#{}"),
+                ("Members",   "members",    "{:,}"),
+                ("Favorites", "favorites",  "{:,}"),
+            ):
+                v = mal_stats.get(key)
+                if v is None:
+                    continue
+                try:
+                    formatted = fmt.format(v)
+                except (TypeError, ValueError):
+                    formatted = str(v)
+                rows.append(f"<tr><td {_TD}>{label}</td><td>{formatted}</td></tr>")
+            if rows:
+                link = f'<a href="{url}" target="_blank">MyAnimeList</a>' if url else "MyAnimeList"
+                parts.append(f"<p>{link} stats as of {as_of}:</p>"
+                             f"<table>{''.join(rows)}</table>")
+
+        desc_raw = (md.get("description") or "").strip()
+        if desc_raw:
+            parts.append(_md_to_html(desc_raw))
+
+        all_alt = self._collect_all_alt_titles(md)
+        if all_alt:
+            label_map = {
+                "en":        "EN",
+                "de":        "DE",
+                "ja":        "JA",
+                "ja-romaji": "JA Romaji",
+                "ko":        "KO",
+                "ko-romaji": "KO Romaji",
+                "zh":        "ZH",
+                "zh-romaji": "ZH Romaji",
+            }
+            alt_rows: list[str] = []
+            for group in ("en", "de", "ja", "ja-romaji",
+                          "ko", "ko-romaji", "zh", "zh-romaji"):
+                titles = all_alt.get(group)
+                if not titles:
+                    continue
+                cell = "<br>".join(titles)
+                alt_rows.append(
+                    f"<tr><td {_TD}>{label_map[group]}</td><td>{cell}</td></tr>")
+            if alt_rows:
+                parts.append(f"<table>{''.join(alt_rows)}</table>")
+
+        return "<br>".join(parts)

src/ln/LightNovelOrchestrator.py

@@ -0,0 +1,260 @@
+"""
+light_novel_orchestrator.py
+===========================
+
+High-level workflow on top of the resolvers, the Kavita client and the
+diff-based updaters.  Exposes three operations to the WebApp:
+
+  - build_matches(library_ids):
+        Scan one or more Kavita libraries, resolve every series against
+        MangaBaka and persist the match in matches.json.
+  - update_series(kavita_series_id):
+        Re-fetch MangaBaka, MAL and AniList data for a single Kavita
+        series and apply the diff (metadata + persons + relationships).
+  - update_all(library_ids):
+        Run update_series for every series that has a match in the
+        cache and lives in the given libraries.
+
+A single shared HTTP session (rate-limited for MangaBaka) and shared
+resolver singletons are used across the whole run to maximise cache
+hits.
+"""
+
+from __future__ import annotations
+
+import requests
+
+from MangaBakaRateLimit import apply_to_session as _apply_mangabaka_rate_limit
+from MALResolver import MALResolver
+from AniListResolver import AniListResolver
+from MatchesCache import MatchesCache
+from KavitaClient import KavitaClient
+from KavitaPersonUpdater import KavitaPersonUpdater
+from KavitaSeriesUpdater import KavitaSeriesUpdater
+from LightNovelMetadataBuilder import (
+    LightNovelMetadataBuilder,
+    pick_thumbnail_url,
+)
+from RelationshipSync import RelationshipSync
+
+
+class LightNovelOrchestrator:
+    def __init__(self, *,
+                 kavita_url: str,
+                 kavita_api_key: str,
+                 matches_cache: MatchesCache,
+                 language: str = "en",
+                 request_timeout: int = 30,
+                 api_base_url: str = "https://api.mangabaka.dev/v1"):
+        self._cache = matches_cache
+        self._timeout = request_timeout
+
+        session = requests.Session()
+        session.headers.setdefault("User-Agent",
+                                   "KavitaLightNovelOrchestrator/1.0")
+        _apply_mangabaka_rate_limit(session)
+        self._session = session
+
+        # First construction in the LN container — pins the singletons to
+        # light-novel search mode (manga container uses the defaults).
+        self._mal = MALResolver(request_timeout=request_timeout,
+                                search_type="lightnovel")
+        self._al  = AniListResolver(request_timeout=request_timeout,
+                                    media_format="novel")
+
+        self._client = KavitaClient(kavita_url, kavita_api_key,
+                                    request_timeout=request_timeout)
+        self._builder = LightNovelMetadataBuilder(
+            api_base_url=api_base_url,
+            language=language,
+            request_timeout=request_timeout,
+            session=session,
+            mal_resolver=self._mal,
+            al_resolver=self._al,
+            matches_cache=matches_cache,
+        )
+        self._series_updater = KavitaSeriesUpdater(self._client)
+        self._person_updater = KavitaPersonUpdater(
+            self._client,
+            mal_resolver=self._mal,
+            al_resolver=self._al,
+        )
+        self._relation_sync = RelationshipSync(
+            self._client, matches_cache, builder=self._builder)
+
+    # ------------------------------------------------------------------
+    # Library listings
+    # ------------------------------------------------------------------
+    def list_libraries(self) -> list[dict]:
+        return self._client.list_libraries()
+
+    def list_series_in_libraries(self, library_ids: list[int]) -> list[dict]:
+        result: list[dict] = []
+        for lib_id in library_ids:
+            try:
+                result.extend(self._client.list_series_in_library(int(lib_id)))
+            except Exception as exc:
+                print(f"[orchestrator] library {lib_id} list failed: {exc}",
+                      flush=True)
+        return result
+
+    # ------------------------------------------------------------------
+    # Matching
+    # ------------------------------------------------------------------
+    def build_matches(self, library_ids: list[int]) -> dict:
+        """
+        Resolves every series in the given libraries against MangaBaka.
+
+        Series already present in matches.json keep their stored
+        mangabakaId; the kavitaSeriesId + libraryId fields are refreshed
+        in case the user moved a series between libraries.
+        """
+        stats = {"checked": 0, "matched": 0, "skipped": 0, "missing": 0}
+        for series in self.list_series_in_libraries(library_ids):
+            title = (series.get("name") or "").strip()
+            if not title:
+                continue
+            stats["checked"] += 1
+            kavita_id = int(series.get("id") or 0)
+            library_id = int(series.get("libraryId") or 0)
+
+            cached = self._cache.get(title)
+            if cached and cached.get("mangabakaId"):
+                self._cache.upsert(
+                    title,
+                    kavita_series_id=kavita_id,
+                    library_id=library_id,
+                )
+                stats["skipped"] += 1
+                continue
+
+            mb_series = self._builder.search_series(title)
+            if not mb_series:
+                self._cache.upsert(
+                    title,
+                    kavita_series_id=kavita_id,
+                    library_id=library_id,
+                )
+                stats["missing"] += 1
+                print(f"[match] {title!r}: no MangaBaka hit", flush=True)
+                continue
+
+            self._cache.upsert(
+                title,
+                mangabaka_id=mb_series.get("id"),
+                mangabaka_name=mb_series.get("title") or "",
+                image_url=pick_thumbnail_url(mb_series.get("cover")),
+                kavita_series_id=kavita_id,
+                library_id=library_id,
+            )
+            stats["matched"] += 1
+            print(f"[match] {title!r} -> {mb_series.get('title')!r} "
+                  f"(id={mb_series.get('id')})", flush=True)
+        return stats
+
+    # ------------------------------------------------------------------
+    # Updating
+    # ------------------------------------------------------------------
+    def update_series(self, kavita_series_id: int) -> dict:
+        """Runs a full metadata update for a single Kavita series."""
+        hit = self._cache.get_by_kavita_id(int(kavita_series_id))
+        if not hit:
+            # Try to resolve via the Kavita series name on the fly.
+            series = self._client.get_series(int(kavita_series_id))
+            title = (series.get("name") or "").strip()
+            if not title:
+                return {"ok": False, "error": "series not in matches.json"}
+            built = self._builder.build(title=title)
+            if not built:
+                return {"ok": False, "error": "no MangaBaka match"}
+            self._cache.upsert(
+                title,
+                mangabaka_id=built.get("mangabakaId"),
+                mangabaka_name=built.get("mangabakaTitle"),
+                image_url=built.get("coverUrl"),
+                kavita_series_id=int(kavita_series_id),
+                library_id=int(series.get("libraryId") or 0),
+            )
+            cached_title = title
+            cached_entry = self._cache.get(title) or {}
+        else:
+            cached_title, cached_entry = hit
+            built = self._builder.build(mangabaka_id=cached_entry.get("mangabakaId"))
+            if not built:
+                return {"ok": False, "error": "mangabaka id no longer resolvable"}
+
+        prev_cover = cached_entry.get("imageUrl") or ""
+        try:
+            series_report = self._series_updater.update_series(
+                int(kavita_series_id), built,
+                previous_cover_url=prev_cover,
+            )
+        except Exception as exc:
+            return {"ok": False, "error": f"series update failed: {exc}"}
+
+        # Persons
+        try:
+            person_report = self._person_updater.update_for_manga(
+                built.get("malId"),
+                al_manga_id=built.get("anilistId"),
+            )
+        except Exception as exc:
+            person_report = {"error": str(exc)}
+
+        # Relationships + collection
+        try:
+            relation_report = self._relation_sync.sync(
+                int(kavita_series_id), built)
+        except Exception as exc:
+            relation_report = {"error": str(exc)}
+
+        # Stamp the new cover URL on the cache so the next run knows when
+        # to re-upload.
+        self._cache.upsert(
+            cached_title,
+            image_url=built.get("coverUrl") or prev_cover,
+        )
+        self._cache.mark_updated(cached_title)
+
+        return {
+            "ok":            True,
+            "title":         cached_title,
+            "mangabakaId":   built.get("mangabakaId"),
+            "series":        series_report,
+            "persons":       person_report,
+            "relationships": relation_report,
+        }
+
+    def update_all(self, library_ids: "list[int] | None") -> dict:
+        """Updates every cached series in the given libraries."""
+        if library_ids is None:
+            entries = self._cache.all()["matches"]
+        else:
+            entries = self._cache.all_in_libraries(library_ids)["matches"]
+
+        results: list[dict] = []
+        ok = fail = 0
+        for title, entry in entries.items():
+            ksid = int(entry.get("kavitaSeriesId") or 0)
+            if not ksid or not entry.get("mangabakaId"):
+                continue
+            try:
+                res = self.update_series(ksid)
+            except Exception as exc:
+                res = {"ok": False, "error": str(exc)}
+            res["title"] = title
+            results.append(res)
+            if res.get("ok"):
+                ok += 1
+            else:
+                fail += 1
+            print(f"[update] {title!r}: "
+                  f"{'ok' if res.get('ok') else 'FAIL ' + str(res.get('error'))}",
+                  flush=True)
+        return {"ok": ok, "failed": fail, "results": results}
+
+    # ------------------------------------------------------------------
+    # Direct helpers exposed to the WebApp
+    # ------------------------------------------------------------------
+    def fetch_series(self, mangabaka_id) -> "dict | None":
+        return self._builder.fetch_series(mangabaka_id)

src/ln/MatchesCache.py

@@ -2,26 +2,30 @@
 matches_cache.py
 ================
 
-Persistent JSON cache that maps a Suwayomi/series search title to the
-MangaBaka series it was matched against.
+Persistent JSON cache that maps a Kavita series title to the MangaBaka
+series it was matched against, plus enough context to update the right
+Kavita record later.
 
 Structure on disk::
 
     {
       "matches": {
-        "<search title>": {
+        "<kavita series name>": {
           "mangabakaId":      "12345",
-          "mangabakaName":  "One-Punch Man",
+          "mangabakaName":    "Re:Zero",
           "imageUrl":         "https://.../cover.jpg",
-          "firstMatchTime": 1700000000
+          "kavitaSeriesId":   42,
+          "libraryId":        3,
+          "firstMatchTime":   1700000000,
+          "lastUpdateTime":   1700100000
         },
         ...
       }
     }
 
-The cache is consulted by ComicInfoBuilder before issuing a MangaBaka
-search request, and is written back to disk on every mutation so a crash
-does not lose matches that were resolved in the current run.
+The cache is the source of truth for the WebUI's matches table and is
+written back on every mutation so a crash mid-batch does not lose
+matches that were resolved in the current run.
 """
 
 from __future__ import annotations
@@ -32,6 +36,14 @@ import time
 from pathlib import Path
 
 
+def _set_int(entry: dict, key: str, value) -> None:
+    """Sets entry[key] = int(value); ignores values that don't coerce."""
+    try:
+        entry[key] = int(value)
+    except (TypeError, ValueError):
+        pass
+
+
 class MatchesCache:
     def __init__(self, path):
         self._path = Path(path)
@@ -47,26 +59,35 @@ class MatchesCache:
             entry = self._data["matches"].get(title)
             return dict(entry) if entry else None
 
-    def add(self, title: str, *,
-            mangabaka_id,
-            mangabaka_name: str,
-            image_url: "str | None") -> dict:
-        entry = {
-            "mangabakaId":    str(mangabaka_id) if mangabaka_id is not None else "",
-            "mangabakaName":  mangabaka_name or "",
-            "imageUrl":       image_url or "",
-            "firstMatchTime": int(time.time()),
-        }
+    def get_by_kavita_id(self, kavita_series_id: int) -> "tuple[str, dict] | None":
         with self._lock:
-            self._data["matches"][title] = entry
-            self._save_unlocked()
-        return dict(entry)
+            for title, entry in self._data["matches"].items():
+                if entry.get("kavitaSeriesId") == kavita_series_id:
+                    return title, dict(entry)
+        return None
+
+    def get_by_mangabaka_id(self, mangabaka_id) -> "tuple[str, dict] | None":
+        target = str(mangabaka_id) if mangabaka_id is not None else ""
+        if not target:
+            return None
+        with self._lock:
+            for title, entry in self._data["matches"].items():
+                if str(entry.get("mangabakaId") or "") == target:
+                    return title, dict(entry)
+        return None
 
     def upsert(self, title: str, *,
                mangabaka_id=None,
                mangabaka_name=None,
                image_url=None,
-               first_match_time=None) -> dict:
+               kavita_series_id=None,
+               library_id=None,
+               first_match_time=None,
+               last_update_time=None) -> dict:
+        """
+        Inserts or updates an entry.  Only fields passed explicitly are
+        modified; the rest are preserved.
+        """
         with self._lock:
             entry = self._data["matches"].get(title)
             if entry is None:
@@ -74,7 +95,10 @@ class MatchesCache:
                     "mangabakaId":    "",
                     "mangabakaName":  "",
                     "imageUrl":       "",
+                    "kavitaSeriesId": 0,
+                    "libraryId":      0,
                     "firstMatchTime": int(time.time()),
+                    "lastUpdateTime": 0,
                 }
                 self._data["matches"][title] = entry
             if mangabaka_id is not None:
@@ -83,14 +107,24 @@ class MatchesCache:
                 entry["mangabakaName"] = mangabaka_name
             if image_url is not None:
                 entry["imageUrl"] = image_url
+            if kavita_series_id is not None:
+                _set_int(entry, "kavitaSeriesId", kavita_series_id)
+            if library_id is not None:
+                _set_int(entry, "libraryId", library_id)
             if first_match_time is not None:
-                try:
-                    entry["firstMatchTime"] = int(first_match_time)
-                except (TypeError, ValueError):
-                    pass
+                _set_int(entry, "firstMatchTime", first_match_time)
+            if last_update_time is not None:
+                _set_int(entry, "lastUpdateTime", last_update_time)
             self._save_unlocked()
             return dict(entry)
 
+    def mark_updated(self, title: str) -> None:
+        with self._lock:
+            entry = self._data["matches"].get(title)
+            if entry is not None:
+                entry["lastUpdateTime"] = int(time.time())
+                self._save_unlocked()
+
     def rename(self, old_title: str, new_title: str) -> bool:
         if not new_title or old_title == new_title:
             return False
@@ -115,6 +149,20 @@ class MatchesCache:
             return {"matches": {k: dict(v)
                                 for k, v in self._data["matches"].items()}}
 
+    def all_in_libraries(self, library_ids: "list[int] | None") -> dict:
+        """
+        Returns the cache filtered to entries whose libraryId is in
+        `library_ids`.  Pass None to return everything.
+        """
+        if library_ids is None:
+            return self.all()
+        ids = {int(i) for i in library_ids}
+        with self._lock:
+            return {"matches": {
+                k: dict(v) for k, v in self._data["matches"].items()
+                if int(v.get("libraryId") or 0) in ids
+            }}
+
     # ------------------------------------------------------------------
     # Internal IO
     # ------------------------------------------------------------------

src/ln/MatchesWebApp.py

@@ -0,0 +1,764 @@
+"""
+matches_web_app.py
+==================
+
+Flask web UI for the Kavita light-novel metadata fetcher.
+
+Pages
+-----
+GET  /                          HTML UI (matches table + actions)
+
+Match cache (JSON)
+------------------
+GET  /api/libraries             Lists Kavita libraries
+GET  /api/matches               Full cache, optionally filtered by libraryIds=
+POST /api/matches               Upsert a single match
+                                  body: {title, mangabakaId}
+POST /api/matches/delete        Remove a match
+                                  body: {title}
+
+Background jobs
+---------------
+POST /api/build                 Build matches for libraries
+                                  body: {libraryIds: [int, ...]}
+POST /api/update                Update a single series
+                                  body: {kavitaSeriesId}
+POST /api/update-all            Update every cached series in libraries
+                                  body: {libraryIds: [int, ...] | null}
+GET  /api/status                Current background job status (status, log)
+"""
+
+from __future__ import annotations
+
+import threading
+import time
+
+from flask import Flask, jsonify, request, Response
+
+from MatchesCache import MatchesCache
+from LightNovelMetadataBuilder import pick_thumbnail_url
+
+
+def _int_list(values) -> list[int]:
+    """Coerces an iterable of mixed values to a list of positive ints."""
+    out: list[int] = []
+    for v in (values or []):
+        try:
+            n = int(v)
+        except (TypeError, ValueError):
+            continue
+        if n > 0:
+            out.append(n)
+    return out
+
+
+_INDEX_HTML = r"""<!doctype html>
+<html lang="en">
+<head>
+  <meta charset="utf-8">
+  <title>Kavita light-novel metadata fetcher</title>
+  <style>
+    body  { font-family: system-ui, sans-serif; margin: 1.5rem; background: #111; color: #eee; }
+    h1    { margin: 0 0 1rem; font-size: 1.4rem; }
+    .bar  { display: flex; gap: .5rem; align-items: center; margin-bottom: 1rem; flex-wrap: wrap; }
+    .bar input[type=search] { padding: .3rem .5rem; min-width: 18rem; background:#222; color:#eee; border:1px solid #444; }
+    .bar select[multiple] { background:#222; color:#eee; border:1px solid #444; min-width: 14rem; min-height: 4.2rem; }
+    button { padding: .35rem .7rem; cursor: pointer; background:#2a2a2a; color:#eee; border:1px solid #555; }
+    button.primary { background:#2563eb; border-color:#2563eb; color:white; }
+    button.danger  { background:#7f1d1d; border-color:#7f1d1d; color:white; }
+    button.success { background:#15803d; border-color:#15803d; color:white; }
+    button:disabled { opacity:.5; cursor:default; }
+    table { border-collapse: collapse; width: 100%; }
+    th, td { border: 1px solid #333; padding: .4rem .6rem; vertical-align: top; }
+    th { background: #1d1d1d; text-align: left; position: sticky; top: 0; }
+    th.sortable { cursor: pointer; user-select: none; }
+    th.sortable:hover { background:#252525; }
+    th .arrow { display:inline-block; width:.8em; color:#9ca3af; }
+    tr:nth-child(even) td { background: #161616; }
+    td.image img { max-width: 90px; max-height: 130px; display:block; }
+    td.id input { width: 12rem; padding: .25rem; background:#222; color:#eee; border:1px solid #444; font-family: monospace; }
+    td.title a { color: #60a5fa; text-decoration: none; }
+    td.title a:hover { text-decoration: underline; }
+    td.actions { white-space: nowrap; }
+    .status { margin-left: .5rem; color:#9ca3af; font-size: .9rem; }
+    .dirty td { background: #1f2937 !important; }
+    .count { color:#9ca3af; font-size:.9rem; margin-left:.5rem; }
+    pre.log { background:#0a0a0a; color:#9ca3af; padding:.5rem .75rem; max-height:18rem; overflow:auto; border:1px solid #333; font-size:.8rem; white-space:pre-wrap; }
+    label { font-size:.9rem; color:#9ca3af; }
+  </style>
+</head>
+<body>
+<h1>Kavita light-novel metadata fetcher <span id="count" class="count"></span></h1>
+
+<div class="bar">
+  <label>Libraries
+    <select id="libraries" multiple size="3"></select>
+  </label>
+  <button id="reload">Reload</button>
+  <button id="build">Match all in libraries</button>
+  <button id="updateAll" class="success">Update all in libraries</button>
+  <button id="batchSave" class="primary">Save dirty (0)</button>
+  <span class="status" id="status"></span>
+</div>
+
+<div class="bar">
+  <input id="filter" type="search" placeholder="Filter by title…">
+  <span class="count" id="jobStatus"></span>
+</div>
+
+<pre id="jobLog" class="log" hidden></pre>
+
+<table>
+  <thead>
+    <tr>
+      <th class="sortable" data-col="title">Title <span class="arrow" id="arrow-title"></span></th>
+      <th>mangabakaId</th>
+      <th>mangabakaName</th>
+      <th>library</th>
+      <th class="sortable" data-col="lastUpdateTime">Last update <span class="arrow" id="arrow-lastUpdateTime"></span></th>
+      <th>Image</th>
+      <th></th>
+    </tr>
+  </thead>
+  <tbody id="rows"></tbody>
+</table>
+
+<script>
+const MB_SEARCH = "https://mangabaka.org/search?q=";
+let matchesData = {};
+let librariesById = {};
+let currentSort = { col: "title", asc: true };
+let jobPollHandle = null;
+
+function fmtTime(unix) {
+  if (!unix) return "";
+  const d = new Date(unix * 1000);
+  return d.toLocaleString();
+}
+
+function setStatus(msg) { document.getElementById("status").textContent = msg; }
+
+function selectedLibraryIds() {
+  const sel = document.getElementById("libraries");
+  return Array.from(sel.selectedOptions).map(o => parseInt(o.value, 10));
+}
+
+function updateDirtyCount() {
+  const n = document.querySelectorAll("#rows tr.dirty").length;
+  const btn = document.getElementById("batchSave");
+  btn.textContent = "Save dirty (" + n + ")";
+  btn.disabled = n === 0;
+}
+
+function makeRow(title, e) {
+  const tr = document.createElement("tr");
+  tr.dataset.title = title;
+
+  // Title — links to MangaBaka search
+  const titleTd = document.createElement("td");
+  titleTd.className = "title";
+  const a = document.createElement("a");
+  a.href = MB_SEARCH + encodeURIComponent(title) + "&type=novel";
+  a.target = "_blank";
+  a.rel = "noopener";
+  a.textContent = title;
+  titleTd.appendChild(a);
+  tr.appendChild(titleTd);
+
+  // mangabakaId (editable)
+  const idTd = document.createElement("td");
+  idTd.className = "id";
+  const idInp = document.createElement("input");
+  idInp.value = e.mangabakaId || "";
+  idInp.dataset.original = e.mangabakaId || "";
+  idInp.addEventListener("input", () => {
+    if (idInp.value !== idInp.dataset.original) tr.classList.add("dirty");
+    else tr.classList.remove("dirty");
+    updateDirtyCount();
+  });
+  idTd.appendChild(idInp);
+  tr.appendChild(idTd);
+
+  // mangabakaName
+  const nameTd = document.createElement("td");
+  nameTd.textContent = e.mangabakaName || "";
+  tr.appendChild(nameTd);
+
+  // library
+  const libTd = document.createElement("td");
+  const libId = e.libraryId || 0;
+  libTd.textContent = librariesById[libId] || (libId ? "#" + libId : "");
+  tr.appendChild(libTd);
+
+  // lastUpdateTime
+  const timeTd = document.createElement("td");
+  timeTd.textContent = e.lastUpdateTime ? fmtTime(e.lastUpdateTime) : "";
+  tr.appendChild(timeTd);
+
+  // Image
+  const imgTd = document.createElement("td");
+  imgTd.className = "image";
+  const img = document.createElement("img");
+  img.src = e.imageUrl || "";
+  img.alt = "";
+  img.loading = "lazy";
+  imgTd.appendChild(img);
+  tr.appendChild(imgTd);
+
+  // Actions
+  const actTd = document.createElement("td");
+  actTd.className = "actions";
+
+  const save = document.createElement("button");
+  save.textContent = "Save";
+  save.className = "primary";
+  save.addEventListener("click", () => saveRow(tr));
+  actTd.appendChild(save);
+
+  const update = document.createElement("button");
+  update.textContent = "Update";
+  update.className = "success";
+  update.style.marginLeft = ".25rem";
+  update.disabled = !e.kavitaSeriesId;
+  update.title = e.kavitaSeriesId
+    ? "Push metadata to Kavita series #" + e.kavitaSeriesId
+    : "Run a Match cycle first so we know the Kavita series id";
+  update.addEventListener("click", () => updateRow(tr));
+  actTd.appendChild(update);
+
+  const del = document.createElement("button");
+  del.textContent = "Delete";
+  del.className = "danger";
+  del.style.marginLeft = ".25rem";
+  del.addEventListener("click", () => deleteRow(tr));
+  actTd.appendChild(del);
+
+  tr.appendChild(actTd);
+
+  tr._idInp = idInp;
+  tr._nameTd = nameTd;
+  tr._img = img;
+  tr._timeTd = timeTd;
+  tr._update = update;
+  return tr;
+}
+
+async function saveRow(tr) {
+  const title = tr.dataset.title;
+  const newId = tr._idInp.value.trim();
+  setStatus("Saving " + title + "…");
+  try {
+    const r = await fetch("/api/matches", {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ title: title, mangabakaId: newId }),
+    });
+    if (!r.ok) throw new Error(await r.text());
+    const data = await r.json();
+    const entry = data.entry || {};
+    matchesData[title] = entry;
+    tr._idInp.value = entry.mangabakaId || "";
+    tr._idInp.dataset.original = entry.mangabakaId || "";
+    tr._nameTd.textContent = entry.mangabakaName || "";
+    tr._img.src = entry.imageUrl || "";
+    tr.classList.remove("dirty");
+    updateDirtyCount();
+    setStatus("Saved " + title);
+    return true;
+  } catch (err) {
+    setStatus("Save failed (" + title + "): " + err.message);
+    return false;
+  }
+}
+
+async function deleteRow(tr) {
+  const title = tr.dataset.title;
+  if (!confirm("Delete " + title + "?")) return;
+  setStatus("Deleting " + title + "…");
+  try {
+    const r = await fetch("/api/matches/delete", {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ title: title }),
+    });
+    if (!r.ok) throw new Error(await r.text());
+    delete matchesData[title];
+    tr.remove();
+    document.getElementById("count").textContent =
+        "(" + Object.keys(matchesData).length + " entries)";
+    setStatus("Deleted");
+  } catch (err) {
+    setStatus("Delete failed: " + err.message);
+  }
+}
+
+async function updateRow(tr) {
+  const title = tr.dataset.title;
+  const entry = matchesData[title] || {};
+  if (!entry.kavitaSeriesId) {
+    setStatus("No kavitaSeriesId for " + title + " — run match first");
+    return;
+  }
+  setStatus("Updating " + title + "…");
+  tr._update.disabled = true;
+  try {
+    const r = await fetch("/api/update", {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ kavitaSeriesId: entry.kavitaSeriesId }),
+    });
+    if (!r.ok) throw new Error(await r.text());
+    const res = await r.json();
+    setStatus(res.ok ? "Updated " + title : "Update failed: " + res.error);
+    if (res.ok) {
+      entry.lastUpdateTime = Math.floor(Date.now() / 1000);
+      tr._timeTd.textContent = fmtTime(entry.lastUpdateTime);
+    }
+  } catch (err) {
+    setStatus("Update failed: " + err.message);
+  } finally {
+    tr._update.disabled = false;
+  }
+}
+
+async function batchSave() {
+  const dirty = Array.from(document.querySelectorAll("#rows tr.dirty"));
+  if (dirty.length === 0) return;
+  if (!confirm("Save " + dirty.length + " changed row(s)?")) return;
+  setStatus("Batch saving " + dirty.length + " rows…");
+  let ok = 0, fail = 0;
+  for (const tr of dirty) {
+    const success = await saveRow(tr);
+    if (success) ok++; else fail++;
+  }
+  setStatus("Batch: " + ok + " ok, " + fail + " failed");
+}
+
+function sortedTitles() {
+  const titles = Object.keys(matchesData);
+  const dir = currentSort.asc ? 1 : -1;
+  if (currentSort.col === "title") {
+    return titles.sort((a, b) => a.localeCompare(b) * dir);
+  }
+  if (currentSort.col === "lastUpdateTime") {
+    return titles.sort((a, b) => {
+      const av = matchesData[a].lastUpdateTime || 0;
+      const bv = matchesData[b].lastUpdateTime || 0;
+      return (av - bv) * dir;
+    });
+  }
+  return titles;
+}
+
+function updateSortArrows() {
+  for (const a of document.querySelectorAll("th .arrow")) a.textContent = "";
+  const id = "arrow-" + currentSort.col;
+  const el = document.getElementById(id);
+  if (el) el.textContent = currentSort.asc ? "▲" : "▼";
+}
+
+function applyFilter() {
+  const q = document.getElementById("filter").value.toLowerCase();
+  const libs = new Set(selectedLibraryIds());
+  for (const tr of document.querySelectorAll("#rows tr")) {
+    const title = tr.dataset.title;
+    const entry = matchesData[title] || {};
+    const titleMatch = title.toLowerCase().includes(q);
+    const libMatch = libs.size === 0 || libs.has(entry.libraryId || 0);
+    tr.style.display = (titleMatch && libMatch) ? "" : "none";
+  }
+}
+
+function render() {
+  const tbody = document.getElementById("rows");
+  tbody.innerHTML = "";
+  for (const t of sortedTitles()) {
+    tbody.appendChild(makeRow(t, matchesData[t]));
+  }
+  updateSortArrows();
+  applyFilter();
+  updateDirtyCount();
+  document.getElementById("count").textContent =
+      "(" + Object.keys(matchesData).length + " entries)";
+}
+
+async function loadLibraries() {
+  try {
+    const r = await fetch("/api/libraries");
+    const data = await r.json();
+    const libs = data.libraries || [];
+    const defaults = new Set(data.defaults || []);
+    librariesById = {};
+    const sel = document.getElementById("libraries");
+    sel.innerHTML = "";
+    for (const lib of libs) {
+      librariesById[lib.id] = lib.name;
+      const opt = document.createElement("option");
+      opt.value = lib.id;
+      opt.textContent = lib.name + " (#" + lib.id + ")";
+      if (defaults.has(lib.id)) opt.selected = true;
+      sel.appendChild(opt);
+    }
+  } catch (err) {
+    setStatus("Failed to load libraries: " + err.message);
+  }
+}
+
+async function load() {
+  setStatus("Loading…");
+  try {
+    const r = await fetch("/api/matches");
+    const data = await r.json();
+    matchesData = data.matches || {};
+    render();
+    setStatus(Object.keys(matchesData).length + " entries");
+  } catch (err) {
+    setStatus("Load failed: " + err.message);
+  }
+}
+
+async function pollJob() {
+  try {
+    const r = await fetch("/api/status");
+    const s = await r.json();
+    const jobStatus = document.getElementById("jobStatus");
+    const jobLog = document.getElementById("jobLog");
+    if (!s.running && !s.lastFinished) {
+      jobStatus.textContent = "";
+      jobLog.hidden = true;
+      stopPolling();
+      return;
+    }
+    jobLog.hidden = false;
+    jobLog.textContent = (s.log || []).join("\n");
+    jobLog.scrollTop = jobLog.scrollHeight;
+    if (s.running) {
+      jobStatus.textContent = "Running: " + (s.label || "");
+    } else {
+      jobStatus.textContent = "Done: " + (s.label || "");
+      stopPolling();
+      load();
+    }
+  } catch (err) {
+    /* keep polling silently */
+  }
+}
+
+function startPolling() {
+  if (jobPollHandle) return;
+  jobPollHandle = setInterval(pollJob, 1000);
+  pollJob();
+}
+
+function stopPolling() {
+  if (jobPollHandle) clearInterval(jobPollHandle);
+  jobPollHandle = null;
+}
+
+async function startBuild() {
+  const libs = selectedLibraryIds();
+  if (libs.length === 0) {
+    setStatus("Pick at least one library");
+    return;
+  }
+  if (!confirm("Match every series in " + libs.length + " library(ies)?")) return;
+  setStatus("Build started");
+  try {
+    const r = await fetch("/api/build", {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ libraryIds: libs }),
+    });
+    if (!r.ok) throw new Error(await r.text());
+    startPolling();
+  } catch (err) {
+    setStatus("Build failed: " + err.message);
+  }
+}
+
+async function startUpdateAll() {
+  const libs = selectedLibraryIds();
+  if (libs.length === 0) {
+    if (!confirm("No libraries selected — update every cached series?")) return;
+  } else if (!confirm("Update every cached series in " + libs.length + " library(ies)?")) {
+    return;
+  }
+  setStatus("Update-all started");
+  try {
+    const r = await fetch("/api/update-all", {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ libraryIds: libs.length ? libs : null }),
+    });
+    if (!r.ok) throw new Error(await r.text());
+    startPolling();
+  } catch (err) {
+    setStatus("Update-all failed: " + err.message);
+  }
+}
+
+document.getElementById("filter").addEventListener("input", applyFilter);
+document.getElementById("libraries").addEventListener("change", applyFilter);
+document.getElementById("reload").addEventListener("click", load);
+document.getElementById("batchSave").addEventListener("click", batchSave);
+document.getElementById("build").addEventListener("click", startBuild);
+document.getElementById("updateAll").addEventListener("click", startUpdateAll);
+for (const th of document.querySelectorAll("th.sortable")) {
+  th.addEventListener("click", () => {
+    const col = th.dataset.col;
+    if (currentSort.col === col) currentSort.asc = !currentSort.asc;
+    else { currentSort.col = col; currentSort.asc = true; }
+    render();
+  });
+}
+
+(async () => {
+  await loadLibraries();
+  await load();
+  // Resume polling if there's a job running from a previous session
+  pollJob();
+})();
+</script>
+</body>
+</html>
+"""
+
+
+class _JobState:
+    """Thread-safe container for the current background job's progress."""
+
+    def __init__(self):
+        self._lock = threading.Lock()
+        self._running = False
+        self._label = ""
+        self._log: list[str] = []
+        self._last_finished_at = 0
+        self._thread: "threading.Thread | None" = None
+
+    def start(self, label: str, target, *args, **kwargs) -> bool:
+        with self._lock:
+            if self._running:
+                return False
+            self._running = True
+            self._label = label
+            self._log = [f"[{time.strftime('%H:%M:%S')}] {label} started"]
+
+        def runner():
+            try:
+                target(self, *args, **kwargs)
+            except Exception as exc:
+                self.append(f"FATAL: {exc}")
+            finally:
+                with self._lock:
+                    self._running = False
+                    self._last_finished_at = int(time.time())
+                self.append(f"[{time.strftime('%H:%M:%S')}] finished")
+
+        self._thread = threading.Thread(target=runner,
+                                        name=f"job:{label}",
+                                        daemon=True)
+        self._thread.start()
+        return True
+
+    def append(self, line: str) -> None:
+        with self._lock:
+            self._log.append(line)
+            # Cap log length so the response stays bounded.
+            if len(self._log) > 1000:
+                self._log = self._log[-800:]
+
+    def snapshot(self) -> dict:
+        with self._lock:
+            return {
+                "running":      self._running,
+                "label":        self._label,
+                "log":          list(self._log),
+                "lastFinished": self._last_finished_at,
+            }
+
+
+class MatchesWebApp:
+    def __init__(self, cache: MatchesCache, *,
+                 orchestrator=None,
+                 default_library_ids: "list[int] | None" = None,
+                 host: str = "0.0.0.0",
+                 port: int = 8080):
+        self._cache = cache
+        self._orchestrator = orchestrator
+        self._defaults = list(default_library_ids or [])
+        self._host = host
+        self._port = port
+        self._job = _JobState()
+        self._app = Flask(__name__)
+        self._thread: "threading.Thread | None" = None
+        self._register_routes()
+
+    @property
+    def app(self) -> Flask:
+        return self._app
+
+    def start(self) -> threading.Thread:
+        if self._thread is not None and self._thread.is_alive():
+            return self._thread
+        self._thread = threading.Thread(
+            target=self._app.run,
+            kwargs={"host": self._host, "port": self._port,
+                    "debug": False, "use_reloader": False,
+                    "threaded": True},
+            name="MatchesWebApp",
+            daemon=False,
+        )
+        self._thread.start()
+        print(f"[MatchesWebApp] listening on {self._host}:{self._port}",
+              flush=True)
+        return self._thread
+
+    def wait(self) -> None:
+        if self._thread is not None:
+            self._thread.join()
+
+    # ------------------------------------------------------------------
+    # Routes
+    # ------------------------------------------------------------------
+    def _register_routes(self) -> None:
+        app = self._app
+        cache = self._cache
+
+        @app.get("/")
+        def index() -> Response:
+            return Response(_INDEX_HTML, mimetype="text/html; charset=utf-8")
+
+        @app.get("/api/libraries")
+        def api_libraries():
+            if self._orchestrator is None:
+                return jsonify([])
+            try:
+                libs = self._orchestrator.list_libraries()
+            except Exception as exc:
+                return Response(f"libraries failed: {exc}", status=502)
+            return jsonify({"libraries": libs, "defaults": self._defaults})
+
+        @app.get("/api/matches")
+        def api_list():
+            raw = request.args.get("libraryIds") or ""
+            lib_ids = _int_list(raw.split(","))
+            if lib_ids:
+                return jsonify(cache.all_in_libraries(lib_ids))
+            return jsonify(cache.all())
+
+        @app.post("/api/matches")
+        def api_upsert():
+            body = request.get_json(silent=True) or {}
+            title = (body.get("title") or "").strip()
+            if not title:
+                return Response("title is required", status=400)
+            new_id_raw = body.get("mangabakaId")
+            new_id = str(new_id_raw).strip() if new_id_raw is not None else ""
+            if not new_id:
+                return Response("mangabakaId is required", status=400)
+
+            new_name: "str | None" = None
+            new_image: "str | None" = None
+            if self._orchestrator is not None:
+                try:
+                    series = self._orchestrator.fetch_series(new_id)
+                except Exception as exc:
+                    return Response(f"resolve failed: {exc}", status=502)
+                if not series:
+                    return Response(
+                        f"MangaBaka has no series with id {new_id}",
+                        status=404)
+                new_name  = series.get("title") or ""
+                new_image = pick_thumbnail_url(series.get("cover")) or ""
+
+            entry = cache.upsert(
+                title,
+                mangabaka_id=new_id,
+                mangabaka_name=new_name,
+                image_url=new_image,
+            )
+            return jsonify({"title": title, "entry": entry})
+
+        @app.post("/api/matches/delete")
+        def api_delete():
+            body = request.get_json(silent=True) or {}
+            title = (body.get("title") or "").strip()
+            if not title:
+                return Response("title is required", status=400)
+            removed = cache.remove(title)
+            return jsonify({"removed": removed, "title": title})
+
+        @app.post("/api/build")
+        def api_build():
+            if self._orchestrator is None:
+                return Response("no orchestrator configured", status=503)
+            body = request.get_json(silent=True) or {}
+            library_ids = _int_list(body.get("libraryIds"))
+            if not library_ids:
+                return Response("libraryIds required", status=400)
+
+            label = f"match libraries {library_ids}"
+
+            def task(job: _JobState, lib_ids):
+                stats = self._orchestrator.build_matches(lib_ids)
+                job.append(f"matched={stats.get('matched')} "
+                           f"skipped={stats.get('skipped')} "
+                           f"missing={stats.get('missing')} "
+                           f"checked={stats.get('checked')}")
+
+            if not self._job.start(label, task, library_ids):
+                return Response("a job is already running", status=409)
+            return jsonify({"started": label})
+
+        @app.post("/api/update")
+        def api_update():
+            if self._orchestrator is None:
+                return Response("no orchestrator configured", status=503)
+            body = request.get_json(silent=True) or {}
+            ksid = body.get("kavitaSeriesId")
+            try:
+                ksid_int = int(ksid)
+            except (TypeError, ValueError):
+                return Response("kavitaSeriesId required", status=400)
+            try:
+                res = self._orchestrator.update_series(ksid_int)
+            except Exception as exc:
+                return Response(f"update failed: {exc}", status=500)
+            return jsonify(res)
+
+        @app.post("/api/update-all")
+        def api_update_all():
+            if self._orchestrator is None:
+                return Response("no orchestrator configured", status=503)
+            body = request.get_json(silent=True) or {}
+            raw = body.get("libraryIds")
+            library_ids = None if raw is None else _int_list(raw)
+
+            label = ("update all (every library)" if library_ids is None
+                     else f"update all in libraries {library_ids}")
+
+            def task(job: _JobState, lib_ids):
+                summary = self._orchestrator.update_all(lib_ids)
+                job.append(f"ok={summary.get('ok')} failed={summary.get('failed')}")
+                for res in summary.get("results", []):
+                    title = res.get("title", "?")
+                    if res.get("ok"):
+                        flags = []
+                        sr = res.get("series") or {}
+                        for k, v in sr.items():
+                            if v == "changed":
+                                flags.append(k)
+                        job.append(
+                            f"  {title}: changed=[{', '.join(flags) or '-'}]")
+                    else:
+                        job.append(f"  {title}: FAIL {res.get('error')}")
+
+            if not self._job.start(label, task, library_ids):
+                return Response("a job is already running", status=409)
+            return jsonify({"started": label})
+
+        @app.get("/api/status")
+        def api_status():
+            snap = self._job.snapshot()
+            snap["defaults"] = self._defaults
+            return jsonify(snap)

src/ln/RelationshipSync.py

@@ -0,0 +1,174 @@
+"""
+relationship_sync.py
+====================
+
+Mirrors MangaBaka's ``relationships_v2`` graph into Kavita:
+
+  1. Every related MangaBaka series that is *also* present in Kavita
+     (resolved via MatchesCache) is added to a shared Kavita collection
+     so the whole franchise can be browsed in one place.
+  2. Series-level relationships (prequel / sequel / spin-off / …) are
+     written via ``POST /api/Series/update-related`` so navigating
+     between entries surfaces the right neighbours.
+
+Only relationships where both endpoints exist in Kavita are written.
+Relationships pointing to series that have not been imported yet are
+silently skipped (the next match run picks them up).
+"""
+
+from __future__ import annotations
+
+from KavitaClient import KavitaClient
+from MatchesCache import MatchesCache
+
+
+# MangaBaka relation_type  ->  Kavita UpdateRelatedSeriesDto bucket
+_RELATION_MAP = {
+    "prequel":             "prequels",
+    "sequel":              "sequels",
+    "side_story":          "sideStories",
+    "spin_off":            "spinOffs",
+    "spinoff":             "spinOffs",
+    "alternative_version": "alternativeVersions",
+    "alternative_story":   "alternativeVersions",
+    "alternative_setting": "alternativeSettings",
+    "adapted_from":        "adaptations",
+    "adaptation":          "adaptations",
+    "doujinshi":           "doujinshis",
+    "parent":              "contains",   # the parent "contains" the child
+}
+
+_ALL_BUCKETS = (
+    "adaptations", "characters", "contains", "others",
+    "prequels", "sequels", "sideStories", "spinOffs",
+    "alternativeSettings", "alternativeVersions", "doujinshis",
+    "editions", "annuals",
+)
+
+
+class RelationshipSync:
+    def __init__(self, client: KavitaClient, cache: MatchesCache, *,
+                 builder=None):
+        """
+        Parameters
+        ----------
+        client  : KavitaClient for collection / relation writes.
+        cache   : MatchesCache to resolve mangabakaId -> kavitaSeriesId.
+        builder : optional LightNovelMetadataBuilder used to fetch parent
+                  series titles when picking the collection name.
+        """
+        self._client = client
+        self._cache = cache
+        self._builder = builder
+
+    # ------------------------------------------------------------------
+    # Public
+    # ------------------------------------------------------------------
+    def sync(self, kavita_series_id: int, built: dict) -> dict:
+        """
+        Applies the relationship and collection links described by
+        `built["relationships"]` (raw MangaBaka relationships_v2 list)
+        for the given Kavita series.  Returns a small status dict.
+        """
+        report: dict = {"relations": {}, "collection": None,
+                        "missing_series": []}
+
+        relationships = built.get("relationships") or []
+        if not relationships:
+            return report
+
+        # Resolve mangabakaId -> kavitaSeriesId for every related entry.
+        related: dict[str, list[int]] = {b: [] for b in _ALL_BUCKETS}
+        all_kavita_ids: set[int] = set()
+        for rel in relationships:
+            mb_id = rel.get("to_series_id")
+            if mb_id is None:
+                continue
+            hit = self._cache.get_by_mangabaka_id(mb_id)
+            if not hit:
+                report["missing_series"].append(int(mb_id))
+                continue
+            _title, entry = hit
+            ksid = int(entry.get("kavitaSeriesId") or 0)
+            if not ksid:
+                report["missing_series"].append(int(mb_id))
+                continue
+            bucket = _RELATION_MAP.get((rel.get("relation_type") or "").lower(),
+                                       "others")
+            if ksid not in related[bucket]:
+                related[bucket].append(ksid)
+            all_kavita_ids.add(ksid)
+
+        # ----- Relationships ------------------------------------------
+        if any(related.values()):
+            payload = {"seriesId": int(kavita_series_id)}
+            for bucket in _ALL_BUCKETS:
+                payload[bucket] = related[bucket]
+            try:
+                self._client.update_related(payload)
+                report["relations"] = {k: v for k, v in related.items() if v}
+            except Exception as exc:
+                report["relations"] = {"error": str(exc)}
+
+        # ----- Collection ---------------------------------------------
+        # Include the current series in the collection so it shows up too.
+        all_kavita_ids.add(int(kavita_series_id))
+        if len(all_kavita_ids) >= 2:
+            collection_name = self._collection_name(built, relationships)
+            collection_id = self._find_collection_id(collection_name)
+            try:
+                self._client.add_series_to_collection(
+                    collection_id=collection_id,
+                    title=collection_name,
+                    series_ids=sorted(all_kavita_ids),
+                )
+                report["collection"] = collection_name
+            except Exception as exc:
+                report["collection"] = f"error: {exc}"
+
+        return report
+
+    # ------------------------------------------------------------------
+    # Internal
+    # ------------------------------------------------------------------
+    def _find_collection_id(self, name: str) -> int:
+        """Returns the id of an existing collection by title, or 0 to create."""
+        if not name:
+            return 0
+        target = name.strip().lower()
+        try:
+            for col in self._client.list_collections():
+                if (col.get("title") or "").strip().lower() == target:
+                    try:
+                        return int(col.get("id") or 0)
+                    except (TypeError, ValueError):
+                        return 0
+        except Exception:
+            pass
+        return 0
+
+    def _collection_name(self, built: dict,
+                         relationships: list[dict]) -> str:
+        """
+        Picks the collection name.  Uses the parent series title from
+        MangaBaka if the current series has one; otherwise falls back to
+        the current series' own title.
+        """
+        for rel in relationships:
+            if (rel.get("relation_type") or "").lower() == "parent":
+                parent_id = rel.get("to_series_id")
+                if parent_id is not None and self._builder is not None:
+                    try:
+                        parent_md = self._builder.fetch_series(parent_id)
+                        if parent_md and parent_md.get("title"):
+                            return parent_md["title"]
+                    except Exception:
+                        pass
+                # Even without a builder, the cache may know the parent.
+                hit = self._cache.get_by_mangabaka_id(parent_id)
+                if hit:
+                    _title, entry = hit
+                    name = entry.get("mangabakaName")
+                    if name:
+                        return name
+        return built.get("mangabakaTitle") or ""

src/manga/ComicInfoBuilder.py

@@ -37,18 +37,26 @@ Data source notes
 
 from __future__ import annotations
 
-import difflib
 import re
+import sys
 import xml.etree.ElementTree as ET
 from pathlib import Path
 
 import requests
 
+# Shared modules live one level up (src/); needed when a module in this
+# folder is run directly as a script (the entry points set the path).
+if __name__ == "__main__":
+    sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
 from MangadexVolumeResolver import MangaDexVolumeResolver
-from MangaBakaWorksResolver import MangaBakaWorksResolver
+from MangaBakaWorksResolver import MangaBakaWorksResolver, _pick_image_url
 from MALResolver import MALResolver
 from AniListResolver import AniListResolver
 from MatchesCache import MatchesCache
+from MangaBakaRateLimit import apply_to_session as _apply_mangabaka_rate_limit
+from CoverCache import CoverCache, _IMAGE_EXTS
+from TextUtils import person_name_with_id
 
 try:
     from PIL import Image
@@ -60,7 +68,11 @@ except ImportError:
 # --------------------------------------------------------------------------
 # Constants
 # --------------------------------------------------------------------------
-_IMAGE_EXTS = {".jpg", ".jpeg", ".png", ".gif", ".webp", ".bmp", ".avif"}
+# Series types accepted by the MangaBaka search endpoint.  Light/web novels
+# are filtered out because this pipeline only handles image-based manga.
+# Passed to `requests` as a list so each value becomes its own `&type=...`
+# query parameter (MangaBaka's API expects repeated keys, not a CSV list).
+_SEARCH_TYPES = ["manga", "manhwa", "manhua"]
 
 _AGE_RATING_MAP = {
     "safe": "Everyone",
@@ -172,7 +184,8 @@ class ComicInfoBuilder:
                  works_resolver: "MangaBakaWorksResolver | None" = None,
                  mal_resolver: "MALResolver | None" = None,
                  al_resolver: "AniListResolver | None" = None,
-                 matches_cache: "MatchesCache | None" = None):
+                 matches_cache: "MatchesCache | None" = None,
+                 cover_cache: "CoverCache | None" = None):
         if not manga_title or not str(manga_title).strip():
             raise ValueError("manga_title must not be empty.")
 
@@ -184,6 +197,9 @@ class ComicInfoBuilder:
         self.request_timeout = request_timeout
         self._session = session or requests.Session()
         self._session.headers.setdefault("User-Agent", "ComicInfoBuilder/1.0")
+        # Throttle every call to api.mangabaka.dev (idempotent — safe even
+        # when the session was already prepared by a parent class).
+        _apply_mangabaka_rate_limit(self._session)
 
         self._volume_resolver = (volume_resolver
                                  or MangaDexVolumeResolver(
@@ -200,6 +216,7 @@ class ComicInfoBuilder:
         self._al_resolver = al_resolver or AniListResolver(
             request_timeout=request_timeout)
         self._matches_cache = matches_cache
+        self._cover_cache = cover_cache or _default_cover_cache()
 
         self._metadata: "dict | None" = None
         self._pages: list[dict] = []
@@ -378,7 +395,8 @@ class ComicInfoBuilder:
 
         url = f"{self.api_base_url}/series/search"
         resp = self._session.get(
-            url, params={"q": title, "page": 1, "limit": 1},
+            url, params={"q": title, "type": _SEARCH_TYPES,
+                         "page": 1, "limit": 1},
             timeout=self.request_timeout)
         resp.raise_for_status()
         data = resp.json().get("data") or []
@@ -389,7 +407,7 @@ class ComicInfoBuilder:
                 title,
                 mangabaka_id=series.get("id"),
                 mangabaka_name=series.get("title") or "",
-                image_url=_pick_cover_url(series.get("cover")),
+                image_url=_pick_thumbnail_url(series.get("cover")),
             )
 
         return series
@@ -428,8 +446,7 @@ class ComicInfoBuilder:
         # ----- Title / Series -----------------------------------------------
         add("Title", sd.get("Title") or f"Chapter {self._chapter}")
         add("Series", md.get("title") or self._manga_title)
-        add("LocalizedSeries",
-            md.get("native_title") or md.get("romanized_title"))
+        add("LocalizedSeries", self._romanized_for_native(md))
         add("SeriesSort", self._get_sort_title(md))
         add("Number", sd.get("Number") or self._chapter)
         add("Count", md.get("total_chapters"))
@@ -473,9 +490,19 @@ class ComicInfoBuilder:
         add("Tags",  ", ".join(_format_term(t) for t in (md.get("tags") or [])))
 
         # ----- Characters — MAL first, AniList fallback ---------------------
-        characters = self._mal_resolver.get_characters(mal_id)
-        if not characters and al_id:
-            characters = self._al_resolver.get_characters(al_id)
+        # Names are disambiguated with the tracker *character* id
+        # ("Rem (MAL 118737)") so same-named characters from different
+        # series stay separate Kavita person records.  The format is shared
+        # with the light-novel updater — see TextUtils.person_name_with_id.
+        char_entries = self._mal_resolver.get_characters_detailed(mal_id)
+        if not char_entries and al_id:
+            char_entries = self._al_resolver.get_characters_detailed(al_id)
+        characters = [
+            person_name_with_id(e.get("name"),
+                                mal_id=e.get("mal_id"),
+                                al_id=e.get("al_id"))
+            for e in char_entries if (e.get("name") or "").strip()
+        ]
         add("Characters", ", ".join(characters) if characters else None)
 
         # ----- Web links ----------------------------------------------------
@@ -570,11 +597,13 @@ class ComicInfoBuilder:
     # ======================================================================
     def _download_cover(self, folder: Path, cover_filename: str) -> "Path | None":
         """
-        Downloads the cover for the current chapter/volume.
+        Fetches the cover for the current chapter/volume and writes it into
+        `folder`.
 
-        If a volume is known and a volume-specific cover exists in MangaBaka
-        works, that cover is used.  Otherwise the series default cover is
-        downloaded (raw variant preferred).
+        If a volume is known and a volume-specific cover exists in MangaBaka,
+        that cover is used; otherwise the series default cover.  The image
+        itself comes from the CoverCache, so a cover shared by many chapters
+        is downloaded only once.
         """
         md = self._get_metadata()
         volume = self._determine_volume()
@@ -592,18 +621,13 @@ class ComicInfoBuilder:
         if not cover_url:
             cover_url = _pick_cover_url(md.get("cover"))
 
-        if not cover_url:
+        fetched = self._cover_cache.get(cover_url) if cover_url else None
+        if not fetched:
             return None
 
-        try:
-            resp = self._session.get(cover_url, timeout=self.request_timeout)
-            resp.raise_for_status()
-        except requests.RequestException:
-            return None
-
-        ext = _guess_extension(cover_url, resp.headers.get("Content-Type", ""))
+        data, ext = fetched
         target = folder / f"{cover_filename}{ext}"
-        target.write_bytes(resp.content)
+        target.write_bytes(data)
         return target
 
     # ======================================================================
@@ -637,6 +661,82 @@ class ComicInfoBuilder:
     # ======================================================================
     # Title helpers
     # ======================================================================
+    # Mapping from series type to the matching romanized language code(s)
+    # in the MangaBaka titles array.  Used to pick the correct romaji /
+    # romaja / pinyin for LocalizedSeries.
+    _ROMANIZED_LANG_BY_TYPE = {
+        "manga":  ("ja-latn", "ja-romaji"),
+        "manhwa": ("ko-latn", "ko-romaji"),
+        "manhua": ("zh-latn",),
+    }
+
+    @staticmethod
+    def _pick_best_title(titles, language_codes: tuple,
+                         prefer_trait: "str | None" = None) -> "str | None":
+        """
+        Picks the highest-scoring entry from a MangaBaka `titles` list for
+        any of the given language codes.
+
+        Scoring: preferred trait (+4) > "official" trait (+2) > is_primary
+        (+1); first seen wins on ties.  Returns None when no entry matches.
+        """
+        if not isinstance(titles, list):
+            return None
+        best_score = -1
+        best_title: "str | None" = None
+        for entry in titles:
+            if not isinstance(entry, dict):
+                continue
+            lang = (entry.get("language") or entry.get("lang") or "").lower()
+            if lang not in language_codes:
+                continue
+            title = entry.get("title")
+            if not title:
+                continue
+            traits = entry.get("traits") or []
+            score = 0
+            if prefer_trait and prefer_trait in traits:
+                score += 4
+            if "official" in traits:
+                score += 2
+            if entry.get("is_primary"):
+                score += 1
+            if score > best_score:
+                best_score, best_title = score, title
+        return best_title
+
+    @classmethod
+    def _romanized_for_native(cls, md: dict) -> "str | None":
+        """
+        Picks the romanized title in the manga's original language from the
+        ``titles`` array.
+
+        The series' original language is inferred from ``type``::
+
+            manga  -> ja-Latn  (Japanese romaji)
+            manhwa -> ko-Latn  (Korean romaja)
+            manhua -> zh-Latn  (Chinese pinyin)
+
+        Among multiple entries for the matching language, the one with the
+        highest "quality score" wins (``official`` trait > ``is_primary`` >
+        first seen).
+
+        The root-level ``romanized_title`` field is **deliberately not used
+        as a fallback** — MangaBaka frequently stores a different language's
+        romanization there (e.g. Korean romaja on a Japanese manga), which
+        is exactly what this function is meant to avoid.
+
+        Returns ``None`` when no romanized title is available for the
+        inferred language.
+        """
+        mtype = (md.get("type") or "").lower()
+        langs = cls._ROMANIZED_LANG_BY_TYPE.get(mtype)
+        if not langs:
+            return None
+
+        titles = md.get("titles") or md.get("alt_titles") or []
+        return cls._pick_best_title(titles, langs)
+
     def _get_sort_title(self, md: dict) -> "str | None":
         """
         Returns the SeriesSort title in the configured language.
@@ -671,31 +771,7 @@ class ComicInfoBuilder:
 
         def pick(language_codes: tuple, prefer_trait: "str | None" = None
                  ) -> "str | None":
-            """Picks the best title entry for any of the given language codes."""
-            if not isinstance(titles, list):
-                return None
-            best_score = -1
-            best_title: "str | None" = None
-            for entry in titles:
-                if not isinstance(entry, dict):
-                    continue
-                lang = (entry.get("language") or entry.get("lang") or "").lower()
-                if lang not in language_codes:
-                    continue
-                title = entry.get("title")
-                if not title:
-                    continue
-                traits = entry.get("traits") or []
-                score = 0
-                if prefer_trait and prefer_trait in traits:
-                    score += 4
-                if "official" in traits:
-                    score += 2
-                if entry.get("is_primary"):
-                    score += 1
-                if score > best_score:
-                    best_score, best_title = score, title
-            return best_title
+            return self._pick_best_title(titles, language_codes, prefer_trait)
 
         result: dict[str, str] = {}
 
@@ -956,12 +1032,14 @@ class ComicInfoBuilder:
         return unique
 
     @staticmethod
-    def _read_existing_comicinfo(folder: Path) -> dict:
-        xml_path = folder / "ComicInfo.xml"
-        if not xml_path.is_file():
-            return {}
+    def read_comicinfo_fields(xml_source) -> dict:
+        """
+        Parses ComicInfo.xml content (bytes or str) and returns the fields
+        relevant as supplementary Suwayomi data.  Returns {} on parse errors.
+        Reusable for XML read directly from a CBZ archive (no extraction).
+        """
         try:
-            root = ET.parse(xml_path).getroot()
+            root = ET.fromstring(xml_source)
         except ET.ParseError:
             return {}
 
@@ -975,6 +1053,16 @@ class ComicInfoBuilder:
                 data[tag] = child.text.strip()
         return data
 
+    @staticmethod
+    def _read_existing_comicinfo(folder: Path) -> dict:
+        xml_path = folder / "ComicInfo.xml"
+        if not xml_path.is_file():
+            return {}
+        try:
+            return ComicInfoBuilder.read_comicinfo_fields(xml_path.read_bytes())
+        except OSError:
+            return {}
+
     @staticmethod
     def _image_dimensions(path: Path):
         if not _HAS_PIL:
@@ -987,22 +1075,37 @@ class ComicInfoBuilder:
 
 
 # --------------------------------------------------------------------------
-# Module-level helpers (shared with MangaBakaWorksResolver logic)
+# Module-level helpers
 # --------------------------------------------------------------------------
-def _pick_cover_url(cover) -> "str | None":
+
+# Alias: _pick_image_url (from MangaBakaWorksResolver) is the canonical
+# generic image-block picker; _pick_cover_url is kept for backward compat.
+_pick_cover_url = _pick_image_url
+
+# Shared fallback CoverCache for builders constructed without an explicit
+# one (temporary directory, removed at process exit).  Created lazily so
+# importing this module never touches the filesystem.
+_shared_cover_cache: "CoverCache | None" = None
+
+
+def _default_cover_cache() -> CoverCache:
+    global _shared_cover_cache
+    if _shared_cover_cache is None:
+        _shared_cover_cache = CoverCache()
+    return _shared_cover_cache
+
+
+def _pick_thumbnail_url(cover) -> "str | None":
     """
-    Selects the best cover URL from a MangaBaka cover object.
+    Picks a small cover variant suitable for a UI thumbnail.
 
-    Real API shape (from `GET /v1/series/{id}` and `/works`):
-        {
-          "raw":  {"url": "...", "size": ..., "height": ..., "width": ...},
-          "x150": {"x1": "...", "x2": "...", "x3": "..."},
-          "x250": {"x1": "...", "x2": "...", "x3": "..."},
-          "x350": {"x1": "...", "x2": "...", "x3": "..."}
-        }
+    Order of preference: x150@x2 > x150@x1 > x150@x3 > x250@x2 > x250@x1 >
+    x250@x3 > x350@x2 > x350@x1 > x350@x3 > raw. x150@x2 is roughly 300px
+    wide — sharp on HiDPI displays at the ~90px thumbnail size used in
+    the matches table, while still being a fraction of the raw image
+    weight (often 50KB vs. several MB).
 
-    Order of preference: raw original  >  x350@x3  >  x250@x3  >  x150@x3
-    (falling through to lower densities and sizes as needed).
+    Falls back to `_pick_cover_url` if no thumbnail variant is available.
     """
     if not cover:
         return None
@@ -1011,46 +1114,17 @@ def _pick_cover_url(cover) -> "str | None":
     if not isinstance(cover, dict):
         return None
 
-    # 1) Preferred: the unscaled "raw" image
-    raw = cover.get("raw")
-    if isinstance(raw, dict):
-        url = raw.get("url")
-        if isinstance(url, str) and url:
-            return url
-    elif isinstance(raw, str) and raw:
-        return raw
-
-    # 2) Fallback: size-keyed variants, largest first, highest density first
-    for size_key in ("x350", "x250", "x150"):
+    for size_key in ("x150", "x250", "x350"):
         variant = cover.get(size_key)
         if isinstance(variant, dict):
-            for density in ("x3", "x2", "x1"):
+            for density in ("x2", "x1", "x3"):
                 url = variant.get(density)
                 if isinstance(url, str) and url:
                     return url
         elif isinstance(variant, str) and variant:
             return variant
 
-    # 3) Last-ditch fallback: any http URL anywhere in the structure
-    for val in cover.values():
-        if isinstance(val, str) and val.startswith("http"):
-            return val
-        if isinstance(val, dict):
-            for sub in val.values():
-                if isinstance(sub, str) and sub.startswith("http"):
-                    return sub
-    return None
-
-
-def _guess_extension(url: str, content_type: str) -> str:
-    url_ext = Path(url.split("?")[0]).suffix.lower()
-    if url_ext in _IMAGE_EXTS:
-        return url_ext
-    ct = (content_type or "").lower()
-    if "png"  in ct: return ".png"
-    if "webp" in ct: return ".webp"
-    if "gif"  in ct: return ".gif"
-    return ".jpg"
+    return _pick_cover_url(cover)
 
 
 # --------------------------------------------------------------------------

src/manga/KavitaVolumeCoverUpdater.py

@@ -0,0 +1,554 @@
+"""
+kavita_volume_cover_updater.py
+==============================
+
+Periodically re-checks chapters already moved to the Kavita library whose
+volume could not be resolved at move time (``"volume": null`` in the
+series' ``chapter_index.json``).
+
+When MangaDex has since assigned the chapter to a volume, the updater:
+
+  1. writes the volume into ``chapter_index.json``,
+  2. updates ``<Volume>`` inside the chapter's ComicInfo.xml (in-archive),
+  3. downloads the MangaBaka volume cover and swaps it in for the
+     placeholder ``000.<ext>`` series cover, and
+  4. refreshes the *first* chapter's ComicInfo.xml with full metadata —
+     Kavita can be configured to take series metadata from the lowest
+     chapter, so it must reflect the latest state.
+
+Host-IO policy
+--------------
+* Per series only ``chapter_index.json`` is read (no archive is opened to
+  discover its contents).
+* Series without null-volume chapters are skipped before any API call.
+* An archive is read+rewritten exactly once per update (single pass,
+  written to a ``.tmp`` file, then atomically replaced).
+
+Every updated chapter is appended to a log file (one line per update).
+
+Reused components
+-----------------
+* ``SuwayomiMover``            — chapter index helpers, dirname sanitizer
+* ``ComicInfoBuilder``         — metadata fetch (matches-cache ID lookup),
+                                 chapter→volume resolution, XML build
+* ``MangaBakaWorksResolver``   — volume covers (/images with /works fallback)
+* ``MangaDexVolumeResolver``   — chapter→volume aggregate (shared cache)
+* ``MangaBakaRateLimit``       — process-wide API throttle
+
+Dependencies
+------------
+    requests    ->  pip install requests
+    Pillow      ->  pip install pillow   (optional, page-0 dimensions)
+"""
+
+from __future__ import annotations
+
+import io
+import sys
+import threading
+import xml.etree.ElementTree as ET
+import zipfile
+from datetime import datetime
+from pathlib import Path
+
+import requests
+
+# Shared modules live one level up (src/); needed when a module in this
+# folder is run directly as a script (the entry points set the path).
+if __name__ == "__main__":
+    sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+from ComicInfoBuilder import ComicInfoBuilder
+from MangadexVolumeResolver import MangaDexVolumeResolver
+from MangaBakaWorksResolver import MangaBakaWorksResolver
+from MALResolver import MALResolver
+from AniListResolver import AniListResolver
+from MatchesCache import MatchesCache
+from SuwayomiMover import (_load_chapter_index, _save_chapter_index,
+                           _sanitize_dirname, _normalise_volume_value)
+from MangaBakaRateLimit import apply_to_session as _apply_mangabaka_rate_limit
+from CronSchedule import CronSchedule
+from CoverCache import CoverCache, _IMAGE_EXTS
+
+try:
+    from PIL import Image
+    _HAS_PIL = True
+except ImportError:
+    _HAS_PIL = False
+
+
+def _now() -> str:
+    return datetime.now().isoformat(timespec="seconds")
+
+
+def _image_dims_from_bytes(data: bytes) -> tuple:
+    """Returns (width, height) of an image byte blob, or (None, None)."""
+    if not _HAS_PIL:
+        return (None, None)
+    try:
+        with Image.open(io.BytesIO(data)) as im:
+            return im.size
+    except Exception:
+        return (None, None)
+
+
+def _chapter_sort_value(num: str) -> float:
+    try:
+        return float(num)
+    except (TypeError, ValueError):
+        return float("inf")
+
+
+def _update_page0_attrs(pages_el: "ET.Element", cover_bytes: bytes) -> None:
+    """Refreshes size/dimension attributes of the FrontCover page entry."""
+    for page in pages_el:
+        if page.get("Image") == "0":
+            page.set("ImageSize", str(len(cover_bytes)))
+            width, height = _image_dims_from_bytes(cover_bytes)
+            if width and height:
+                page.set("ImageWidth", str(width))
+                page.set("ImageHeight", str(height))
+            return
+
+
+def _serialize_tree(root: "ET.Element") -> str:
+    tree = ET.ElementTree(root)
+    try:
+        ET.indent(tree, space="  ")
+    except AttributeError:
+        pass
+    return ('<?xml version="1.0" encoding="UTF-8"?>\n'
+            + ET.tostring(root, encoding="unicode"))
+
+
+class KavitaVolumeCoverUpdater:
+    """
+    Scans the Kavita library for chapters whose volume was unknown at move
+    time and back-fills volume + volume cover once MangaDex / MangaBaka
+    provide the data.  Runs periodically on a background thread.
+
+    Parameters
+    ----------
+    kavita_path      : Root of the Kavita library (series folders inside).
+    matches_cache    : MatchesCache — provides the MangaBaka series ID per
+                       series (mandatory; folders without a match are skipped).
+    language         : ComicInfo language (passed to ComicInfoBuilder).
+    request_timeout  : HTTP timeout in seconds.
+    log_path         : File that receives one line per updated chapter.
+                       Default: <kavita_path>/volume_updater.log
+    schedule         : Cron expression (5 fields) defining when scans run,
+                       e.g. "0 19 * * 1,4" = 19:00 every Monday and
+                       Thursday.  Evaluated in local time — set the TZ env
+                       var inside Docker.  Default: "0 19 * * 1,4".
+    cover_cache_dir  : Directory for the persistent cover cache.  None ->
+                       temporary cache, deleted at process exit.
+    """
+
+    def __init__(self,
+                 kavita_path,
+                 *,
+                 matches_cache: MatchesCache,
+                 language: str = "en",
+                 request_timeout: int = 30,
+                 api_base_url: str = "https://api.mangabaka.dev/v1",
+                 log_path=None,
+                 schedule: str = "0 19 * * 1,4",
+                 cover_cache_dir=None):
+        self._dst = Path(kavita_path)
+        self._matches_cache = matches_cache
+        self._language = language
+        self._timeout = request_timeout
+        self._api_base_url = api_base_url.rstrip("/")
+        self._log_path = (Path(log_path) if log_path
+                          else self._dst / "volume_updater.log")
+        self._cron = CronSchedule(schedule)
+
+        session = requests.Session()
+        session.headers.setdefault("User-Agent", "KavitaVolumeCoverUpdater/1.0")
+        _apply_mangabaka_rate_limit(session)
+        self._session = session
+
+        self._mal = MALResolver(request_timeout=request_timeout)
+        self._al  = AniListResolver(request_timeout=request_timeout)
+        self._vol_resolver = MangaDexVolumeResolver(
+            request_timeout=request_timeout, session=session)
+        self._works_resolver = MangaBakaWorksResolver(
+            api_base_url=api_base_url,
+            request_timeout=request_timeout, session=session)
+        self._cover_cache = CoverCache(
+            cover_cache_dir, session=session, request_timeout=request_timeout)
+
+        self._stop = threading.Event()
+        self._thread: "threading.Thread | None" = None
+
+    # ------------------------------------------------------------------
+    # Cron API (mirrors SuwayomiFolderWatcher)
+    # ------------------------------------------------------------------
+    def start(self) -> None:
+        """Starts the periodic scan thread.  Non-blocking."""
+        if self._thread is not None and self._thread.is_alive():
+            return
+        self._stop.clear()
+        self._thread = threading.Thread(
+            target=self._loop, name="KavitaVolumeCoverUpdater", daemon=True)
+        self._thread.start()
+        print(f"[{_now()}] [updater] scanning {self._dst} "
+              f"on cron '{self._cron.expression}'", flush=True)
+
+    def stop(self) -> None:
+        """Stops the scan thread (current scan finishes its series first)."""
+        self._stop.set()
+        if self._thread is not None:
+            self._thread.join(timeout=10)
+
+    def wait(self) -> None:
+        """Blocks the calling thread until stop() is invoked."""
+        self._stop.wait()
+
+    def _loop(self) -> None:
+        while not self._stop.is_set():
+            next_run = self._cron.next_after(datetime.now())
+            wait = max(0.0, (next_run - datetime.now()).total_seconds())
+            print(f"[{_now()}] [updater] next scheduled scan: "
+                  f"{next_run.isoformat(timespec='minutes')}", flush=True)
+            if self._stop.wait(wait):
+                break
+
+            try:
+                summary = self.update_all()
+                print(f"[{_now()}] [updater] scan done: "
+                      f"{summary['series_updated']} series / "
+                      f"{summary['chapters_updated']} chapters updated",
+                      flush=True)
+            except Exception as exc:
+                print(f"[{_now()}] [updater] scan ERROR: {exc}", flush=True)
+
+    # ------------------------------------------------------------------
+    # Public scan API
+    # ------------------------------------------------------------------
+    def update_all(self) -> dict:
+        """
+        Scans every series folder under the Kavita root once.
+        Returns {"series_scanned": n, "series_updated": n, "chapters_updated": n}.
+        """
+        summary = {"series_scanned": 0, "series_updated": 0,
+                   "chapters_updated": 0}
+        if not self._dst.is_dir():
+            print(f"[updater] kavita path missing: {self._dst}", flush=True)
+            return summary
+
+        # The whole point of a scan is detecting volume assignments added
+        # since the previous run — start from fresh API data, not the
+        # process-lifetime resolver caches.
+        self._vol_resolver.clear_cache()
+        self._works_resolver.clear_cache()
+
+        for series_dir in sorted(self._dst.iterdir()):
+            if self._stop.is_set():
+                break
+            if not series_dir.is_dir():
+                continue
+            summary["series_scanned"] += 1
+            try:
+                updated = self.update_series(series_dir)
+            except Exception as exc:
+                print(f"[updater] {series_dir.name}: ERROR {exc}", flush=True)
+                continue
+            if updated:
+                summary["series_updated"] += 1
+                summary["chapters_updated"] += updated
+        return summary
+
+    def update_series(self, series_dir: Path) -> int:
+        """
+        Updates one series folder.  Returns the number of updated chapters.
+
+        Only chapters listed in chapter_index.json with ``"volume": null``
+        are candidates; everything else costs no further host reads.
+        """
+        index = _load_chapter_index(series_dir)
+        chapters: dict = index["chapter"]
+        if not chapters:
+            return 0
+
+        missing = [num for num, e in chapters.items()
+                   if isinstance(e, dict) and e.get("volume") is None]
+        if not missing:
+            return 0
+
+        match_key, match = self._find_match_for_folder(series_dir.name)
+        if not match or not match.get("mangabakaId"):
+            print(f"[updater] {series_dir.name}: no matches.json entry — skip",
+                  flush=True)
+            return 0
+
+        # Builder resolves metadata via the cached MangaBaka ID and gives us
+        # the exact same chapter→volume logic the mover uses.
+        builder = ComicInfoBuilder(
+            match_key, chapter=missing[0],
+            api_base_url=self._api_base_url,
+            language=self._language,
+            request_timeout=self._timeout,
+            session=self._session,
+            volume_resolver=self._vol_resolver,
+            works_resolver=self._works_resolver,
+            mal_resolver=self._mal,
+            al_resolver=self._al,
+            matches_cache=self._matches_cache,
+            cover_cache=self._cover_cache,
+        )
+        md = builder.fetch_metadata()
+        series_id = str(md.get("id") or "")
+
+        # Resolve volumes for all null-volume chapters first (API only).
+        updates: dict[str, dict] = {}   # num -> {"volume": str, "cover": tuple|None}
+        for num in sorted(missing, key=_chapter_sort_value):
+            builder.chapter = num
+            try:
+                volume = builder._determine_volume()
+            except Exception:
+                volume = None
+            if not volume:
+                continue
+            updates[num] = {"volume": volume,
+                            "cover": self._fetch_cover(series_id, volume)}
+
+        if not updates:
+            return 0
+
+        first = min(chapters, key=_chapter_sort_value)
+        updated = 0
+
+        for num, up in updates.items():
+            entry = chapters[num]
+            cbz = series_dir / (entry.get("archiveName") or "")
+            if not entry.get("archiveName") or not cbz.is_file():
+                print(f"[updater] {series_dir.name} ch.{num}: archive missing "
+                      f"({entry.get('archiveName')!r}) — skip", flush=True)
+                continue
+            # The first chapter gets a full metadata rebuild (Kavita reads
+            # series metadata from it); other chapters only a volume edit.
+            ok, cover_swapped = self._apply_update(
+                cbz, builder, num,
+                volume=up["volume"], cover=up["cover"],
+                full_rebuild=(num == first))
+            if not ok:
+                continue
+            entry["volume"] = _normalise_volume_value(up["volume"])
+            updated += 1
+            self._log(f"{series_dir.name} | chapter {num} -> volume "
+                      f"{up['volume']} | cover "
+                      f"{'replaced' if cover_swapped else 'kept'} | {cbz.name}")
+
+        # Refresh the first chapter's metadata when any other chapter changed
+        # (skip when it was already fully rebuilt in the loop above).
+        if updated and first not in updates:
+            first_entry = chapters.get(first) or {}
+            cbz = series_dir / (first_entry.get("archiveName") or "")
+            if first_entry.get("archiveName") and cbz.is_file():
+                ok, _ = self._apply_update(
+                    cbz, builder, first,
+                    volume=None, cover=None, full_rebuild=True)
+                if ok:
+                    self._log(f"{series_dir.name} | chapter {first} | "
+                              f"first-chapter metadata refreshed | {cbz.name}")
+
+        if updated:
+            _save_chapter_index(series_dir, index)
+        return updated
+
+    # ------------------------------------------------------------------
+    # Matching Kavita folder -> matches.json entry
+    # ------------------------------------------------------------------
+    def _find_match_for_folder(self, folder_name: str) -> tuple:
+        """
+        Maps a Kavita series folder back to its matches.json entry.
+
+        The folder was created as ``_sanitize_dirname(mangabaka_title)``, so
+        the comparison sanitizes each entry's mangabakaName the same way.
+        Falls back to the folderTitle (Suwayomi name) for robustness.
+        Returns (match_key, entry) or (None, None).
+        """
+        target = folder_name.strip().casefold()
+        matches = self._matches_cache.all()["matches"]
+        for key, entry in matches.items():
+            name = entry.get("mangabakaName") or ""
+            if name and _sanitize_dirname(name).strip().casefold() == target:
+                return key, entry
+        for key, entry in matches.items():
+            folder = entry.get("folderTitle") or key
+            if _sanitize_dirname(folder).strip().casefold() == target:
+                return key, entry
+        return None, None
+
+    # ------------------------------------------------------------------
+    # Cover download
+    # ------------------------------------------------------------------
+    def _fetch_cover(self, series_id: str, volume) -> "tuple[str, bytes] | None":
+        """
+        Fetches the MangaBaka volume cover via the CoverCache (one download
+        per unique URL, even across chapters sharing a volume).
+        Returns ("000<ext>", bytes) or None when no cover is available.
+        """
+        try:
+            url = self._works_resolver.get_cover_for_volume(series_id, volume)
+        except Exception:
+            url = None
+        if not url:
+            return None
+        fetched = self._cover_cache.get(url)
+        if not fetched:
+            return None
+        data, ext = fetched
+        return (f"000{ext}", data)
+
+    # ------------------------------------------------------------------
+    # Archive update (single read + single write per archive)
+    # ------------------------------------------------------------------
+    def _apply_update(self, cbz_path: Path, builder: ComicInfoBuilder,
+                      chapter_num: str, *,
+                      volume, cover, full_rebuild: bool) -> tuple:
+        """
+        Rewrites one CBZ archive with an updated ComicInfo.xml and (when
+        provided and a placeholder exists) a new cover image.
+
+        Returns (ok, cover_swapped).
+        """
+        try:
+            with zipfile.ZipFile(cbz_path, "r") as zin:
+                try:
+                    old_xml = zin.read("ComicInfo.xml")
+                except KeyError:
+                    old_xml = None
+
+                if full_rebuild or old_xml is None:
+                    new_xml = self._build_full_xml(
+                        builder, chapter_num, old_xml, cover)
+                else:
+                    new_xml = self._edit_volume_xml(old_xml, volume, cover)
+                    if new_xml is None:           # parse error -> full rebuild
+                        new_xml = self._build_full_xml(
+                            builder, chapter_num, None, cover)
+
+                infos = zin.infolist()
+                # Cover is only ever *replaced*: inserting one would shift
+                # every <Pages> image index in the existing XML.
+                has_placeholder = any(
+                    Path(i.filename).stem == "000"
+                    and Path(i.filename).suffix.lower() in _IMAGE_EXTS
+                    for i in infos)
+                swap_cover = cover is not None and has_placeholder
+
+                tmp = cbz_path.with_suffix(cbz_path.suffix + ".tmp")
+                wrote_xml = False
+                with zipfile.ZipFile(tmp, "w", zipfile.ZIP_STORED) as zout:
+                    for info in infos:
+                        p = Path(info.filename)
+                        if (swap_cover and p.stem == "000"
+                                and p.suffix.lower() in _IMAGE_EXTS):
+                            zout.writestr(cover[0], cover[1])
+                        elif info.filename == "ComicInfo.xml":
+                            zout.writestr("ComicInfo.xml", new_xml)
+                            wrote_xml = True
+                        else:
+                            zout.writestr(info, zin.read(info.filename))
+                    if not wrote_xml:
+                        zout.writestr("ComicInfo.xml", new_xml)
+            tmp.replace(cbz_path)
+            return True, swap_cover
+        except Exception as exc:
+            print(f"[updater] {cbz_path.name}: update failed: {exc}",
+                  flush=True)
+            return False, False
+
+    # ------------------------------------------------------------------
+    # XML builders
+    # ------------------------------------------------------------------
+    def _edit_volume_xml(self, old_xml: bytes, volume,
+                         cover) -> "str | None":
+        """
+        Sets <Volume> in an existing ComicInfo.xml and refreshes the
+        FrontCover page attributes when the cover gets replaced.
+        Returns None when the XML is unparseable.
+        """
+        try:
+            root = ET.fromstring(old_xml)
+        except ET.ParseError:
+            return None
+
+        el = root.find("Volume")
+        if el is None:
+            el = ET.SubElement(root, "Volume")
+        el.text = str(volume)
+
+        if cover is not None:
+            pages = root.find("Pages")
+            if pages is not None:
+                _update_page0_attrs(pages, cover[1])
+
+        return _serialize_tree(root)
+
+    def _build_full_xml(self, builder: ComicInfoBuilder, chapter_num: str,
+                        old_xml: "bytes | None", cover) -> str:
+        """
+        Rebuilds the complete ComicInfo.xml via ComicInfoBuilder (fresh
+        MangaBaka/MAL metadata).  Suwayomi-derived fields and the <Pages>
+        section are carried over from the previous XML.
+        """
+        builder.chapter = chapter_num   # also clears builder page state
+        builder._suwayomi_data = (
+            ComicInfoBuilder.read_comicinfo_fields(old_xml) if old_xml else {})
+        root = builder._build_tree().getroot()
+
+        if old_xml:
+            try:
+                old_root = ET.fromstring(old_xml)
+            except ET.ParseError:
+                old_root = None
+            if old_root is not None:
+                pages = old_root.find("Pages")
+                if pages is not None and cover is not None:
+                    _update_page0_attrs(pages, cover[1])
+                page_count = old_root.find("PageCount")
+                if page_count is not None:
+                    root.append(page_count)
+                if pages is not None:
+                    root.append(pages)
+
+        return _serialize_tree(root)
+
+    # ------------------------------------------------------------------
+    # Logging
+    # ------------------------------------------------------------------
+    def _log(self, msg: str) -> None:
+        line = f"[{_now()}] {msg}"
+        print(f"[updater] {msg}", flush=True)
+        try:
+            self._log_path.parent.mkdir(parents=True, exist_ok=True)
+            with self._log_path.open("a", encoding="utf-8") as f:
+                f.write(line + "\n")
+        except OSError as exc:
+            print(f"[updater] cannot write log file {self._log_path}: {exc}",
+                  flush=True)
+
+
+# --------------------------------------------------------------------------
+# Usage example
+# --------------------------------------------------------------------------
+if __name__ == "__main__":
+    # Local (no-Docker) smoke test.  Adjust paths to your environment.
+    KAVITA_PATH  = r"\\192.168.2.2\root\ServerData\Kavita\test"
+    MATCHES_PATH = Path(__file__).resolve().parent.parent / "matches.json"
+
+    updater = KavitaVolumeCoverUpdater(
+        KAVITA_PATH,
+        matches_cache=MatchesCache(MATCHES_PATH),
+    )
+
+    # One-shot scan (no cron thread):
+    summary = updater.update_all()
+    print(f"\n[updater] {summary}")
+
+    # Or run on the cron schedule (default: 19:00 every Mon + Thu):
+    # updater.start()
+    # updater.wait()

src/manga/MangadexVolumeResolver.py

@@ -43,7 +43,6 @@ Dependencies
 from __future__ import annotations
 
 import difflib
-import re
 
 import requests
 

src/manga/MatchesCache.py

@@ -0,0 +1,218 @@
+"""
+matches_cache.py
+================
+
+Persistent JSON cache that maps a normalised (lowercase) search title to the
+MangaBaka series it was matched against.
+
+Structure on disk::
+
+    {
+      "matches": {
+        "<normalised lowercase key>": {
+          "folderTitle":    "Original Folder Name",
+          "mangabakaId":    "12345",
+          "mangabakaName":  "One-Punch Man",
+          "imageUrl":       "https://.../cover.jpg",
+          "firstMatchTime": 1700000000
+        },
+        ...
+      }
+    }
+
+Keys are always stored lowercase so that folder names differing only in
+capitalisation (e.g. "[Oshi No Ko]" vs "[oshi no ko]") are treated as
+identical entries.  The original casing is preserved in the ``folderTitle``
+field and is used for display purposes (e.g. the web UI title link).
+
+The cache is consulted by ComicInfoBuilder before issuing a MangaBaka
+search request, and is written back to disk on every mutation so a crash
+does not lose matches that were resolved in the current run.
+"""
+
+from __future__ import annotations
+
+import json
+import threading
+import time
+from pathlib import Path
+
+
+def _norm_key(title: str) -> str:
+    """Normalises a cache key to lowercase for case-insensitive deduplication."""
+    return title.lower()
+
+
+class MatchesCache:
+    def __init__(self, path):
+        self._path = Path(path)
+        self._lock = threading.RLock()
+        self._data: dict = {"matches": {}}
+        self._load()
+
+    # ------------------------------------------------------------------
+    # Public lookup / mutation API
+    # ------------------------------------------------------------------
+    def get(self, title: str) -> "dict | None":
+        with self._lock:
+            entry = self._data["matches"].get(_norm_key(title))
+            return dict(entry) if entry else None
+
+    def add(self, title: str, *,
+            mangabaka_id,
+            mangabaka_name: str,
+            image_url: "str | None") -> dict:
+        entry = {
+            "folderTitle":    title,
+            "mangabakaId":    str(mangabaka_id) if mangabaka_id is not None else "",
+            "mangabakaName":  mangabaka_name or "",
+            "imageUrl":       image_url or "",
+            "firstMatchTime": int(time.time()),
+        }
+        with self._lock:
+            self._data["matches"][_norm_key(title)] = entry
+            self._save_unlocked()
+        return dict(entry)
+
+    def upsert(self, title: str, *,
+               mangabaka_id=None,
+               mangabaka_name=None,
+               image_url=None,
+               first_match_time=None) -> dict:
+        norm = _norm_key(title)
+        with self._lock:
+            entry = self._data["matches"].get(norm)
+            if entry is None:
+                entry = {
+                    "folderTitle":    title,
+                    "mangabakaId":    "",
+                    "mangabakaName":  "",
+                    "imageUrl":       "",
+                    "firstMatchTime": int(time.time()),
+                }
+                self._data["matches"][norm] = entry
+            # folderTitle is only set on creation; preserve original casing on updates.
+            if mangabaka_id is not None:
+                entry["mangabakaId"] = str(mangabaka_id)
+            if mangabaka_name is not None:
+                entry["mangabakaName"] = mangabaka_name
+            if image_url is not None:
+                entry["imageUrl"] = image_url
+            if first_match_time is not None:
+                try:
+                    entry["firstMatchTime"] = int(first_match_time)
+                except (TypeError, ValueError):
+                    pass
+            self._save_unlocked()
+            return dict(entry)
+
+    def rename(self, old_title: str, new_title: str) -> bool:
+        old_norm = _norm_key(old_title)
+        new_norm = _norm_key(new_title)
+        if not new_title or old_norm == new_norm:
+            return False
+        with self._lock:
+            entry = self._data["matches"].pop(old_norm, None)
+            if entry is None:
+                return False
+            entry["folderTitle"] = new_title
+            self._data["matches"][new_norm] = entry
+            self._save_unlocked()
+            return True
+
+    def remove(self, title: str) -> bool:
+        norm = _norm_key(title)
+        with self._lock:
+            existed = norm in self._data["matches"]
+            if existed:
+                del self._data["matches"][norm]
+                self._save_unlocked()
+            return existed
+
+    def all(self) -> dict:
+        with self._lock:
+            return {"matches": {k: dict(v)
+                                for k, v in self._data["matches"].items()}}
+
+    # ------------------------------------------------------------------
+    # Internal IO
+    # ------------------------------------------------------------------
+    def _load(self) -> None:
+        if not self._path.is_file():
+            return
+        try:
+            with self._path.open("r", encoding="utf-8") as f:
+                loaded = json.load(f)
+        except (OSError, json.JSONDecodeError) as exc:
+            print(f"[MatchesCache] failed to load {self._path}: {exc}",
+                  flush=True)
+            return
+        if not isinstance(loaded, dict) or not isinstance(loaded.get("matches"), dict):
+            return
+
+        normalized, changed = self._normalize_on_load(loaded["matches"])
+        loaded["matches"] = normalized
+        self._data = loaded
+        if changed:
+            print(f"[MatchesCache] migrated {changed} entr{'y' if changed == 1 else 'ies'} "
+                  f"(lowercase keys / folderTitle), saving", flush=True)
+            self._save_unlocked()
+
+    @staticmethod
+    def _normalize_on_load(raw: dict) -> "tuple[dict, int]":
+        """
+        Normalises the raw matches dict loaded from disk.
+
+        - Keys are lowercased.
+        - ``folderTitle`` is added from the original key when missing.
+        - Duplicate keys (same normalised form) are merged by keeping the
+          entry with the higher ``firstMatchTime``.
+
+        Returns (normalised_dict, number_of_changed_entries).
+        """
+        result: dict = {}
+        changed = 0
+
+        for orig_key, entry in raw.items():
+            if not isinstance(entry, dict):
+                continue
+            norm = _norm_key(orig_key)
+            entry = dict(entry)
+
+            # Add folderTitle if absent
+            if "folderTitle" not in entry:
+                entry["folderTitle"] = orig_key
+                changed += 1
+
+            if norm != orig_key:
+                changed += 1
+
+            # Merge duplicates: keep data from the more recent entry, but
+            # prefer the folderTitle that contains uppercase letters (= the
+            # original folder name) regardless of which entry is newer.
+            if norm in result:
+                existing = result[norm]
+                if entry.get("firstMatchTime", 0) > existing.get("firstMatchTime", 0):
+                    # Newer entry wins for data; preserve better-cased folderTitle
+                    existing_ft = existing.get("folderTitle", norm)
+                    new_ft      = entry.get("folderTitle", norm)
+                    if existing_ft != existing_ft.lower() and new_ft == new_ft.lower():
+                        entry["folderTitle"] = existing_ft
+                    result[norm] = entry
+                else:
+                    # Existing entry stays; but adopt new folderTitle if it has casing
+                    existing_ft = existing.get("folderTitle", norm)
+                    new_ft      = entry.get("folderTitle", norm)
+                    if new_ft != new_ft.lower() and existing_ft == existing_ft.lower():
+                        existing["folderTitle"] = new_ft
+            else:
+                result[norm] = entry
+
+        return result, changed
+
+    def _save_unlocked(self) -> None:
+        self._path.parent.mkdir(parents=True, exist_ok=True)
+        tmp = self._path.with_suffix(self._path.suffix + ".tmp")
+        with tmp.open("w", encoding="utf-8") as f:
+            json.dump(self._data, f, ensure_ascii=False, indent=2)
+        tmp.replace(self._path)

src/manga/MatchesWebApp.py

@@ -0,0 +1,500 @@
+"""
+matches_web_app.py
+==================
+
+Flask web UI for inspecting and editing the matches.json file produced by
+MatchesCache.
+
+Routes
+------
+GET  /                       HTML table view (one row per cached match)
+GET  /api/matches            JSON dump of the full cache
+POST /api/matches            Update an entry's mangabakaId
+                             body: {title, mangabakaId}
+                             Server resolves the id against MangaBaka and
+                             refreshes the mangabakaName + imageUrl fields.
+POST /api/matches/delete     Remove an entry          body: {title}
+POST /api/build              Trigger a full re-scan via
+                             SuwayomiMover.build_matches_only
+
+The Title cell is rendered as a link to MangaBaka's search page restricted
+to the manga / manhwa / manhua types. Only mangabakaId is editable; title
+(folder name) and mangabakaName (info only) are read-only.
+"""
+
+from __future__ import annotations
+
+import threading
+
+from flask import Flask, jsonify, request, Response
+
+from MatchesCache import MatchesCache
+from ComicInfoBuilder import _pick_thumbnail_url
+
+
+_INDEX_HTML = """<!doctype html>
+<html lang="en">
+<head>
+  <meta charset="utf-8">
+  <title>MangaBaka matches</title>
+  <style>
+    body  { font-family: system-ui, sans-serif; margin: 1.5rem; background: #111; color: #eee; }
+    h1    { margin: 0 0 1rem; font-size: 1.4rem; }
+    .bar  { display: flex; gap: .5rem; align-items: center; margin-bottom: 1rem; flex-wrap: wrap; }
+    .bar input[type=search] { padding: .3rem .5rem; min-width: 18rem; background:#222; color:#eee; border:1px solid #444; }
+    button { padding: .35rem .7rem; cursor: pointer; background:#2a2a2a; color:#eee; border:1px solid #555; }
+    button.primary { background:#2563eb; border-color:#2563eb; color:white; }
+    button.danger  { background:#7f1d1d; border-color:#7f1d1d; color:white; }
+    button:disabled { opacity:.5; cursor:default; }
+    table { border-collapse: collapse; width: 100%; }
+    th, td { border: 1px solid #333; padding: .4rem .6rem; vertical-align: top; }
+    th { background: #1d1d1d; text-align: left; position: sticky; top: 0; }
+    th.sortable { cursor: pointer; user-select: none; }
+    th.sortable:hover { background:#252525; }
+    th .arrow { display:inline-block; width:.8em; color:#9ca3af; }
+    tr:nth-child(even) td { background: #161616; }
+    td.image img { max-width: 90px; max-height: 130px; display:block; }
+    td.id input { width: 14rem; padding: .25rem; background:#222; color:#eee; border:1px solid #444; font-family: monospace; }
+    td.title a { color: #60a5fa; text-decoration: none; }
+    td.title a:hover { text-decoration: underline; }
+    td.actions { white-space: nowrap; }
+    .status { margin-left: .5rem; color:#9ca3af; font-size: .9rem; }
+    .dirty td { background: #1f2937 !important; }
+    .count { color:#9ca3af; font-size:.9rem; margin-left:.5rem; }
+  </style>
+</head>
+<body>
+<h1>MangaBaka matches <span id="count" class="count"></span></h1>
+<div class="bar">
+  <input id="filter" type="search" placeholder="Filter by title…">
+  <button id="reload">Reload</button>
+  <button id="batchSave" class="primary">Save dirty (0)</button>
+  <button id="build">Build all (rescan)</button>
+  <button id="move">Start move</button>
+  <span class="status" id="status"></span>
+</div>
+
+<table>
+  <thead>
+    <tr>
+      <th class="sortable" data-col="title">Title <span class="arrow" id="arrow-title"></span></th>
+      <th>mangabakaId</th>
+      <th>mangabakaName</th>
+      <th class="sortable" data-col="firstMatchTime">firstMatchTime <span class="arrow" id="arrow-firstMatchTime"></span></th>
+      <th>Image</th>
+      <th></th>
+    </tr>
+  </thead>
+  <tbody id="rows"></tbody>
+</table>
+
+<script>
+const TYPES = "&type=manhwa&type=manhua&type=manga";
+let matchesData = {};
+let currentSort = { col: "title", asc: true };
+
+function fmtTime(unix) {
+  if (!unix) return "";
+  const d = new Date(unix * 1000);
+  return d.toLocaleString();
+}
+
+function searchUrl(title) {
+  return "https://mangabaka.org/search?q=" + encodeURIComponent(title) + TYPES;
+}
+
+function setStatus(msg) { document.getElementById("status").textContent = msg; }
+
+function updateDirtyCount() {
+  const n = document.querySelectorAll("#rows tr.dirty").length;
+  const btn = document.getElementById("batchSave");
+  btn.textContent = "Save dirty (" + n + ")";
+  btn.disabled = n === 0;
+}
+
+function makeRow(title, e) {
+  const tr = document.createElement("tr");
+  tr.dataset.title = title;
+  const displayTitle = e.folderTitle || title;
+  tr.dataset.folderTitle = displayTitle;
+
+  // Title — link only, not editable; shows folderTitle (original casing)
+  const titleTd = document.createElement("td");
+  titleTd.className = "title";
+  const titleLink = document.createElement("a");
+  titleLink.href = searchUrl(displayTitle);
+  titleLink.target = "_blank";
+  titleLink.rel = "noopener";
+  titleLink.textContent = displayTitle;
+  titleTd.appendChild(titleLink);
+  tr.appendChild(titleTd);
+
+  // mangabakaId — editable
+  const idTd = document.createElement("td");
+  idTd.className = "id";
+  const idInp = document.createElement("input");
+  idInp.value = e.mangabakaId || "";
+  idInp.dataset.original = e.mangabakaId || "";
+  idInp.addEventListener("input", () => {
+    if (idInp.value !== idInp.dataset.original) tr.classList.add("dirty");
+    else tr.classList.remove("dirty");
+    updateDirtyCount();
+  });
+  idTd.appendChild(idInp);
+  tr.appendChild(idTd);
+
+  // mangabakaName — plain text (info only)
+  const nameTd = document.createElement("td");
+  nameTd.className = "name";
+  nameTd.textContent = e.mangabakaName || "";
+  tr.appendChild(nameTd);
+
+  // firstMatchTime — plain text
+  const timeTd = document.createElement("td");
+  timeTd.textContent = fmtTime(e.firstMatchTime);
+  tr.appendChild(timeTd);
+
+  // Image
+  const imgTd = document.createElement("td");
+  imgTd.className = "image";
+  const img = document.createElement("img");
+  img.src = e.imageUrl || "";
+  img.alt = "";
+  img.loading = "lazy";
+  imgTd.appendChild(img);
+  tr.appendChild(imgTd);
+
+  // Actions
+  const actTd = document.createElement("td");
+  actTd.className = "actions";
+  const save = document.createElement("button");
+  save.textContent = "Save";
+  save.className = "primary";
+  save.addEventListener("click", () => saveRow(tr));
+  const del = document.createElement("button");
+  del.textContent = "Delete";
+  del.className = "danger";
+  del.style.marginLeft = ".25rem";
+  del.addEventListener("click", () => deleteRow(tr));
+  actTd.append(save, del);
+  tr.appendChild(actTd);
+
+  tr._idInp = idInp;
+  tr._nameTd = nameTd;
+  tr._img = img;
+  return tr;
+}
+
+async function saveRow(tr) {
+  const title = tr.dataset.title;
+  const newId = tr._idInp.value.trim();
+  setStatus("Saving " + (tr.dataset.folderTitle || title) + "…");
+  try {
+    const r = await fetch("/api/matches", {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ title: title, mangabakaId: newId }),
+    });
+    if (!r.ok) throw new Error(await r.text());
+    const data = await r.json();
+    const entry = data.entry || {};
+    matchesData[title] = entry;
+    tr._idInp.value = entry.mangabakaId || "";
+    tr._idInp.dataset.original = entry.mangabakaId || "";
+    tr._nameTd.textContent = entry.mangabakaName || "";
+    tr._img.src = entry.imageUrl || "";
+    tr.classList.remove("dirty");
+    updateDirtyCount();
+    setStatus("Saved " + (tr.dataset.folderTitle || title));
+    return true;
+  } catch (err) {
+    setStatus("Save failed (" + title + "): " + err.message);
+    return false;
+  }
+}
+
+async function deleteRow(tr) {
+  const title = tr.dataset.title;
+  if (!confirm("Delete " + title + "?")) return;
+  setStatus("Deleting " + title + "…");
+  try {
+    const r = await fetch("/api/matches/delete", {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ title: title }),
+    });
+    if (!r.ok) throw new Error(await r.text());
+    delete matchesData[title];
+    tr.remove();
+    updateDirtyCount();
+    document.getElementById("count").textContent =
+        "(" + Object.keys(matchesData).length + " entries)";
+    setStatus("Deleted");
+  } catch (err) {
+    setStatus("Delete failed: " + err.message);
+  }
+}
+
+async function batchSave() {
+  const dirty = Array.from(document.querySelectorAll("#rows tr.dirty"));
+  if (dirty.length === 0) return;
+  if (!confirm("Save " + dirty.length + " changed row(s)?")) return;
+  setStatus("Batch saving " + dirty.length + " rows…");
+  let ok = 0, fail = 0;
+  for (const tr of dirty) {
+    const success = await saveRow(tr);
+    if (success) ok++; else fail++;
+  }
+  setStatus("Batch: " + ok + " ok, " + fail + " failed");
+}
+
+function sortedTitles() {
+  const titles = Object.keys(matchesData);
+  const dir = currentSort.asc ? 1 : -1;
+  if (currentSort.col === "title") {
+    return titles.sort((a, b) => {
+      const fa = (matchesData[a].folderTitle || a).toLowerCase();
+      const fb = (matchesData[b].folderTitle || b).toLowerCase();
+      return fa.localeCompare(fb) * dir;
+    });
+  }
+  if (currentSort.col === "firstMatchTime") {
+    return titles.sort((a, b) => {
+      const av = matchesData[a].firstMatchTime || 0;
+      const bv = matchesData[b].firstMatchTime || 0;
+      return (av - bv) * dir;
+    });
+  }
+  return titles;
+}
+
+function updateSortArrows() {
+  for (const a of document.querySelectorAll("th .arrow")) a.textContent = "";
+  const id = "arrow-" + currentSort.col;
+  const el = document.getElementById(id);
+  if (el) el.textContent = currentSort.asc ? "▲" : "▼";
+}
+
+function render() {
+  const tbody = document.getElementById("rows");
+  tbody.innerHTML = "";
+  for (const t of sortedTitles()) {
+    tbody.appendChild(makeRow(t, matchesData[t]));
+  }
+  updateSortArrows();
+  applyFilter();
+  updateDirtyCount();
+  document.getElementById("count").textContent =
+      "(" + Object.keys(matchesData).length + " entries)";
+}
+
+async function load() {
+  setStatus("Loading…");
+  try {
+    const r = await fetch("/api/matches");
+    const data = await r.json();
+    matchesData = data.matches || {};
+    render();
+    setStatus(Object.keys(matchesData).length + " entries");
+  } catch (err) {
+    setStatus("Load failed: " + err.message);
+  }
+}
+
+function applyFilter() {
+  const q = document.getElementById("filter").value.toLowerCase();
+  for (const tr of document.querySelectorAll("#rows tr")) {
+    const t = (tr.dataset.folderTitle || tr.dataset.title).toLowerCase();
+    tr.style.display = t.includes(q) ? "" : "none";
+  }
+}
+
+document.getElementById("filter").addEventListener("input", applyFilter);
+document.getElementById("reload").addEventListener("click", load);
+document.getElementById("batchSave").addEventListener("click", batchSave);
+document.getElementById("build").addEventListener("click", async () => {
+  if (!confirm("Run full scan? This may take several minutes.")) return;
+  setStatus("Building… (running on the server)");
+  try {
+    const r = await fetch("/api/build", { method: "POST" });
+    if (!r.ok) throw new Error(await r.text());
+    setStatus("Build finished");
+    load();
+  } catch (err) {
+    setStatus("Build failed: " + err.message);
+  }
+});
+document.getElementById("move").addEventListener("click", async () => {
+  if (!confirm("Start move operation? This will process all series and may take a long time.")) return;
+  const btn = document.getElementById("move");
+  btn.disabled = true;
+  setStatus("Moving… (running on the server)");
+  try {
+    const r = await fetch("/api/move", { method: "POST" });
+    if (!r.ok) throw new Error(await r.text());
+    const data = await r.json();
+    const total = Object.keys(data.results || {}).length;
+    setStatus("Move finished — " + total + " series processed");
+  } catch (err) {
+    setStatus("Move failed: " + err.message);
+  } finally {
+    btn.disabled = false;
+  }
+});
+for (const th of document.querySelectorAll("th.sortable")) {
+  th.addEventListener("click", () => {
+    const col = th.dataset.col;
+    if (currentSort.col === col) currentSort.asc = !currentSort.asc;
+    else { currentSort.col = col; currentSort.asc = true; }
+    render();
+  });
+}
+
+load();
+</script>
+</body>
+</html>
+"""
+
+
+class MatchesWebApp:
+    """
+    Flask app exposing the MatchesCache. `mover` is required when you want
+    POST /api/matches to resolve a new mangabakaId against MangaBaka (it
+    uses the mover's rate-limited session) and when POST /api/build should
+    work.
+    """
+
+    def __init__(self, cache: MatchesCache, *,
+                 mover=None,
+                 host: str = "0.0.0.0",
+                 port: int = 8080):
+        self._cache = cache
+        self._mover = mover
+        self._host = host
+        self._port = port
+        self._build_lock = threading.Lock()
+        self._move_lock  = threading.Lock()
+        self._app = Flask(__name__)
+        self._thread: "threading.Thread | None" = None
+        self._register_routes()
+
+    @property
+    def app(self) -> Flask:
+        return self._app
+
+    def start(self) -> threading.Thread:
+        """
+        Starts the Flask server on a background thread and returns it.
+
+        The thread is non-daemon so the process stays alive even when the
+        caller does not explicitly join() — important when this is the
+        only foreground task (e.g. watcher disabled for testing).
+        """
+        if self._thread is not None and self._thread.is_alive():
+            return self._thread
+        self._thread = threading.Thread(
+            target=self._app.run,
+            kwargs={"host": self._host, "port": self._port,
+                    "debug": False, "use_reloader": False,
+                    "threaded": True},
+            name="MatchesWebApp",
+            daemon=False,
+        )
+        self._thread.start()
+        print(f"[MatchesWebApp] listening on {self._host}:{self._port}",
+              flush=True)
+        return self._thread
+
+    def wait(self) -> None:
+        """Blocks until the Flask thread exits."""
+        if self._thread is not None:
+            self._thread.join()
+
+    # ------------------------------------------------------------------
+    # Routes
+    # ------------------------------------------------------------------
+    def _register_routes(self) -> None:
+        app = self._app
+        cache = self._cache
+
+        @app.get("/")
+        def index() -> Response:
+            return Response(_INDEX_HTML, mimetype="text/html; charset=utf-8")
+
+        @app.get("/api/matches")
+        def api_list():
+            return jsonify(cache.all())
+
+        @app.post("/api/matches")
+        def api_upsert():
+            body = request.get_json(silent=True) or {}
+            title = (body.get("title") or "").strip()
+            if not title:
+                return Response("title is required", status=400)
+
+            new_id_raw = body.get("mangabakaId")
+            new_id = str(new_id_raw).strip() if new_id_raw is not None else ""
+            if not new_id:
+                return Response("mangabakaId is required", status=400)
+
+            # Resolve the id against MangaBaka so mangabakaName + imageUrl
+            # always reflect what the id actually points to.
+            new_name: "str | None" = None
+            new_image: "str | None" = None
+            if self._mover is not None:
+                try:
+                    series = self._mover.fetch_series(new_id)
+                except Exception as exc:
+                    return Response(f"resolve failed: {exc}", status=502)
+                if not series:
+                    return Response(
+                        f"MangaBaka has no series with id {new_id}",
+                        status=404)
+                new_name  = series.get("title") or ""
+                new_image = _pick_thumbnail_url(series.get("cover")) or ""
+
+            entry = cache.upsert(
+                title,
+                mangabaka_id=new_id,
+                mangabaka_name=new_name,
+                image_url=new_image,
+            )
+            return jsonify({"title": title, "entry": entry})
+
+        @app.post("/api/matches/delete")
+        def api_delete():
+            body = request.get_json(silent=True) or {}
+            title = (body.get("title") or "").strip()
+            if not title:
+                return Response("title is required", status=400)
+            removed = cache.remove(title)
+            return jsonify({"removed": removed, "title": title})
+
+        @app.post("/api/build")
+        def api_build():
+            if self._mover is None:
+                return Response("no mover configured", status=503)
+            if not self._build_lock.acquire(blocking=False):
+                return Response("build already running", status=409)
+            try:
+                result = self._mover.build_matches_only()
+            except Exception as exc:
+                return Response(f"build failed: {exc}", status=500)
+            finally:
+                self._build_lock.release()
+            return jsonify(result)
+
+        @app.post("/api/move")
+        def api_move():
+            if self._mover is None:
+                return Response("no mover configured", status=503)
+            if not self._move_lock.acquire(blocking=False):
+                return Response("move already running", status=409)
+            try:
+                results = self._mover.process_all()
+            except Exception as exc:
+                return Response(f"move failed: {exc}", status=500)
+            finally:
+                self._move_lock.release()
+            return jsonify({"results": results})

src/manga/SuwayomiFolderWatcher.py

@@ -29,7 +29,6 @@ from __future__ import annotations
 
 import queue
 import threading
-import time
 from datetime import datetime
 from pathlib import Path
 

src/manga/SuwayomiMover.py

@@ -43,26 +43,92 @@ Dependencies
 
 from __future__ import annotations
 
+import json
 import re
 import shutil
+import sys
 import xml.etree.ElementTree as ET
 import zipfile
 from pathlib import Path
 
 import requests
 
-from ComicInfoBuilder import ComicInfoBuilder, _pick_cover_url
+# Shared modules live one level up (src/); needed when a module in this
+# folder is run directly as a script (the entry points set the path).
+if __name__ == "__main__":
+    sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+from ComicInfoBuilder import (ComicInfoBuilder, _pick_thumbnail_url,
+                              _SEARCH_TYPES, _natural_key)
 from MangadexVolumeResolver import MangaDexVolumeResolver
 from MangaBakaWorksResolver import MangaBakaWorksResolver
 from MALResolver import MALResolver
 from AniListResolver import AniListResolver
+from KavitaClient import KavitaClient
 from KavitaPersonUpdater import KavitaPersonUpdater
 from MatchesCache import MatchesCache
+from MangaBakaRateLimit import apply_to_session as _apply_mangabaka_rate_limit
+from CoverCache import CoverCache, _IMAGE_EXTS
 
 
-_IMAGE_EXTS = {".jpg", ".jpeg", ".png", ".gif", ".webp", ".bmp", ".avif"}
 _CHAPTER_RE = re.compile(r'[Cc]hapter\s+(\d+(?:\.\d+)?)')
 
+# JSON file written into each Kavita series folder, listing every chapter
+# already moved.  Avoids opening CBZ archives to determine what is present.
+# Absence is interpreted as "folder empty" (per spec), not "scan the folder".
+_CHAPTER_INDEX_FILENAME = "chapter_index.json"
+
+
+def _normalise_volume_value(value):
+    """
+    Normalises a volume identifier for storage in chapter_index.json.
+
+    Returns int when the value is a whole number, float for fractional
+    volumes, None when missing.  Mirrors how the user wants volumes
+    rendered (``"volume": 1`` rather than ``"volume": "1"``).
+    """
+    if value is None:
+        return None
+    text = str(value).strip()
+    if not text:
+        return None
+    try:
+        f = float(text)
+        return int(f) if f.is_integer() else f
+    except (TypeError, ValueError):
+        return text
+
+
+def _load_chapter_index(dest_series: Path) -> dict:
+    """
+    Reads chapter_index.json from a Kavita series folder.
+
+    Returns ``{"chapter": {}}`` when the file is missing or unreadable —
+    per the project spec, absence means "no chapters are present yet".
+    """
+    path = dest_series / _CHAPTER_INDEX_FILENAME
+    if not path.is_file():
+        return {"chapter": {}}
+    try:
+        with path.open("r", encoding="utf-8") as f:
+            data = json.load(f)
+    except (OSError, json.JSONDecodeError) as exc:
+        print(f"  [warn] chapter_index unreadable ({path.name}): {exc} — "
+              f"treating folder as empty")
+        return {"chapter": {}}
+    if not isinstance(data, dict) or not isinstance(data.get("chapter"), dict):
+        return {"chapter": {}}
+    return data
+
+
+def _save_chapter_index(dest_series: Path, index: dict) -> None:
+    """Writes chapter_index.json atomically into a Kavita series folder."""
+    path = dest_series / _CHAPTER_INDEX_FILENAME
+    tmp = path.with_suffix(path.suffix + ".tmp")
+    with tmp.open("w", encoding="utf-8") as f:
+        json.dump(index, f, ensure_ascii=False, indent=2)
+    tmp.replace(path)
+
 # Parenthetical source labels that Suwayomi appends to series names.
 # These are not part of the actual title and confuse MangaBaka searches.
 _SOURCE_LABEL_RE = re.compile(
@@ -75,11 +141,6 @@ _SOURCE_LABEL_RE = re.compile(
 _WIN_ILLEGAL_RE = re.compile(r'[\\/*?"<>|]')
 
 
-def _natural_key(name: str) -> list:
-    return [int(p) if p.isdigit() else p.lower()
-            for p in re.split(r"(\d+)", name)]
-
-
 def _sanitize_dirname(name: str) -> str:
     """
     Makes a string safe to use as a Windows (or SMB) directory name.
@@ -134,34 +195,6 @@ def _clean_suwayomi_title(title: str) -> str:
     return _SOURCE_LABEL_RE.sub("", title).strip()
 
 
-def _mal_id_from_metadata(md: dict) -> "int | None":
-    """Extracts the MAL ID from a MangaBaka series dict's source map."""
-    for raw_key, info in (md.get("source") or {}).items():
-        if re.sub(r"[^a-z0-9]", "", raw_key.lower()) in ("myanimelist", "mal"):
-            if isinstance(info, dict):
-                mal_id = info.get("id")
-                if mal_id is not None:
-                    try:
-                        return int(mal_id)
-                    except (TypeError, ValueError):
-                        pass
-    return None
-
-
-def _al_id_from_metadata(md: dict) -> "int | None":
-    """Extracts the AniList ID from a MangaBaka series dict's source map."""
-    for raw_key, info in (md.get("source") or {}).items():
-        if re.sub(r"[^a-z0-9]", "", raw_key.lower()) == "anilist":
-            if isinstance(info, dict):
-                al_id = info.get("id")
-                if al_id is not None:
-                    try:
-                        return int(al_id)
-                    except (TypeError, ValueError):
-                        pass
-    return None
-
-
 def _chapter_image_size(chapter_dir: Path) -> int:
     """Returns the total file size of all images in a chapter folder."""
     return sum(
@@ -278,6 +311,8 @@ class SuwayomiMover:
     language        : ComicInfo LanguageISO and SeriesSort language ("en").
     request_timeout : HTTP timeout in seconds for all API / image requests.
     delete_source   : Remove the source chapter folder after successful pack.
+    cover_cache_dir : Directory for the persistent cover cache.  None ->
+                      temporary cache, deleted at process exit.
     """
 
     def __init__(self,
@@ -290,7 +325,8 @@ class SuwayomiMover:
                  request_timeout: int = 30,
                  delete_source: bool = True,
                  matches_cache: "MatchesCache | None" = None,
-                 api_base_url: str = "https://api.mangabaka.dev/v1"):
+                 api_base_url: str = "https://api.mangabaka.dev/v1",
+                 cover_cache_dir=None):
         self._src = Path(suwayomi_path)
         self._dst = Path(kavita_path)
         self._language = language
@@ -303,6 +339,8 @@ class SuwayomiMover:
         # to maximise cache hits and minimise API round-trips.
         session = requests.Session()
         session.headers.setdefault("User-Agent", "SuwayomiMover/1.0")
+        # Throttle every call to api.mangabaka.dev (>=1s gap + retry on 429).
+        _apply_mangabaka_rate_limit(session)
         self._session = session
 
         self._mal = MALResolver(request_timeout=request_timeout)
@@ -311,14 +349,18 @@ class SuwayomiMover:
             request_timeout=request_timeout, session=session)
         self._works_resolver = MangaBakaWorksResolver(
             request_timeout=request_timeout, session=session)
+        self._cover_cache = CoverCache(
+            cover_cache_dir, session=session, request_timeout=request_timeout)
 
         self._person_updater: "KavitaPersonUpdater | None" = None
         if kavita_base_url and kavita_api_key:
-            self._person_updater = KavitaPersonUpdater(
+            kavita_client = KavitaClient(
                 kavita_base_url, kavita_api_key,
-                mal_resolver=self._mal,
-                al_resolver=self._al,
                 request_timeout=request_timeout)
+            self._person_updater = KavitaPersonUpdater(
+                kavita_client,
+                mal_resolver=self._mal,
+                al_resolver=self._al)
 
     # ------------------------------------------------------------------
     # Public API
@@ -362,6 +404,18 @@ class SuwayomiMover:
         raise FileNotFoundError(
             f"No Suwayomi directory found for '{manga_title}' under {self._src}")
 
+    def fetch_series(self, series_id) -> "dict | None":
+        """
+        Fetches a MangaBaka series by id via the shared (rate-limited) session.
+        Returns the inner `data` dict, or None if not found / empty.
+        """
+        if series_id is None or str(series_id).strip() == "":
+            return None
+        url = f"{self._api_base_url}/series/{series_id}"
+        resp = self._session.get(url, timeout=self._timeout)
+        resp.raise_for_status()
+        return resp.json().get("data")
+
     def build_matches_only(self) -> dict:
         """
         Walks every series under the Suwayomi root and resolves each one
@@ -410,7 +464,8 @@ class SuwayomiMover:
                 try:
                     resp = self._session.get(
                         search_url,
-                        params={"q": builder_title, "page": 1, "limit": 1},
+                        params={"q": builder_title, "type": _SEARCH_TYPES,
+                                "page": 1, "limit": 1},
                         timeout=self._timeout)
                     resp.raise_for_status()
                     data = resp.json().get("data") or []
@@ -422,7 +477,7 @@ class SuwayomiMover:
                         builder_title,
                         mangabaka_id=series.get("id"),
                         mangabaka_name=series.get("title") or "",
-                        image_url=_pick_cover_url(series.get("cover")),
+                        image_url=_pick_thumbnail_url(series.get("cover")),
                     )
                 except Exception as exc:
                     print(f"  [warn] search failed for {builder_title!r}: {exc}")
@@ -477,6 +532,7 @@ class SuwayomiMover:
             mal_resolver=self._mal,
             al_resolver=self._al,
             matches_cache=self._matches_cache,
+            cover_cache=self._cover_cache,
         )
 
         # Fetch MangaBaka metadata now to get the canonical title and MAL ID.
@@ -493,22 +549,47 @@ class SuwayomiMover:
         dest_series = self._dst / _sanitize_dirname(mangabaka_title)
         dest_series.mkdir(parents=True, exist_ok=True)
 
+        # Skip chapters that have already been moved to Kavita.  The index
+        # file in the destination folder is the authoritative source — we
+        # never open CBZ archives or stat them individually.
+        chapter_index = _load_chapter_index(dest_series)
+        already_moved = chapter_index["chapter"]
+        skipped: list[tuple[Path, str]] = []
+        pending:  list[tuple[Path, dict, str]] = []
+        for item in chapter_items:
+            chapter_dir, _fields, chapter_num = item
+            if chapter_num in already_moved:
+                skipped.append((chapter_dir, chapter_num))
+            else:
+                pending.append(item)
+
+        for chapter_dir, chapter_num in skipped:
+            print(f"  Chapter {chapter_num}: skip (already in Kavita)")
+            if self._delete_source:
+                shutil.rmtree(chapter_dir, ignore_errors=True)
+
         chapter_results: list[dict] = []
-        for chapter_dir, _fields, chapter_num in chapter_items:
+        for chapter_dir, _fields, chapter_num in pending:
             result = self._process_chapter(
                 builder, chapter_num, chapter_dir, dest_series)
             chapter_results.append(result)
             status = "ok" if result["ok"] else f"ERROR: {result.get('error')}"
             print(f"  Chapter {chapter_num}: {status}")
+            if result["ok"]:
+                already_moved[chapter_num] = {
+                    "volume":      _normalise_volume_value(result.get("volume")),
+                    "archiveName": Path(result["cbz"]).name,
+                }
+                _save_chapter_index(dest_series, chapter_index)
 
         # Sync Kavita persons once per series.
         # Both MAL and AniList IDs come from MangaBaka's source map;
         # AniList is used as fallback when MAL returns no characters/staff.
         person_result: "dict | None" = None
         if self._person_updater:
-            mal_id = (_mal_id_from_metadata(md) if md else None
+            mal_id = ((ComicInfoBuilder._mal_id_from_source(md) if md else None)
                       or self._mal.find_mal_id(builder_title))
-            al_id  = _al_id_from_metadata(md) if md else None
+            al_id  = ComicInfoBuilder._al_id_from_source(md) if md else None
             if mal_id or al_id:
                 try:
                     person_result = self._person_updater.update_for_manga(
@@ -541,11 +622,19 @@ class SuwayomiMover:
         try:
             builder.chapter = chapter_num
             builder.add_pages_from_folder(chapter_dir, cover_filename="000")
+            # Resolving the volume here piggy-backs on caches already warmed
+            # by add_pages_from_folder, so it's effectively free.  Used by
+            # the chapter index in the Kavita destination folder.
+            try:
+                volume = builder._determine_volume()
+            except Exception:
+                volume = None
             builder.save_xml(chapter_dir)
             _pack_to_cbz(chapter_dir, cbz_path)
             if self._delete_source:
                 shutil.rmtree(chapter_dir)
-            return {"chapter": chapter_num, "cbz": str(cbz_path), "ok": True}
+            return {"chapter": chapter_num, "cbz": str(cbz_path),
+                    "ok": True, "volume": volume}
         except Exception as exc:
             return {"chapter": chapter_num, "cbz": str(cbz_path),
                     "ok": False, "error": str(exc)}
@@ -555,21 +644,38 @@ class SuwayomiMover:
 # Usage example
 # --------------------------------------------------------------------------
 if __name__ == "__main__":
-    SUWAYOMI_PATH = r"\\192.168.2.2\root\Temp\managdl\mangas"
+    import os
+
+    # Local (no-Docker) smoke test.  Adjust paths to your environment.
+    # Set the KAVITA_API_KEY env var — never commit API keys to the repo.
+    SUWAYOMI_PATH = r"M:\config\downloads\mangas"
     KAVITA_PATH   = r"\\192.168.2.2\root\ServerData\Kavita\test"
     KAVITA_URL    = "http://192.168.2.2:5000"
-    KAVITA_KEY    = "Sq4a3hcV171dn3gzCl0K4eN7hZNk4sOA"
+    KAVITA_KEY    = os.environ.get("KAVITA_API_KEY", "")
+
+    # matches.json lives next to this script during local testing.
+    MATCHES_PATH  = Path(__file__).resolve().parent.parent / "matches.json"
+    matches_cache = MatchesCache(MATCHES_PATH)
 
     mover = SuwayomiMover(
         SUWAYOMI_PATH,
         KAVITA_PATH,
         kavita_base_url=KAVITA_URL,
         kavita_api_key=KAVITA_KEY,
-        delete_source=False
+        delete_source=False,
+        matches_cache=matches_cache,
     )
 
-    # Process a single series
-    result = mover.process_series("Yofukashi no Uta")
+    # ---- Option A: build matches.json only (no moves / no Kavita sync) ----
+    # data = mover.build_matches_only()
+    # matches = data.get("matches", {})
+    # print(f"\n[matches] {len(matches)} entries total — file: {MATCHES_PATH}")
+    # for title, entry in list(matches.items())[:10]:
+    #     print(f"  {title!r:50s}  id={entry.get('mangabakaId')}  "
+    #           f"name={entry.get('mangabakaName')!r}")
+
+    # ---- Option B: full pipeline for one series (uses the cache too) ----
+    result = mover.process_series("Wistoria - Wand and Sword")
     ok     = sum(1 for c in result["chapters"] if c["ok"])
     failed = sum(1 for c in result["chapters"] if not c["ok"])
     print(f"\nDone: {ok} ok, {failed} failed")