init

2026-05-27 18:58:42 +02:00
parent cd0757ab6d
commit 01c562dc46
20 changed files with 4579 additions and 0 deletions
@@ -0,0 +1,6 @@
+KAVITA_URL=http://192.168.1.100:5000
+KAVITA_API_KEY=your-api-key-here
+LIBRARY_IDS=3,5
+LANGUAGE=en
+MATCH_PATH=matches.json
+WEB_PORT=8080
@@ -267,3 +267,10 @@ pyvenv.cfg
 .venv
 pip-selfcheck.json

+manga-mover-and-metadata-collector/
+
+# Project-local state
+matches.json
+config/
+output/
+
@@ -0,0 +1,18 @@
+FROM python:3.12-slim
+
+WORKDIR /app
+
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY src/     /app/src/
+COPY main.py  /app/main.py
+
+ENV PYTHONUNBUFFERED=1 \
+    PYTHONDONTWRITEBYTECODE=1
+
+VOLUME ["/config"]
+
+EXPOSE 8080
+
+CMD ["python", "/app/main.py"]
@@ -1,2 +1,56 @@
 # kavita-lightnovel-metadata-fetcher

+Pulls metadata (summary, tags, genres, characters, staff, score,
+cover, links, related series) for light novels from **MangaBaka**,
+enriched with **MyAnimeList** and **AniList** data, and writes it
+back to a **Kavita** server through its REST API.
+
+No file mover, no ComicInfo.xml — the source of truth is Kavita
+itself.  Series are discovered via the Kavita library API.
+
+## Features
+
+- Match every series in one or more Kavita libraries against
+  MangaBaka and persist the match in `matches.json` (editable via
+  the web UI).
+- Update metadata for a single series or all matched series at
+  once.  Updates are diff-based:
+  - Locked fields in Kavita are never overwritten.
+  - List fields (tags, genres, characters, writers, …) are merged:
+    new items are added, removed items are dropped.
+  - Cover images are only re-uploaded when MangaBaka's cover URL
+    actually changed.
+- Characters and authors are synced to Kavita Person records
+  (image, description, MAL/AniList id) via Kavita's `/api/Person`
+  endpoints.
+- MangaBaka relationships (sequel / prequel / spin-off / …) are
+  mirrored as Kavita series relationships, and every related
+  series that exists in Kavita is added to a shared collection.
+
+## Environment
+
+| Variable           | Default                   | Description                                              |
+| ------------------ | ------------------------- | -------------------------------------------------------- |
+| `KAVITA_URL`       | —                         | Base URL of the Kavita server, e.g. `http://kavita:5000` |
+| `KAVITA_API_KEY`   | —                         | API key from Kavita user settings                        |
+| `LIBRARY_IDS`      | _(empty)_                 | Default libraries (CSV of ids).  Empty = pick in WebUI.  |
+| `LANGUAGE`         | `en`                      | Series language ISO code (used for `language` field)     |
+| `REQUEST_TIMEOUT`  | `30`                      | HTTP timeout in seconds                                  |
+| `MATCH_PATH`       | `/config/matches.json`    | Where to persist the match cache                         |
+| `WEB_HOST`         | `0.0.0.0`                 | Bind host for the Flask UI                               |
+| `WEB_PORT`         | `8080`                    | Bind port for the Flask UI                               |
+
+## Running locally
+
+```bash
+pip install -r requirements.txt
+KAVITA_URL=http://localhost:5000 KAVITA_API_KEY=... python main.py
+```
+
+Then open <http://localhost:8080/>.
+
+## Docker
+
+```bash
+docker compose -f docker-compose.prod.yml up -d
+```
@@ -0,0 +1,16 @@
+services:
+  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:    "${LIBRARY_IDS}"
+      LANGUAGE:       "${LANGUAGE:-en}"
+      MATCH_PATH:     "${MATCH_PATH:-/config/matches.json}"
+      WEB_PORT:       "${WEB_PORT:-8080}"
+    ports:
+      - "${WEB_PORT:-8080}:${WEB_PORT:-8080}"
+    volumes:
+      - "${HOST_CONFIG_PATH}:/config"
@@ -0,0 +1,122 @@
+"""
+main.py
+=======
+
+Container entry point for the Kavita light-novel metadata fetcher.
+
+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
+
+# Make src/ importable when running as `python main.py`.
+sys.path.insert(0, str(Path(__file__).resolve().parent / "src"))
+
+from src.MatchesCache import MatchesCache                       # noqa: E402
+from src.LightNovelOrchestrator import LightNovelOrchestrator   # noqa: E402
+from src.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())
@@ -0,0 +1,3 @@
+requests>=2.31
+Flask>=3.0
+python-dotenv>=1.0
@@ -0,0 +1,507 @@
+"""
+anilist_resolver.py
+===================
+
+Fetches and caches AniList manga metadata (statistics, characters, staff)
+using the public AniList GraphQL API.
+
+AniList API: https://graphql.anilist.co  (no authentication required)
+Rate limit: 90 req/min  ->  a 700 ms guard between calls is applied.
+On HTTP 429 (rate-limit exceeded) the response Retry-After header is
+honoured; the request is retried once automatically.
+
+Singleton
+---------
+Only one instance of this class exists per process.  Subsequent calls to
+AniListResolver() return the same object with its warm caches intact.
+
+Provided features
+-----------------
+- Title-based AniList ID lookup with best-match scoring
+- Manga statistics: score (0–10), rank, popularity, members, favorites
+- Character list for a manga (names only — for <Characters> XML tag)
+- Detailed character list: name, AniList character ID, image URL, role
+- Detailed staff list: name, AniList person ID, image URL, positions
+- Lazy full-detail fetches per character / person (for descriptions)
+
+Dependencies
+------------
+    requests    ->  pip install requests
+"""
+
+from __future__ import annotations
+
+import datetime
+import difflib
+import time
+
+import requests
+
+from MediaResolver import MediaResolver
+
+
+# --------------------------------------------------------------------------
+# GraphQL query strings
+# --------------------------------------------------------------------------
+_SEARCH_MANGA = """
+query ($search: String) {
+  Page(page: 1, perPage: 5) {
+    media(search: $search, type: MANGA, format_in: [NOVEL]) {
+      id title { romaji english native } siteUrl
+    }
+  }
+}
+"""
+
+_MANGA_STATS = """
+query ($id: Int) {
+  Media(id: $id, type: MANGA) {
+    id title { romaji english native }
+    meanScore popularity favourites
+    rankings { rank type allTime }
+    siteUrl
+  }
+}
+"""
+
+_MANGA_CHARACTERS = """
+query ($id: Int) {
+  Media(id: $id, type: MANGA) {
+    characters(sort: [ROLE, RELEVANCE], perPage: 25) {
+      nodes { id name { full } image { large } siteUrl }
+      edges { role }
+    }
+  }
+}
+"""
+
+_MANGA_STAFF = """
+query ($id: Int) {
+  Media(id: $id, type: MANGA) {
+    staff(perPage: 25) {
+      nodes { id name { full } image { large } siteUrl }
+      edges { role }
+    }
+  }
+}
+"""
+
+_CHARACTER_DETAILS = """
+query ($id: Int) {
+  Character(id: $id) {
+    id name { full } image { large }
+    description(asHtml: false)
+    favourites siteUrl
+  }
+}
+"""
+
+_PERSON_DETAILS = """
+query ($id: Int) {
+  Staff(id: $id) {
+    id name { full native } image { large }
+    description(asHtml: false)
+    favourites siteUrl
+    dateOfBirth { year month day }
+    primaryOccupations
+    homeTown
+  }
+}
+"""
+
+_ANILIST_GQL = "https://graphql.anilist.co"
+
+
+class AniListResolver(MediaResolver):
+    """
+    Singleton: fetches and caches AniList manga data via GraphQL API.
+
+    The first call to AniListResolver() creates and initialises the instance;
+    all subsequent calls return the same object.
+    """
+
+    _instance: "AniListResolver | None" = None
+
+    # ------------------------------------------------------------------
+    # Singleton machinery
+    # ------------------------------------------------------------------
+    def __new__(cls, **kwargs):
+        if cls._instance is None:
+            cls._instance = super().__new__(cls)
+            cls._instance._initialized = False
+        return cls._instance
+
+    def __init__(self, *, request_timeout: int = 30):
+        if self._initialized:
+            return
+
+        self.request_timeout = request_timeout
+
+        self._session = requests.Session()
+        self._session.headers.update({
+            "User-Agent":   "AniListResolver/1.0",
+            "Content-Type": "application/json",
+            "Accept":       "application/json",
+        })
+
+        # title_lower -> al_id
+        self._id_cache: dict[str, "int | None"] = {}
+        # al_id -> stats dict
+        self._stats_cache: dict[int, dict] = {}
+        # manga_al_id -> [name_str, ...]
+        self._char_names_cache: dict[int, list[str]] = {}
+        # manga_al_id -> [{al_id, name, image_url, role}]
+        self._char_detailed_cache: dict[int, list[dict]] = {}
+        # manga_al_id -> [{al_id, name, image_url, positions}]
+        self._staff_detailed_cache: dict[int, list[dict]] = {}
+        # char_al_id -> {al_id, name, image_url, about, favorites, url}
+        self._char_info_cache: dict[int, dict] = {}
+        # person_al_id -> {al_id, name, image_url, about, favorites, url, ...}
+        self._person_info_cache: dict[int, dict] = {}
+
+        self._last_request_at: float = 0.0
+        self._initialized = True
+
+    # ------------------------------------------------------------------
+    # Public: ID lookup
+    # ------------------------------------------------------------------
+    def find_id(self, title: str) -> "int | None":
+        """
+        Searches AniList for a manga by title and returns the best-matching
+        AniList ID.  Returns None on failure or when no result is found.
+        """
+        if not title or not title.strip():
+            return None
+
+        key = title.strip().lower()
+        if key in self._id_cache:
+            return self._id_cache[key]
+
+        try:
+            data = self._gql(_SEARCH_MANGA, {"search": title})
+            results = ((data.get("data") or {})
+                       .get("Page", {})
+                       .get("media") or [])
+        except requests.RequestException:
+            return None
+
+        if not results:
+            self._id_cache[key] = None
+            return None
+
+        results.sort(key=lambda e: _score_title(title, e), reverse=True)
+        al_id = results[0].get("id")
+        self._id_cache[key] = al_id
+        return al_id
+
+    # ------------------------------------------------------------------
+    # Public: statistics
+    # ------------------------------------------------------------------
+    def get_stats(self, tracker_id: "int | None") -> "dict | None":
+        """
+        Returns a statistics dict for the given AniList manga ID:
+
+            {score, rank, scored_by, popularity, members, favorites,
+             url, title, as_of (DD-MM-YYYY)}
+
+        Returns None if tracker_id is None or on network failure.
+        """
+        if tracker_id is None:
+            return None
+        if tracker_id in self._stats_cache:
+            return self._stats_cache[tracker_id]
+
+        try:
+            data = self._gql(_MANGA_STATS, {"id": tracker_id})
+            entry = (data.get("data") or {}).get("Media") or {}
+        except requests.RequestException:
+            return None
+
+        title_obj = entry.get("title") or {}
+        title = (title_obj.get("romaji")
+                 or title_obj.get("english")
+                 or title_obj.get("native") or "")
+
+        # AniList meanScore is 0–100; normalise to 0.0–10.0 for consistency
+        # with the MALResolver stats dict shape.
+        raw_score = entry.get("meanScore")
+        score = round(raw_score / 10, 1) if raw_score is not None else None
+
+        # Ranked and popularity ranks are in the rankings array.
+        rated_rank  = None
+        popular_rank = None
+        for r in (entry.get("rankings") or []):
+            if r.get("allTime"):
+                if r.get("type") == "RATED"   and rated_rank  is None:
+                    rated_rank  = r.get("rank")
+                if r.get("type") == "POPULAR" and popular_rank is None:
+                    popular_rank = r.get("rank")
+
+        stats: dict = {
+            "score":      score,
+            "rank":       rated_rank,
+            "scored_by":  None,            # not exposed by AniList API
+            "popularity": popular_rank,
+            "members":    entry.get("popularity"),   # AniList's popularity = member count
+            "favorites":  entry.get("favourites"),
+            "url":        entry.get("siteUrl") or f"https://anilist.co/manga/{tracker_id}",
+            "title":      title,
+            "as_of":      datetime.date.today().strftime("%d-%m-%Y"),
+        }
+        self._stats_cache[tracker_id] = stats
+        return stats
+
+    # ------------------------------------------------------------------
+    # Public: character names (for ComicInfo <Characters> tag)
+    # ------------------------------------------------------------------
+    def get_characters(self, tracker_id: "int | None") -> list[str]:
+        """Returns a flat list of character names for the manga."""
+        if tracker_id is None:
+            return []
+        if tracker_id in self._char_names_cache:
+            return self._char_names_cache[tracker_id]
+
+        detailed = self.get_characters_detailed(tracker_id)
+        names = [e["name"] for e in detailed if e.get("name")]
+        if names:
+            self._char_names_cache[tracker_id] = names
+        return names
+
+    # ------------------------------------------------------------------
+    # Public: detailed character data
+    # ------------------------------------------------------------------
+    def get_characters_detailed(self, tracker_id: "int | None") -> list[dict]:
+        """
+        Returns detailed character entries for a manga:
+            [{al_id, mal_id, name, image_url, role, about=None}, ...]
+        """
+        if tracker_id is None:
+            return []
+        if tracker_id in self._char_detailed_cache:
+            return self._char_detailed_cache[tracker_id]
+
+        try:
+            data = self._gql(_MANGA_CHARACTERS, {"id": tracker_id})
+            chars = ((data.get("data") or {})
+                     .get("Media", {})
+                     .get("characters") or {})
+            nodes = chars.get("nodes") or []
+            edges = chars.get("edges") or []
+        except requests.RequestException:
+            return []
+
+        results = []
+        for node, edge in zip(nodes, edges):
+            name = (node.get("name") or {}).get("full") or ""
+            if not name:
+                continue
+            results.append({
+                "al_id":     node.get("id"),
+                "mal_id":    None,
+                "name":      name,
+                "raw_name":  name,
+                "image_url": (node.get("image") or {}).get("large"),
+                "role":      edge.get("role") or "SUPPORTING",
+                "about":     None,
+            })
+
+        if results:
+            self._char_detailed_cache[tracker_id] = results
+        return results
+
+    # ------------------------------------------------------------------
+    # Public: detailed staff data
+    # ------------------------------------------------------------------
+    def get_staff_detailed(self, tracker_id: "int | None") -> list[dict]:
+        """
+        Returns detailed staff entries for a manga:
+            [{al_id, mal_id, name, image_url, positions, about=None}, ...]
+        """
+        if tracker_id is None:
+            return []
+        if tracker_id in self._staff_detailed_cache:
+            return self._staff_detailed_cache[tracker_id]
+
+        try:
+            data = self._gql(_MANGA_STAFF, {"id": tracker_id})
+            staff = ((data.get("data") or {})
+                     .get("Media", {})
+                     .get("staff") or {})
+            nodes = staff.get("nodes") or []
+            edges = staff.get("edges") or []
+        except requests.RequestException:
+            return []
+
+        results = []
+        for node, edge in zip(nodes, edges):
+            name = (node.get("name") or {}).get("full") or ""
+            if not name:
+                continue
+            results.append({
+                "al_id":     node.get("id"),
+                "mal_id":    None,
+                "name":      name,
+                "raw_name":  name,
+                "image_url": (node.get("image") or {}).get("large"),
+                "positions": [edge.get("role")] if edge.get("role") else [],
+                "about":     None,
+            })
+
+        if results:
+            self._staff_detailed_cache[tracker_id] = results
+        return results
+
+    # ------------------------------------------------------------------
+    # Public: individual character / person details
+    # ------------------------------------------------------------------
+    def get_character_details(self, char_id: "int | None") -> "dict | None":
+        """Returns full details for a single AniList character."""
+        if char_id is None:
+            return None
+        if char_id in self._char_info_cache:
+            return self._char_info_cache[char_id]
+
+        try:
+            data = self._gql(_CHARACTER_DETAILS, {"id": char_id})
+            entry = (data.get("data") or {}).get("Character") or {}
+        except requests.RequestException:
+            return None
+
+        result = {
+            "al_id":     entry.get("id"),
+            "mal_id":    None,
+            "name":      (entry.get("name") or {}).get("full") or "",
+            "image_url": (entry.get("image") or {}).get("large"),
+            "about":     entry.get("description"),
+            "favorites": entry.get("favourites"),
+            "url":       entry.get("siteUrl") or f"https://anilist.co/character/{char_id}",
+        }
+        self._char_info_cache[char_id] = result
+        return result
+
+    def get_person_details(self, person_id: "int | None") -> "dict | None":
+        """Returns full details for a single AniList staff person."""
+        if person_id is None:
+            return None
+        if person_id in self._person_info_cache:
+            return self._person_info_cache[person_id]
+
+        try:
+            data = self._gql(_PERSON_DETAILS, {"id": person_id})
+            entry = (data.get("data") or {}).get("Staff") or {}
+        except requests.RequestException:
+            return None
+
+        # dateOfBirth: {year, month, day} → ISO string for _format_birthday
+        dob = entry.get("dateOfBirth") or {}
+        birthday: "str | None" = None
+        if dob.get("year"):
+            m = dob.get("month") or 1
+            d = dob.get("day") or 1
+            birthday = f"{dob['year']}-{m:02d}-{d:02d}"
+
+        name_obj = entry.get("name") or {}
+        result = {
+            "al_id":       entry.get("id"),
+            "mal_id":      None,
+            "name":        name_obj.get("full") or "",
+            "given_name":  None,      # AniList does not break names into given/family
+            "family_name": None,
+            "birthday":    birthday,
+            "image_url":   (entry.get("image") or {}).get("large"),
+            "about":       entry.get("description"),
+            "favorites":   entry.get("favourites"),
+            "website_url": None,      # not exposed by AniList public API
+            "url":         entry.get("siteUrl") or f"https://anilist.co/staff/{person_id}",
+        }
+        self._person_info_cache[person_id] = result
+        return result
+
+    # ------------------------------------------------------------------
+    # Public: cache management
+    # ------------------------------------------------------------------
+    def clear_cache(self) -> None:
+        """Clears all internal caches (the Singleton instance is retained)."""
+        self._id_cache.clear()
+        self._stats_cache.clear()
+        self._char_names_cache.clear()
+        self._char_detailed_cache.clear()
+        self._staff_detailed_cache.clear()
+        self._char_info_cache.clear()
+        self._person_info_cache.clear()
+
+    # ------------------------------------------------------------------
+    # Internal: rate-limited GraphQL POST
+    # ------------------------------------------------------------------
+    def _gql(self, query: str, variables: "dict | None" = None) -> dict:
+        """
+        Rate-limited GraphQL POST request (respects AniList's 90 req/min limit).
+
+        On HTTP 429 the Retry-After header is honoured and the request is
+        retried once.
+        """
+        elapsed = time.monotonic() - self._last_request_at
+        if elapsed < 0.7:
+            time.sleep(0.7 - elapsed)
+
+        payload: dict = {"query": query}
+        if variables:
+            payload["variables"] = variables
+
+        resp = self._session.post(
+            _ANILIST_GQL, json=payload, timeout=self.request_timeout)
+        self._last_request_at = time.monotonic()
+
+        if resp.status_code == 429:
+            retry_after = int(resp.headers.get("Retry-After", 60))
+            time.sleep(retry_after)
+            resp = self._session.post(
+                _ANILIST_GQL, json=payload, timeout=self.request_timeout)
+            self._last_request_at = time.monotonic()
+
+        resp.raise_for_status()
+        return resp.json()
+
+
+# --------------------------------------------------------------------------
+# Module helpers
+# --------------------------------------------------------------------------
+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
+
+
+# --------------------------------------------------------------------------
+# Usage example
+# --------------------------------------------------------------------------
+if __name__ == "__main__":
+    r1 = AniListResolver()
+    r2 = AniListResolver()
+    assert r1 is r2, "AniListResolver must be a Singleton"
+
+    al_id = r1.find_id("Yofukashi no Uta")
+    print("AniList ID   :", al_id)
+
+    stats = r1.get_stats(al_id)
+    if stats:
+        print("Score        :", stats["score"])
+        print("Rank         :", stats["rank"])
+        print("Members      :", stats["members"])
+
+    chars = r1.get_characters_detailed(al_id)
+    print("Characters (first 3):", [c["name"] for c in chars[:3]])
+
+    staff = r1.get_staff_detailed(al_id)
+    print("Staff        :", [s["name"] for s in staff])
@@ -0,0 +1,229 @@
+"""
+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 {}
+
+    # ------------------------------------------------------------------
+    # Series cover upload
+    # ------------------------------------------------------------------
+    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.
+
+        Mirrors the cover-upload trick used in KavitaPersonUpdater:
+        Kavita's `/api/Upload/series` accepts a raw base64 blob (no
+        ``data:`` prefix) in the ``url`` field.
+        """
+        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}/api/Upload/series",
+            json={"id": series_id, "url": b64, "lockCover": lock},
+            timeout=self._timeout)
+        r.raise_for_status()
+
+    # ------------------------------------------------------------------
+    # 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)
@@ -0,0 +1,545 @@
+"""
+kavita_person_updater.py
+========================
+
+Synchronises Kavita person / character records with MyAnimeList data.
+
+For every character and staff member that MAL knows about for a given manga
+the updater:
+  1. Searches Kavita for a matching Person record (by name similarity /
+     alias match, configurable threshold).
+  2. Sets the MAL ID on the Kavita person if it is not yet linked.
+  3. Uploads the MAL profile image when the cover is not locked and has
+     not been set in a previous sync run.
+  4. Populates the description field when Kavita has none and MAL provides
+     an 'about' text (requires an extra Jikan request per character; only
+     performed when update_descriptions=True).
+
+Kavita API version
+------------------
+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 MALResolver import MALResolver
+from AniListResolver import AniListResolver
+
+
+class KavitaPersonUpdater:
+    """
+    Syncs Kavita Person records with MyAnimeList data.
+
+    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)
+    mal_resolver    : Shared MALResolver singleton (created automatically if omitted)
+    request_timeout : HTTP timeout in seconds for both Kavita and image requests
+    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, *,
+                 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._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]] = {}
+
+    # ------------------------------------------------------------------
+    # Public: combined update
+    # ------------------------------------------------------------------
+    def update_for_manga(self, mal_manga_id: "int | None", *,
+                         al_manga_id: "int | None" = None,
+                         update_covers: bool = True,
+                         update_descriptions: bool = True) -> dict:
+        """
+        Runs a full update pass for both characters and staff of the manga.
+        MAL is tried first; AniList is used as fallback when MAL returns nothing.
+
+        Returns
+        -------
+        {
+          "characters": {"updated": n, "skipped": n, "not_found": n},
+          "staff":      {"updated": n, "skipped": n, "not_found": n},
+        }
+        """
+        return {
+            "characters": self.update_characters(
+                mal_manga_id, al_manga_id=al_manga_id,
+                update_covers=update_covers,
+                update_descriptions=update_descriptions),
+            "staff": self.update_staff(
+                mal_manga_id, al_manga_id=al_manga_id,
+                update_covers=update_covers,
+                update_descriptions=update_descriptions),
+        }
+
+    # ------------------------------------------------------------------
+    # Public: character update
+    # ------------------------------------------------------------------
+    def update_characters(self, mal_manga_id: "int | None", *,
+                          al_manga_id: "int | None" = None,
+                          update_covers: bool = True,
+                          update_descriptions: bool = True) -> dict:
+        """
+        Updates Kavita persons that match MAL/AniList characters for the manga.
+        MAL is tried first; AniList is the fallback when MAL returns nothing.
+
+        Returns {"updated": n, "skipped": n, "not_found": n}.
+        """
+        entries = self._mal.get_characters_detailed(mal_manga_id) if mal_manga_id else []
+        resolver = self._mal
+        if not entries and al_manga_id:
+            entries = self._al.get_characters_detailed(al_manga_id)
+            resolver = self._al
+        return self._sync_entries(entries, "character", resolver,
+                                  update_covers=update_covers,
+                                  update_descriptions=update_descriptions)
+
+    # ------------------------------------------------------------------
+    # Public: staff update
+    # ------------------------------------------------------------------
+    def update_staff(self, mal_manga_id: "int | None", *,
+                     al_manga_id: "int | None" = None,
+                     update_covers: bool = True,
+                     update_descriptions: bool = True) -> dict:
+        """
+        Updates Kavita persons that match MAL/AniList staff for the manga.
+        MAL is tried first; AniList is the fallback when MAL returns nothing.
+
+        Returns {"updated": n, "skipped": n, "not_found": n}.
+        """
+        entries = self._mal.get_staff_detailed(mal_manga_id) if mal_manga_id else []
+        resolver = self._mal
+        if not entries and al_manga_id:
+            entries = self._al.get_staff_detailed(al_manga_id)
+            resolver = self._al
+        return self._sync_entries(entries, "staff", resolver,
+                                  update_covers=update_covers,
+                                  update_descriptions=update_descriptions)
+
+    # ------------------------------------------------------------------
+    # Public: cache management
+    # ------------------------------------------------------------------
+    def clear_cache(self) -> None:
+        """Clears the Kavita person search cache."""
+        self._person_search_cache.clear()
+
+    # ------------------------------------------------------------------
+    # Internal: main sync loop
+    # ------------------------------------------------------------------
+    def _sync_entries(self, entries: list[dict], kind: str, resolver, *,
+                      update_covers: bool,
+                      update_descriptions: bool) -> dict:
+        result: dict = {"updated": 0, "skipped": 0, "not_found": 0,
+                        "errors": []}
+        for entry in entries:
+            name = (entry.get("name") or "").strip()
+            raw_name = (entry.get("raw_name") or "").strip()
+            if not name and not raw_name:
+                continue
+
+            # Search by the 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)
+
+            if not matches:
+                result["not_found"] += 1
+                continue
+
+            changed = self._apply_mal_data(
+                matches[0], entry, kind, resolver,
+                update_cover=update_covers,
+                update_desc=update_descriptions,
+                errors=result["errors"])
+            result["updated" if changed else "skipped"] += 1
+
+        return result
+
+    # ------------------------------------------------------------------
+    # Internal: Kavita person search
+    # ------------------------------------------------------------------
+    def _find_kavita_person(self, name: str) -> list[dict]:
+        """
+        Searches Kavita for persons matching `name`.
+
+        Checks both the main name and any stored aliases.
+        Returns persons sorted by similarity, filtered by min_name_score.
+        Results are cached per (normalised) query name.
+        """
+        key = name.lower().strip()
+        if key in self._person_search_cache:
+            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 []
+        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]
+        self._person_search_cache[key] = filtered
+        return filtered
+
+    # ------------------------------------------------------------------
+    # Internal: apply MAL data to a single Kavita person
+    # ------------------------------------------------------------------
+    def _apply_mal_data(self, person: dict, mal_entry: dict, kind: str,
+                        resolver, *,
+                        update_cover: bool, update_desc: bool,
+                        errors: "list | None" = None) -> bool:
+        """
+        Applies tracker data (MAL or AniList) to one Kavita person record.
+
+        Fields updated
+        --------------
+        - malId      : set when the entry carries a MAL ID and it differs
+        - aniListId  : set when the entry carries an AniList ID and it differs
+        - description: set when empty and the tracker provides a description
+        - cover image: uploaded when not locked and no prior sync cover exists
+
+        Returns True if any change was made.  Failures are appended to the
+        `errors` list (if provided) instead of being silently swallowed.
+        """
+        person_id: "int | None" = person.get("id")
+        if not person_id:
+            return False
+
+        person_name = person.get("name") or ""
+
+        # Tracker IDs — a MAL entry has mal_id set; an AniList entry has al_id.
+        mal_id: "int | None" = mal_entry.get("mal_id")
+        al_id:  "int | None" = mal_entry.get("al_id")
+        entity_id = mal_id or al_id   # used for resolver detail calls
+
+        current_mal_id: int = person.get("malId") or 0
+        current_al_id:  int = person.get("aniListId") or 0
+        needs_mal_id = bool(mal_id and current_mal_id != mal_id)
+        needs_al_id  = bool(al_id  and current_al_id  != al_id)
+
+        # ------ Lazy description fetch -----------------------------------
+        description: "str | None" = None
+        if update_desc and not (person.get("description") or "").strip():
+            if entity_id:
+                if kind == "character":
+                    details = resolver.get_character_details(entity_id)
+                    if details:
+                        description = _build_character_description(details) or None
+                else:
+                    details = resolver.get_person_details(entity_id)
+                    if details:
+                        description = _build_person_description(details) or None
+
+        needs_desc = bool(description)
+
+        # ------ Metadata update ------------------------------------------
+        changed = False
+        if needs_mal_id or needs_al_id or needs_desc:
+            payload: dict = {
+                "id":               person_id,
+                "name":             person_name,
+                # MUST stay a boolean — the cover image itself is uploaded
+                # separately via POST /api/Upload/person (below).  Putting a
+                # URL here makes Kavita reject the whole payload with HTTP 400.
+                "coverImageLocked": bool(person.get("coverImageLocked", False)),
+                "aliases":          person.get("aliases") or [],
+                "description":      description or person.get("description"),
+                "malId":    mal_id if needs_mal_id else (current_mal_id or None),
+                "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()
+                changed = True
+            except requests.RequestException as e:
+                if errors is not None:
+                    errors.append(
+                        f"Person/update failed for #{person_id} "
+                        f"'{person_name}': {e}")
+
+        # ------ Cover image upload ----------------------------------------
+        # Upload whenever:
+        #   - caller requested cover updates
+        #   - cover is NOT locked (user did not manually pin it)
+        #   - we have not already uploaded this exact tracker entity's image
+        #     (i.e. the tracked ID differs OR there is no cover yet).
+        if update_cover and not person.get("coverImageLocked"):
+            image_url = mal_entry.get("image_url")
+            already_uploaded = (
+                entity_id is not None
+                and (current_mal_id == mal_id or current_al_id == al_id)
+                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):
+                    changed = True
+
+        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:
+        return ""
+    try:
+        dt = datetime.date.fromisoformat(birthday.split("T")[0])
+        return f"{dt.day} {dt.strftime('%B %Y')}"
+    except (ValueError, AttributeError):
+        return ""
+
+
+def _build_character_description(details: dict) -> str:
+    """
+    Builds a Kavita-safe HTML description for a MAL character.
+
+    Top line: "Favorites: N" as a link to the character's MAL page.
+    Remainder: the character's `about` text converted to HTML paragraphs.
+    """
+    parts: list[str] = []
+    url = details.get("url") or ""
+    favorites = details.get("favorites")
+    if url and favorites is not None:
+        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))
+    return "<br>".join(parts)
+
+
+def _build_person_description(details: dict) -> str:
+    """
+    Builds a Kavita-safe HTML description for a MAL person (mangaka / staff).
+
+    Renders a summary table (given name, family name, birthday, website,
+    member favorites) followed by the `about` biography as HTML paragraphs.
+    """
+    _TD = 'style="padding-right:1.5em"'
+    rows: list[str] = []
+
+    given = (details.get("given_name") or "").strip()
+    family = (details.get("family_name") or "").strip()
+    birthday = details.get("birthday") or ""
+    favorites = details.get("favorites")
+    website = (details.get("website_url") or "").strip()
+    url = (details.get("url") or "").strip()
+
+    if given:
+        rows.append(f"<tr><td {_TD}>Given name</td><td>{given}</td></tr>")
+    if family:
+        rows.append(f"<tr><td {_TD}>Family name</td><td>{family}</td></tr>")
+    bday_str = _format_birthday(birthday)
+    if bday_str:
+        rows.append(f"<tr><td {_TD}>Birthday</td><td>{bday_str}</td></tr>")
+    if website:
+        rows.append(
+            f'<tr><td {_TD}>Website</td>'
+            f'<td><a href="{website}">{website}</a></td></tr>'
+        )
+    if favorites is not None:
+        fav_cell = (f'<a href="{url}" target="_blank">{favorites:,}</a>' if url
+                    else f"{favorites:,}")
+        rows.append(
+            f"<tr><td {_TD}>Member Favorites</td><td>{fav_cell}</td></tr>")
+
+    parts: list[str] = []
+    if rows:
+        parts.append(f'<table>{"".join(rows)}</table>')
+    about = (details.get("about") or "").strip()
+    if about:
+        parts.append(_plain_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"
+
+    updater = KavitaPersonUpdater(KAVITA_URL, KAVITA_KEY)
+
+    mal = MALResolver()
+    mal_id = mal.find_mal_id("よふかしのうた")
+    print("MAL ID:", mal_id)
+
+    if mal_id:
+        result = updater.update_for_manga(mal_id)
+        print("Characters:", {k: v for k, v in result["characters"].items()
+                              if k != "errors"})
+        print("Staff     :", {k: v for k, v in result["staff"].items()
+                              if k != "errors"})
+        # Surface any non-fatal upload / API errors for debugging
+        for section in ("characters", "staff"):
+            for err in result[section].get("errors", []):
+                print(f"[{section}] {err}")
@@ -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
@@ -0,0 +1,560 @@
+"""
+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
+
+
+# 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)
+    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 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)
+        self._al  = al_resolver  or AniListResolver(request_timeout=request_timeout)
+        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."""
+        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.request_timeout)
+        resp.raise_for_status()
+        data = resp.json().get("data")
+        if data and data.get("state") == "merged" and data.get("merged_with"):
+            return self.fetch_series(data["merged_with"])
+        return data
+
+    # ------------------------------------------------------------------
+    # 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 / writer name lists for SeriesMetadata
+        character_names = [c["name"] 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
+        publishers = self._publishers_by_type(md, "English") \
+                     or self._publishers_by_type(md, "Original")
+        imprint = None
+        if self._publishers_by_type(md, "English") and \
+           self._publishers_by_type(md, "Original"):
+            imprint = self._publishers_by_type(md, "Original")[0] if \
+                      self._publishers_by_type(md, "Original") 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)
@@ -0,0 +1,257 @@
+"""
+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
+
+        self._mal = MALResolver(request_timeout=request_timeout)
+        self._al  = AniListResolver(request_timeout=request_timeout)
+
+        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(
+            kavita_url, kavita_api_key,
+            mal_resolver=self._mal,
+            al_resolver=self._al,
+            request_timeout=request_timeout,
+        )
+        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)
@@ -0,0 +1,442 @@
+"""
+mal_resolver.py
+===============
+
+Fetches and caches MyAnimeList manga metadata (statistics, characters, staff)
+using the public Jikan REST API v4.
+
+Jikan API: https://api.jikan.moe/v4  (no authentication required)
+Rate limit: 3 req/s, 60 req/min  ->  a 400 ms guard between calls is applied.
+
+Singleton
+---------
+Only one instance of this class exists per process.  Subsequent calls to
+MALResolver() return the same object with its warm caches intact.
+
+Provided features
+-----------------
+- Title-based MAL ID lookup with best-match scoring
+- MAL statistics: score, rank, scored_by, popularity, members, favorites
+- Character list for a manga (names only — for <Characters> XML tag)
+- Detailed character list: name, MAL character ID, image URL, role
+- Detailed staff list: name, MAL person ID, image URL, positions
+- Lazy full-detail fetches per character / person (for descriptions)
+
+Dependencies
+------------
+    requests    ->  pip install requests
+"""
+
+from __future__ import annotations
+
+import datetime
+import difflib
+import time
+
+import requests
+
+from MediaResolver import MediaResolver
+
+
+class MALResolver(MediaResolver):
+    """
+    Singleton: fetches and caches MAL manga data via Jikan API v4.
+
+    The first call to MALResolver() creates and initialises the instance;
+    all subsequent calls return the same object.
+    """
+
+    _instance: "MALResolver | None" = None
+
+    # ------------------------------------------------------------------
+    # Singleton machinery
+    # ------------------------------------------------------------------
+    def __new__(cls, **kwargs):
+        if cls._instance is None:
+            cls._instance = super().__new__(cls)
+            cls._instance._initialized = False
+        return cls._instance
+
+    def __init__(self, *, request_timeout: int = 30):
+        if self._initialized:
+            return
+
+        self.JIKAN_BASE = "https://api.jikan.moe/v4"
+        self.request_timeout = request_timeout
+
+        self._session = requests.Session()
+        self._session.headers.setdefault("User-Agent", "MALResolver/1.0")
+
+        # title_lower -> mal_id
+        self._id_cache: dict[str, "int | None"] = {}
+        # mal_id -> stats dict
+        self._stats_cache: dict[int, dict] = {}
+        # manga_mal_id -> [name_str, ...]           (for ComicInfo <Characters>)
+        self._char_names_cache: dict[int, list[str]] = {}
+        # manga_mal_id -> [{mal_id, name, image_url, role}]
+        self._char_detailed_cache: dict[int, list[dict]] = {}
+        # manga_mal_id -> [{mal_id, name, image_url, positions}]
+        self._staff_detailed_cache: dict[int, list[dict]] = {}
+        # char_mal_id  -> {mal_id, name, image_url, about}
+        self._char_info_cache: dict[int, dict] = {}
+        # person_mal_id -> {mal_id, name, image_url, about, website_url}
+        self._person_info_cache: dict[int, dict] = {}
+
+        self._last_request_at: float = 0.0
+        self._initialized = True
+
+    # ------------------------------------------------------------------
+    # Public: ID lookup
+    # ------------------------------------------------------------------
+    def find_id(self, title: str) -> "int | None":
+        """MediaResolver interface — delegates to find_mal_id."""
+        return self.find_mal_id(title)
+
+    def find_mal_id(self, title: str) -> "int | None":
+        """
+        Searches MAL for a manga by title and returns the best-matching MAL ID.
+        Returns None on failure or when no result is found.
+        """
+        if not title or not title.strip():
+            return None
+
+        key = title.strip().lower()
+        if key in self._id_cache:
+            return self._id_cache[key]
+
+        try:
+            data = self._get(f"{self.JIKAN_BASE}/manga",
+                             {"q": title, "limit": 5, "type": "lightnovel"})
+            results = data.get("data") or []
+        except requests.RequestException:
+            return None
+
+        if not results:
+            self._id_cache[key] = None
+            return None
+
+        results.sort(key=lambda e: _score_title(title, e), reverse=True)
+        mal_id = results[0].get("mal_id")
+        self._id_cache[key] = mal_id
+        return mal_id
+
+    # ------------------------------------------------------------------
+    # Public: statistics
+    # ------------------------------------------------------------------
+    def get_stats(self, mal_id: "int | None") -> "dict | None":
+        """
+        Returns a statistics dict for the given MAL manga ID:
+
+            {score, rank, scored_by, popularity, members, favorites,
+             url, title, as_of (DD-MM-YYYY)}
+
+        Returns None if mal_id is None or on network failure.
+        """
+        if mal_id is None:
+            return None
+        if mal_id in self._stats_cache:
+            return self._stats_cache[mal_id]
+
+        try:
+            data = self._get(f"{self.JIKAN_BASE}/manga/{mal_id}")
+            entry = data.get("data") or {}
+        except requests.RequestException:
+            return None
+
+        stats: dict = {
+            "score":      entry.get("score"),
+            "rank":       entry.get("rank"),
+            "scored_by":  entry.get("scored_by"),
+            "popularity": entry.get("popularity"),
+            "members":    entry.get("members"),
+            "favorites":  entry.get("favorites"),
+            "url":        (entry.get("url")
+                           or f"https://myanimelist.net/manga/{mal_id}"),
+            "title":      entry.get("title") or "",
+            "as_of":      datetime.date.today().strftime("%d-%m-%Y"),
+        }
+        self._stats_cache[mal_id] = stats
+        return stats
+
+    def get_stats_for_manga(self, title: str) -> "dict | None":
+        """Convenience: find MAL ID by title, then return stats."""
+        return self.get_stats(self.find_mal_id(title))
+
+    # ------------------------------------------------------------------
+    # Public: character names (for ComicInfo <Characters> tag)
+    # ------------------------------------------------------------------
+    def get_characters(self, mal_id: "int | None") -> list[str]:
+        """
+        Returns a flat list of character names for the manga.
+        Used by ComicInfoBuilder to populate the <Characters> XML element.
+        """
+        if mal_id is None:
+            return []
+        if mal_id in self._char_names_cache:
+            return self._char_names_cache[mal_id]
+
+        detailed = self.get_characters_detailed(mal_id)
+        names = [e["name"] for e in detailed if e.get("name")]
+        if names:
+            # Only cache a successful result — empty could be a transient
+            # API failure and we want the next call to retry.
+            self._char_names_cache[mal_id] = names
+        return names
+
+    def get_characters_for_manga(self, title: str) -> list[str]:
+        """Convenience: search by title, then return character names."""
+        return self.get_characters(self.find_mal_id(title))
+
+    # ------------------------------------------------------------------
+    # Public: detailed character data (for KavitaPersonUpdater)
+    # ------------------------------------------------------------------
+    def get_characters_detailed(self, mal_id: "int | None") -> list[dict]:
+        """
+        Returns detailed character entries for a manga:
+            [{mal_id, name, image_url, role, about=None}, ...]
+
+        `about` is not populated here; call get_character_details(char_mal_id)
+        to fetch it lazily when needed.
+        """
+        if mal_id is None:
+            return []
+        if mal_id in self._char_detailed_cache:
+            return self._char_detailed_cache[mal_id]
+
+        try:
+            data = self._get(f"{self.JIKAN_BASE}/manga/{mal_id}/characters")
+            entries = data.get("data") or []
+        except requests.RequestException:
+            return []
+
+        results = []
+        for entry in entries:
+            char = entry.get("character") or {}
+            raw_name = char.get("name") or ""
+            if not raw_name:
+                continue
+            jpg = (char.get("images") or {}).get("jpg") or {}
+            results.append({
+                "mal_id":    char.get("mal_id"),
+                # Cleaned name: "Hibino, Susuki" -> "Susuki Hibino".  ComicInfo
+                # <Characters> is comma-separated, so commas in names would
+                # cause Kavita to split a single character into two persons.
+                "name":      _clean_mal_name(raw_name),
+                "raw_name":  raw_name,
+                "image_url": jpg.get("image_url") or jpg.get("small_image_url"),
+                "role":      entry.get("role") or "Supporting",
+                "about":     None,
+            })
+
+        if results:
+            self._char_detailed_cache[mal_id] = results
+        return results
+
+    # ------------------------------------------------------------------
+    # Public: detailed staff data (for KavitaPersonUpdater)
+    # ------------------------------------------------------------------
+    def get_staff_detailed(self, mal_id: "int | None") -> list[dict]:
+        """
+        Returns detailed staff (author) entries for a manga:
+            [{mal_id, name, image_url, positions, about=None}, ...]
+
+        Jikan has no `/manga/{id}/staff` endpoint — that route only exists for
+        anime.  For manga the authors are listed on `/manga/{id}` under
+        `data.authors`, but each entry only has {mal_id, name, url}; the image
+        URL is fetched lazily via get_person_details (cached, so the later
+        description fetch is free).
+        """
+        if mal_id is None:
+            return []
+        if mal_id in self._staff_detailed_cache:
+            return self._staff_detailed_cache[mal_id]
+
+        try:
+            data = self._get(f"{self.JIKAN_BASE}/manga/{mal_id}")
+            entry = data.get("data") or {}
+        except requests.RequestException:
+            return []
+
+        results = []
+        for author in (entry.get("authors") or []):
+            raw_name = author.get("name") or ""
+            person_mal_id = author.get("mal_id")
+            if not raw_name or person_mal_id is None:
+                continue
+            details = self.get_person_details(person_mal_id) or {}
+            results.append({
+                "mal_id":    person_mal_id,
+                "name":      _clean_mal_name(raw_name),
+                "raw_name":  raw_name,
+                "image_url": details.get("image_url"),
+                "positions": [],
+                "about":     None,
+            })
+
+        if results:
+            self._staff_detailed_cache[mal_id] = results
+        return results
+
+    # ------------------------------------------------------------------
+    # Public: individual character / person details (lazy, with description)
+    # ------------------------------------------------------------------
+    def get_character_details(self, char_mal_id: "int | None") -> "dict | None":
+        """
+        Returns full details for a single MAL character, including `about`.
+        Result is cached.
+        """
+        if char_mal_id is None:
+            return None
+        if char_mal_id in self._char_info_cache:
+            return self._char_info_cache[char_mal_id]
+
+        try:
+            data = self._get(f"{self.JIKAN_BASE}/characters/{char_mal_id}")
+            entry = data.get("data") or {}
+        except requests.RequestException:
+            return None
+
+        jpg = (entry.get("images") or {}).get("jpg") or {}
+        result = {
+            "mal_id":    entry.get("mal_id"),
+            "name":      entry.get("name") or "",
+            "image_url": jpg.get("image_url") or jpg.get("small_image_url"),
+            "about":     entry.get("about"),
+            "favorites": entry.get("favorites"),
+            "url":       (entry.get("url")
+                          or f"https://myanimelist.net/character/{char_mal_id}"),
+        }
+        self._char_info_cache[char_mal_id] = result
+        return result
+
+    def get_person_details(self, person_mal_id: "int | None") -> "dict | None":
+        """
+        Returns full details for a single MAL person (staff), including `about`.
+        Result is cached.
+        """
+        if person_mal_id is None:
+            return None
+        if person_mal_id in self._person_info_cache:
+            return self._person_info_cache[person_mal_id]
+
+        try:
+            data = self._get(f"{self.JIKAN_BASE}/people/{person_mal_id}")
+            entry = data.get("data") or {}
+        except requests.RequestException:
+            return None
+
+        jpg = (entry.get("images") or {}).get("jpg") or {}
+        result = {
+            "mal_id":      entry.get("mal_id"),
+            "name":        entry.get("name") or "",
+            "given_name":  entry.get("given_name"),
+            "family_name": entry.get("family_name"),
+            "birthday":    entry.get("birthday"),
+            "image_url":   jpg.get("image_url") or jpg.get("small_image_url"),
+            "about":       entry.get("about"),
+            "favorites":   entry.get("favorites"),
+            "website_url": entry.get("website_url"),
+            "url":         (entry.get("url")
+                            or f"https://myanimelist.net/people/{person_mal_id}"),
+        }
+        self._person_info_cache[person_mal_id] = result
+        return result
+
+    # ------------------------------------------------------------------
+    # Public: cache management
+    # ------------------------------------------------------------------
+    def clear_cache(self) -> None:
+        """Clears all internal caches (the Singleton instance is retained)."""
+        self._id_cache.clear()
+        self._stats_cache.clear()
+        self._char_names_cache.clear()
+        self._char_detailed_cache.clear()
+        self._staff_detailed_cache.clear()
+        self._char_info_cache.clear()
+        self._person_info_cache.clear()
+
+    # ------------------------------------------------------------------
+    # Internal: rate-limited HTTP
+    # ------------------------------------------------------------------
+    def _get(self, url: str, params: "dict | None" = None) -> dict:
+        """Rate-limited GET request (respects Jikan's ~3 req/s limit)."""
+        elapsed = time.monotonic() - self._last_request_at
+        if elapsed < 0.4:
+            time.sleep(0.4 - elapsed)
+        resp = self._session.get(url, params=params, timeout=self.request_timeout)
+        self._last_request_at = time.monotonic()
+        resp.raise_for_status()
+        return resp.json()
+
+
+# --------------------------------------------------------------------------
+# Module helper
+# --------------------------------------------------------------------------
+def _clean_mal_name(name: str) -> str:
+    """
+    Converts an MAL name into a comma-free, ComicInfo-safe form.
+
+    The ComicInfo <Characters> tag is comma-separated, so a single MAL
+    character "Hibino, Susuki" written into the XML would be parsed by
+    Kavita as two persons ("Hibino" and "Susuki").
+
+    Conversion:
+      "Hibino, Susuki"   ->  "Susuki Hibino"   (Western: First Last)
+      "Yamori, Kou"      ->  "Kou Yamori"
+      "Kotoyama"         ->  "Kotoyama"        (unchanged)
+
+    Trailing/leading commas and stray whitespace are stripped defensively.
+    """
+    if not name:
+        return ""
+    name = name.strip()
+    if "," in name:
+        last, _, first = name.partition(",")
+        first = first.strip()
+        last = last.strip()
+        if first and last:
+            return f"{first} {last}"
+        # Fallback: strip any remaining commas
+        return name.replace(",", " ").strip()
+    return name
+
+
+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 "",
+    ]
+    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
+
+
+# --------------------------------------------------------------------------
+# Usage example
+# --------------------------------------------------------------------------
+if __name__ == "__main__":
+    r1 = MALResolver()
+    r2 = MALResolver()
+    assert r1 is r2, "MALResolver must be a Singleton"
+
+    mal_id = r1.find_mal_id("Yofukashi no Uta")
+    print("MAL ID       :", mal_id)
+
+    stats = r1.get_stats(mal_id)
+    if stats:
+        print("Score        :", stats["score"])
+        print("Rank         :", stats["rank"])
+
+    chars = r1.get_characters_detailed(mal_id)
+    print("Characters (first 3):", [c["name"] for c in chars[:3]])
+
+    staff = r1.get_staff_detailed(mal_id)
+    print("Staff        :", [s["name"] for s in staff])
@@ -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)
@@ -0,0 +1,195 @@
+"""
+mangabaka_works_resolver.py
+===========================
+
+Fetches volume-level (work) data from the MangaBaka API.
+
+Each "work" is a physical tankobon volume and may carry:
+  - volume number
+  - ISBN / GTIN
+  - page count   (used for chapter-to-volume estimation)
+  - 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.
+
+Dependencies
+------------
+    requests    ->  pip install requests
+"""
+
+from __future__ import annotations
+
+import requests
+
+
+class MangaBakaWorksResolver:
+    """
+    Fetches and caches MangaBaka volume (work) data for a series.
+    Only works that have a cover image are retained in the cache.
+    """
+
+    def __init__(self, api_base_url: str = "https://api.mangabaka.dev/v1",
+                 request_timeout: int = 30,
+                 session: "requests.Session | None" = None):
+        self.api_base_url = api_base_url.rstrip("/")
+        self.request_timeout = request_timeout
+        self._session = session or requests.Session()
+        self._session.headers.setdefault("User-Agent", "MangaBakaWorksResolver/1.0")
+
+        # Cache: series_id (str) -> list of work dicts (only those with covers)
+        self._cache: dict[str, list[dict]] = {}
+
+    # ------------------------------------------------------------------
+    # Public API
+    # ------------------------------------------------------------------
+    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.  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.
+        """
+        if not series_id:
+            return []
+
+        if series_id in self._cache:
+            return self._cache[series_id]
+
+        all_works: list[dict] = []
+        page = 1
+        try:
+            while True:
+                resp = self._session.get(
+                    f"{self.api_base_url}/series/{series_id}/works",
+                    params={"limit": 50, "page": page},
+                    timeout=self.request_timeout,
+                )
+                resp.raise_for_status()
+                page_data = resp.json().get("data") or []
+                if not page_data:
+                    break
+                all_works.extend(page_data)
+                if len(page_data) < 50:
+                    break
+                page += 1
+        except requests.RequestException:
+            if not all_works:
+                return []
+
+        # Discard works that carry no usable cover
+        works_with_cover = [w for w in all_works if w.get("images")]
+        self._cache[series_id] = works_with_cover
+        return works_with_cover
+
+    def get_work_for_volume(self, series_id: str, volume) -> "dict | None":
+        """
+        Returns the work dict for a specific volume number, or None.
+        Volume comparison normalises trailing ".0" (e.g. "1.0" == "1").
+        """
+        works = self.get_works(series_id)
+        if not works:
+            return None
+
+        target = _norm_vol(volume)
+        for work in works:
+            if _norm_vol(work.get("sequence_string")) == target:
+                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:
+            return None
+        return self._pick_cover_url(work.get("images")[0].get("image"))
+
+    def get_page_counts(self, series_id: str) -> "dict[str, int]":
+        """
+        Returns {volume_str: page_count} for all cached works.
+        Used by MangaDexVolumeResolver for chapter-to-volume estimation.
+        """
+        result: dict[str, int] = {}
+        for work in self.get_works(series_id):
+            vol = _norm_vol(work.get("volume"))
+            pages = work.get("pages")
+            if vol and pages is not None:
+                try:
+                    result[vol] = int(pages)
+                except (TypeError, ValueError):
+                    pass
+        return result
+
+    def clear_cache(self) -> None:
+        """Clears the internal works 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
+
+
+# --------------------------------------------------------------------------
+# Module helper
+# --------------------------------------------------------------------------
+def _norm_vol(value) -> str:
+    """Normalises a volume identifier: strips whitespace, removes trailing .0."""
+    text = str(value or "").strip()
+    try:
+        f = float(text)
+        if f.is_integer():
+            return str(int(f))
+    except ValueError:
+        pass
+    return text
@@ -0,0 +1,191 @@
+"""
+matches_cache.py
+================
+
+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": {
+        "<kavita series name>": {
+          "mangabakaId":      "12345",
+          "mangabakaName":    "Re:Zero",
+          "imageUrl":         "https://.../cover.jpg",
+          "kavitaSeriesId":   42,
+          "libraryId":        3,
+          "firstMatchTime":   1700000000,
+          "lastUpdateTime":   1700100000
+        },
+        ...
+      }
+    }
+
+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
+
+import json
+import threading
+import time
+from pathlib import Path
+
+
+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(title)
+            return dict(entry) if entry else None
+
+    def get_by_kavita_id(self, kavita_series_id: int) -> "tuple[str, dict] | None":
+        with self._lock:
+            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,
+               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:
+                entry = {
+                    "mangabakaId":    "",
+                    "mangabakaName":  "",
+                    "imageUrl":       "",
+                    "kavitaSeriesId": 0,
+                    "libraryId":      0,
+                    "firstMatchTime": int(time.time()),
+                    "lastUpdateTime": 0,
+                }
+                self._data["matches"][title] = entry
+            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 kavita_series_id is not None:
+                try:
+                    entry["kavitaSeriesId"] = int(kavita_series_id)
+                except (TypeError, ValueError):
+                    pass
+            if library_id is not None:
+                try:
+                    entry["libraryId"] = int(library_id)
+                except (TypeError, ValueError):
+                    pass
+            if first_match_time is not None:
+                try:
+                    entry["firstMatchTime"] = int(first_match_time)
+                except (TypeError, ValueError):
+                    pass
+            if last_update_time is not None:
+                try:
+                    entry["lastUpdateTime"] = int(last_update_time)
+                except (TypeError, ValueError):
+                    pass
+            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
+        with self._lock:
+            entry = self._data["matches"].pop(old_title, None)
+            if entry is None:
+                return False
+            self._data["matches"][new_title] = entry
+            self._save_unlocked()
+            return True
+
+    def remove(self, title: str) -> bool:
+        with self._lock:
+            existed = title in self._data["matches"]
+            if existed:
+                del self._data["matches"][title]
+                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()}}
+
+    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
+    # ------------------------------------------------------------------
+    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 isinstance(loaded, dict) and isinstance(loaded.get("matches"), dict):
+            self._data = loaded
+
+    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)
@@ -0,0 +1,757 @@
+"""
+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
+
+
+_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(p) for p in raw.split(",") if p.strip().isdigit()]
+            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(i) for i in (body.get("libraryIds") or [])
+                           if str(i).strip().lstrip("-").isdigit()]
+            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: "list[int] | None"
+            if raw is None:
+                library_ids = None
+            else:
+                library_ids = [int(i) for i in raw
+                               if str(i).strip().lstrip("-").isdigit()]
+
+            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)
@@ -0,0 +1,91 @@
+"""
+media_resolver.py
+=================
+
+Abstract base class for tracker-specific manga metadata resolvers.
+
+Concrete implementations (MALResolver, AniListResolver) must implement
+every abstract method, ensuring a uniform interface regardless of the
+underlying data source (Jikan/MAL, AniList GraphQL, …).
+"""
+
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+
+
+class MediaResolver(ABC):
+    """
+    Abstract base for tracker-specific manga metadata resolvers.
+
+    Subclasses connect to a specific tracker API and expose a common
+    interface for:
+      - Searching a manga by title → tracker-specific numeric ID
+      - Fetching summary statistics (score, rank, popularity, …)
+      - Listing characters and staff (name-only and detailed forms)
+      - Fetching full details for a single character or person
+
+    Methods that accept a tracker ID treat None as "unknown" and return
+    a safe empty value rather than raising.
+    """
+
+    @abstractmethod
+    def find_id(self, title: str) -> "int | None":
+        """
+        Searches the tracker for a manga by title.
+        Returns the best-matching tracker ID, or None on failure.
+        """
+
+    @abstractmethod
+    def get_stats(self, tracker_id: "int | None") -> "dict | None":
+        """
+        Returns a statistics dict for the given tracker ID:
+
+            {score, rank, scored_by, popularity, members, favorites,
+             url, title, as_of (DD-MM-YYYY)}
+
+        Returns None if tracker_id is None or on network failure.
+        """
+
+    @abstractmethod
+    def get_characters(self, tracker_id: "int | None") -> "list[str]":
+        """
+        Returns a flat list of character name strings for the manga.
+        Used to populate the ComicInfo <Characters> XML element.
+        """
+
+    @abstractmethod
+    def get_characters_detailed(self, tracker_id: "int | None") -> "list[dict]":
+        """
+        Returns detailed character entries for a manga:
+            [{id, name, image_url, role, about=None, ...}, ...]
+
+        'about' is not populated here; call get_character_details() lazily.
+        """
+
+    @abstractmethod
+    def get_staff_detailed(self, tracker_id: "int | None") -> "list[dict]":
+        """
+        Returns detailed staff/author entries for a manga:
+            [{id, name, image_url, positions, about=None, ...}, ...]
+
+        'about' is not populated here; call get_person_details() lazily.
+        """
+
+    @abstractmethod
+    def get_character_details(self, char_id: "int | None") -> "dict | None":
+        """
+        Returns full details for a single character, including description.
+        Implementations should cache the result.
+        """
+
+    @abstractmethod
+    def get_person_details(self, person_id: "int | None") -> "dict | None":
+        """
+        Returns full details for a single person (staff), including description.
+        Implementations should cache the result.
+        """
+
+    @abstractmethod
+    def clear_cache(self) -> None:
+        """Clears all internal caches."""
@@ -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 ""