feat: rename to manga-chapters and use openai api to allow for swapping in another provider

Author: RobBrazier

plugins/manga-toc/README.md plugins/manga-chapters/README.mdplugins/manga-toc/README.md renamed to plugins/manga-chapters/README.md

@@ -3,20 +3,20 @@ requires = ["setuptools"]               #, "setuptools-scm"]
 build-backend = "setuptools.build_meta"
 
 [project]
-name = "manga-toc"
+name = "manga-chapters"
 version = "0.0.0"
 # dynamic = ["version"]
 description = "Editor Plugin to generate Manga Table of Contents from an Image Contents page"
 readme = "README.md"
 authors = [{ name = "Rob Brazier", email = "git+github@brzr.co" }]
-requires-python = ">=3.9"
-dependencies = ["google-genai", "typing-extensions"]
+requires-python = ">=3.11"
+dependencies = ["openai>=1.74.0", "pillow>=11.2.1"]
 
 [tool.setuptools_scm]
 fallback_version = "0.0.0+unknown"
-version_file = "src/manga-toc/_version.py"
+version_file = "src/manga-chapters/_version.py"
 root = "../../"
-tag_regex = '^manga-toc-(?P<version>\d+(?:\.\d+){0,2})$'
+tag_regex = '^manga-chapters-(?P<version>\d+(?:\.\d+){0,2})$'
 version_file_template = '''
 __version__ = "{version}"
 __version_tuple__ = {version_tuple}

plugins/manga-toc/pyproject.toml plugins/manga-chapters/pyproject.tomlplugins/manga-toc/pyproject.toml renamed to plugins/manga-chapters/pyproject.toml

@@ -0,0 +1,40 @@
+from qt.core import QWidget, QVBoxLayout, QLabel, QLineEdit
+from calibre.utils.config import JSONConfig
+
+# Create a configuration
+prefs = JSONConfig("plugins/manga_toc_generator")
+
+# Set defaults
+prefs.defaults["llm_endpoint"] = (
+    "https://generativelanguage.googleapis.com/v1beta/openai/"
+)
+prefs.defaults["llm_model"] = "gemini-2.0-flash"
+prefs.defaults["api_key"] = ""
+
+
+class ConfigWidget(QWidget):
+    def __init__(self):
+        QWidget.__init__(self)
+        self.layout = QVBoxLayout()
+        self.setLayout(self.layout)
+
+        # API Key
+        self.layout.addWidget(QLabel("OpenAI-Compatible Endpoint:"))
+        self.llm_endpoint = QLineEdit(self)
+        self.llm_endpoint.setText(prefs["llm_endpoint"])
+        self.layout.addWidget(self.llm_endpoint)
+        self.layout.addWidget(QLabel("LLM Model Identifier:"))
+        self.llm_model = QLineEdit(self)
+        self.llm_model.setText(prefs["llm_model"])
+        self.layout.addWidget(self.llm_model)
+        self.layout.addWidget(QLabel("API Key"))
+        self.api_key = QLineEdit(self)
+        self.api_key.setText(prefs["api_key"])
+        self.layout.addWidget(self.api_key)
+
+        self.layout.addStretch(1)
+
+    def save_settings(self):
+        prefs["llm_endpoint"] = self.llm_endpoint.text().strip()
+        prefs["llm_model"] = self.llm_model.text().strip()
+        prefs["api_key"] = self.api_key.text().strip()

…gins/manga-toc/src/manga_toc/__init__.py …-chapters/src/manga_chapters/__init__.pyplugins/manga-toc/src/manga_toc/__init__.py renamed to plugins/manga-chapters/src/manga_chapters/__init__.py plugins/manga-toc/src/manga_toc/__init__.py renamed to plugins/manga-chapters/src/manga_chapters/__init__.py

@@ -0,0 +1,47 @@
+from io import BytesIO
+import base64
+import openai
+from PIL import Image
+from pydantic import BaseModel
+
+PROMPT = """
+Look at the image and output the chapters that you see. Focus only on the text content visible in the image. Do not generate any content that isn't visible in the image.
+Format the output in Title Case. Do not assume the ordering based on the first number that is read, only use numbering that is visible.
+
+For numbered chapters (numbered being either numeric (e.g. 1,2,3) or roman numerals (e.g. I,II,X), format as 'Chapter [number]: [title]'. Ensure that all roman numerals are converted to their numerical equivalents.
+For other chapters without numbers, format as '[category]: [title]' when a category is present. Omit the category when not.
+"""
+
+
+class ChapterResponse(BaseModel):
+    chapters: list[str]
+
+
+class LLMReader:
+    def __init__(self, url: str, model: str, api_key: str) -> None:
+        self.client = openai.OpenAI(api_key=api_key, base_url=url)
+        self.model = model
+
+    @staticmethod
+    def _trim(input: list[str]) -> list[str]:
+        return filter(None, [line.trim() for line in input])
+
+    def read_chapters(self, image_bytes: bytes) -> list[str]:
+        image = Image.open(BytesIO(image_bytes))
+        encoded_image = base64.b64encode(image_bytes).decode("utf-8")
+        image_url = f"data:{image.get_format_mimetype()};base64,{encoded_image}"
+        response = self.client.beta.chat.completions.parse(
+            model=self.model,
+            messages=[
+                {
+                    "role": "user",
+                    "content": [
+                        {"type": "text", "text": PROMPT},
+                        {"type": "image_url", "image_url": {"url": image_url}},
+                    ],
+                },
+            ],
+            response_format=ChapterResponse,
+        )
+        chapters = response.choices[0].message.parsed
+        return chapters.chapters

plugins/manga-chapters/src/manga_chapters/config.py

@@ -1,12 +1,12 @@
 import os
-from io import BytesIO
 
 from calibre.customize import Plugin
 from calibre.ebooks.oeb.polish.toc import get_toc, commit_toc
 from calibre.gui2 import error_dialog, question_dialog
 from calibre.gui2.toc.main import TOC
 from calibre.gui2.tweak_book.plugin import Tool
 from qt.core import QAction
+from .config import prefs
 
 
 class MangaTocTool(Tool):
@@ -16,6 +16,7 @@ class MangaTocTool(Tool):
 
     def __init__(self):
         self.plugin_path = os.path.dirname(os.path.abspath(__file__))
+        self.prefs = prefs
 
     def create_action(self, for_toolbar=True):
         action = QAction(get_icons("images/chapters.png"), _("Generate ToC"), self.gui)
@@ -30,18 +31,6 @@ def __enter__(self, *args):
     def __exit__(self, *args):
         Plugin.__exit__(self, *args)
 
-    def _get_client(self):
-        from .config import prefs
-
-        api_key = prefs["api_key"]
-        if not api_key:
-            raise Exception(
-                "A Gemini API Key is required - please configure in settings"
-            )
-        from google import genai
-
-        return genai.Client(api_key=api_key)
-
     @staticmethod
     def _normalise_path(base, path) -> str:
         base_dir = os.path.dirname(base)
@@ -77,28 +66,17 @@ def parse_links(self, toc, container) -> tuple[str, list[str], int]:
         ]
         return image, links, contents_index
 
-    def _get_image_contents(self, container, path):
-        from PIL import Image
-
-        data = container.raw_data(path, decode=False)
-        image = Image.open(BytesIO(data))
-        return image
-
-    def _gemini_read_chapters(self, image) -> list[str]:
-        client = self._get_client()
-        result = client.models.generate_content(
-            model="gemini-2.0-flash",
-            contents=[
-                image,
-                "\n\n",
-                "Please can you extract the chapter names from this attached image. Only return the chapter names, no page numbers or any other information that is present in the image. Just chapter names.\n",
-                "Convert the text to Title Case and ensure that the chapters are numerically numbered (converting other schemes into numbers).\n",
-                "e.g. CH102 -> Chapter 102, Chapter I -> Chapter 1, Chapter X -> Chapter 10.\n",
-                "In the case of additional chapters e.g. 'Extra Story', 'Origial Story', 'Bonus' etc., leave those as-is, but still performing the Title Case transformations",
-                "Do not format the response in any way, just do one chapter per line.",
-            ],
-        )
-        return result.text.splitlines()
+    def _get_image_contents(self, container, path) -> bytes:
+        return container.raw_data(path, decode=False)
+
+    def _read_chapters(self, image: bytes) -> list[str]:
+        from .llm import LLMReader
+
+        url = self.prefs["llm_endpoint"]
+        model = self.prefs["llm_model"]
+        api_key = self.prefs["api_key"]
+        reader = LLMReader(url, model, api_key)
+        return reader.read_chapters(image)
 
     def _confirm_apply(self, changes):
         mappings_string = "\n".join(changes)
@@ -134,7 +112,7 @@ def generate_toc(self):
                 toc = get_toc(container)
                 image, links, contents_idx = self.parse_links(toc, container)
                 contents_image = self._get_image_contents(container, image)
-                chapters = self._gemini_read_chapters(contents_image)
+                chapters = self._read_chapters(contents_image)
                 entries = {}
                 if len(links) != len(chapters):
                     raise Exception(