codebee 0.1.2 → 0.1.4

package/app/core/bookmeta.py

@@ -43,12 +43,15 @@ def needs_book_meta(task):
         return True
 
 
-# 各平台建书表单的字段展示顺序（前端弹框与 Markdown 归档共用）
+# 各平台建书表单的字段展示顺序（前端卡片与 Markdown 归档共用）。
+# 番茄官方表单：目标读者 → 分类 → 标签（主题/角色/情节 三组多选）；
+# 签约模式官方叫法「连载签约/完本签约」。
 FIELD_LABELS = {
     "fanqie": [
         ("book_name", "作品名"), ("signing_mode", "签约模式"),
-        ("target_reader", "目标读者"), ("read_tags", "阅读标签"),
-        ("content_tags", "内容标签"), ("protagonist_1", "主角名1"),
+        ("target_reader", "目标读者"), ("category", "分类"),
+        ("tags_theme", "主题标签"), ("tags_role", "角色标签"),
+        ("tags_plot", "情节标签"), ("protagonist_1", "主角名1"),
         ("protagonist_2", "主角名2"), ("summary", "作品简介"),
     ],
     "qimao": [
@@ -68,21 +71,59 @@ __SKILLS__
 不要输出其他内容。JSON 结构：
 __SCHEMA__
 硬性要求：
-- 字段值必须贴合素材内容，不编造素材里没有的设定；简介 50-500 字，突出卖点与钩子；
-- 标签选平台热门且与题材匹配的，不要堆砌；
+- 字段值必须贴合素材内容，不编造素材里没有的设定；简介不超过 500 字，突出卖点与钩子；
+- __OPTIONS_RULE__
 - 题材健康，无违规内容，符合平台签约调性。
 """
 
-FANQIE_SCHEMA = ('{"book_name": "作品名(15字内)", "signing_mode": "连载模式 或 完本模式", '
-                 '"target_reader": "男频 或 女频", "read_tags": ["阅读标签", …最多6个], '
-                 '"content_tags": ["内容标签", …最多6个], "protagonist_1": "主角名1(5字内)", '
-                 '"protagonist_2": "主角名2(5字内，没有就空串)", "summary": "作品简介(50-500字)"}')
-
-QIMAO_SCHEMA = ('{"book_name": "作品名称(18字内)", "target_reader": "男生 或 女生", '
-                '"category_main": "一级分类", "category_sub": "二级分类", '
-                '"tags": ["作品标签", …最多6个], "protagonist_1": "主角名1(5字内)", '
+# 签约模式官方叫法（番茄帮助中心）：连载签约 / 完本签约
+FANQIE_MODES = ("连载签约", "完本签约")
+
+FANQIE_SCHEMA = ('{"book_name": "作品名(15字内)", "signing_mode": "连载签约 或 完本签约", '
+                 '"target_reader": "男频 或 女频", "category": "分类(从下方分类表选一个)", '
+                 '"tags_theme": ["主题标签", …最多3个], "tags_role": ["角色标签", …最多3个], '
+                 '"tags_plot": ["情节标签", …最多5个], '
+                 '"protagonist_1": "主角名1(5字内)", '
+                 '"protagonist_2": "主角名2(5字内，没有就空串)", "summary": "作品简介(500字内)"}')
+
+QIMAO_SCHEMA = ('{"book_name": "作品名称(15字内)", "target_reader": "男生 或 女生", '
+                '"category_main": "一级分类(从下方一级分类表选一个)", '
+                '"category_sub": "二级分类(必须是该一级下的二级)", '
+                '"tags": ["作品标签", …最多6个，从标签池选], '
+                '"protagonist_1": "主角名1(5字内)", '
                 '"protagonist_2": "主角名2(5字内，没有就空串)", "status": "连载中 或 已完结", '
-                '"summary": "作品简介(50-500字)"}')
+                '"summary": "作品简介(200字内)"}')
+
+
+def _fq_tables(reader):
+    """番茄按目标读者取对应选项表（男频/女频两套完全独立的分类与标签）。"""
+    from . import bookmeta_catalog as cat
+    if reader == "女频":
+        return (cat.FANQIE_FEMALE_CATEGORIES, cat.FANQIE_FEMALE_TAGS_THEME,
+                cat.FANQIE_FEMALE_TAGS_ROLE, cat.FANQIE_FEMALE_TAGS_PLOT)
+    return (cat.FANQIE_MALE_CATEGORIES, cat.FANQIE_MALE_TAGS_THEME,
+            cat.FANQIE_MALE_TAGS_ROLE, cat.FANQIE_MALE_TAGS_PLOT)
+
+
+def _fq_options_block(goal):
+    """番茄提示词的真实选项注入：按 goal 推断读者给出对应分类与三组标签表。"""
+    reader = "女频" if "女" in (goal or "") else "男频"
+    cats, theme, role, plot = _fq_tables(reader)
+    return ("目标读者（%s）可用的分类表：%s\n"
+            "标签分三组多选——主题组：%s\n角色组：%s\n情节组（部分，选最贴合的）：%s"
+            % (reader, "、".join(cats), "、".join(theme), "、".join(role),
+               "、".join(plot[:50])))
+
+
+def _qm_options_block():
+    """七猫提示词的真实选项注入：一级分类 + 已确认级联 + 标签池。"""
+    from . import bookmeta_catalog as cat
+    cascade = ["%s（二级：%s）" % (k, "、".join(v)) for k, v in cat.QIMAO_CASCADE.items()]
+    pool = ["%s组：%s" % (g, "、".join(t)) for g, t in cat.QIMAO_TAG_GROUPS.items()]
+    return ("一级分类表：%s\n一级→二级级联（已确认部分，二级必须从所属一级下选）：%s\n"
+            "标签池（选最贴合的 3-6 个）：%s"
+            % ("、".join(cat.QIMAO_MAIN_CATEGORIES), "；".join(cascade),
+               "；".join(pool)))
 
 
 def _collect_material(task):
@@ -141,11 +182,34 @@ def _summary(v):
     return str(v or "").strip()[:500]
 
 
+def _in_table(v, table, limit=10):
+    """值必须在平台真实选项表内；不在返回空串（宁缺勿错——填了假选项用户还得手动改）。"""
+    s = _name(v, limit)
+    return s if s in (table or []) else ""
+
+
+def _filter_table(v, table, limit=6):
+    """多选标签：只保留表内项（保持模型给的顺序），去重截断。"""
+    if not isinstance(v, list):
+        return []
+    out = []
+    for x in v:
+        s = str(x).strip()[:12]
+        if s and s in (table or []) and s not in out:
+            out.append(s)
+        if len(out) >= limit:
+            break
+    return out
+
+
 def _norm(data, platform, goal=""):
-    """规范化平台字段输出；不合规返回 None。"""
+    """规范化平台字段输出；不合规返回 None。
+
+    分类/标签按平台真实选项表校验（bookmeta_catalog）：分类单选不在表内清空、
+    标签多选只留表内项；七猫二级按级联校验（一级无级联记录时宽松放行）。"""
     if not isinstance(data, dict):
         return None
-    book_name = _name(data.get("book_name"), 15 if platform == "fanqie" else 18)
+    book_name = _name(data.get("book_name"), 15 if platform == "fanqie" else 15)
     summary = _summary(data.get("summary"))
     if not book_name and not summary:
         return None
@@ -154,23 +218,34 @@ def _norm(data, platform, goal=""):
         if reader not in ("男频", "女频"):
             reader = "女频" if "女" in (goal or "") else "男频"
         mode = str(data.get("signing_mode") or "").strip()
-        if mode not in ("连载模式", "完本模式"):
-            mode = "连载模式"
+        if mode not in FANQIE_MODES:
+            mode = "连载签约"
+        cats, theme, role, plot = _fq_tables(reader)
+        # 兼容旧字段：read_tags/content_tags 并入情节/主题组（老数据不丢）
+        legacy_read = data.get("read_tags") if isinstance(data.get("read_tags"), list) else []
         return {"book_name": book_name, "signing_mode": mode, "target_reader": reader,
-                "read_tags": _tags(data.get("read_tags")),
-                "content_tags": _tags(data.get("content_tags")),
+                "category": _in_table(data.get("category"), cats),
+                "tags_theme": _filter_table(data.get("tags_theme"), theme),
+                "tags_role": _filter_table(data.get("tags_role"), role),
+                "tags_plot": _filter_table(data.get("tags_plot"), plot),
+                "legacy_read_tags": _filter_table(legacy_read, plot + theme + role),
                 "protagonist_1": _name(data.get("protagonist_1"), 5),
                 "protagonist_2": _name(data.get("protagonist_2"), 5),
                 "summary": summary}
+    from . import bookmeta_catalog as cat
     reader = str(data.get("target_reader") or "").strip()
     if reader not in ("男生", "女生"):
         reader = "女生" if "女" in (goal or "") else "男生"
     status = str(data.get("status") or "").strip()
     if status not in ("连载中", "已完结"):
         status = "连载中"
+    c_main = _in_table(data.get("category_main"), cat.QIMAO_MAIN_CATEGORIES)
+    c_sub = _name(data.get("category_sub"), 10)
+    subs = cat.QIMAO_CASCADE.get(c_main)
+    if subs and c_sub not in subs:
+        c_sub = ""     # 级联不匹配清空（二级必须挂在一级下）
     return {"book_name": book_name, "target_reader": reader,
-            "category_main": _name(data.get("category_main"), 10),
-            "category_sub": _name(data.get("category_sub"), 10),
+            "category_main": c_main, "category_sub": c_sub,
             "tags": _tags(data.get("tags")),
             "protagonist_1": _name(data.get("protagonist_1"), 5),
             "protagonist_2": _name(data.get("protagonist_2"), 5),
@@ -178,13 +253,69 @@ def _norm(data, platform, goal=""):
 
 
 def _template_meta(task, platform, outline):
-    """兜底模板：只做素材搬运归纳（书名/简介来自大纲与目标），其余留空待补。"""
+    """兜底模板：只做素材搬运归纳（书名/简介来自大纲与目标），其余留空待补。
+
+    标签/分类尽力而为：扫描素材文本（goal/圣经/大纲书名）里恰好命中平台
+    选项表的词就带上——模板没有模型理解力，不做语义归纳，命中多少算多少，
+    其余留「待补充」让用户手选（宁缺勿错，不填表外词）。"""
     goal = task.get("goal") or ""
     book_name = str((outline or {}).get("book_title") or task.get("title") or "").strip()
     summary = goal.splitlines()[0][:500] if goal.strip() else ""
     if not summary:
         summary = book_name or "（待补充）"    # 全空素材也要有返回，_norm 才不会拒绝
-    return _norm({"book_name": book_name, "summary": summary}, platform, goal)
+    material = goal
+    try:
+        _, bible, _ = store_read_bible(task)
+        material += "\n" + (bible or "")
+    except Exception:
+        pass
+    if outline:
+        material += "\n" + str(outline.get("book_title") or "")
+    data = {"book_name": book_name, "summary": summary}
+    if platform == "fanqie":
+        reader = "女频" if "女" in goal else "男频"
+        cats, theme, role, plot = _fq_tables(reader)
+        data.update({"target_reader": reader,
+                     "category": _scan_table(material, cats, 1),
+                     "tags_theme": _scan_table(material, theme, 3),
+                     "tags_role": _scan_table(material, role, 3),
+                     "tags_plot": _scan_table(material, plot, 5)})
+    else:
+        from . import bookmeta_catalog as cat
+        data["target_reader"] = "女生" if "女" in goal else "男生"
+        for main, subs in cat.QIMAO_CASCADE.items():
+            if main in material:
+                data["category_main"] = main
+                for s in subs:               # 二级名撞素材即选（如「宫闱宅斗」）
+                    if s in material:
+                        data["category_sub"] = s
+                        break
+                break
+        pool = [t for ts in cat.QIMAO_TAG_GROUPS.values() for t in ts]
+        data["tags"] = _scan_table(material, pool, 6)
+    return _norm(data, platform, goal)
+
+
+def _scan_table(text, table, limit):
+    """素材文本里命中的选项表词（按表序取，保证真实存在）。"""
+    if not text:
+        return []
+    out = []
+    for word in table:
+        if len(word) >= 2 and word in text:
+            out.append(word)
+            if len(out) >= limit:
+                break
+    return out
+
+
+def store_read_bible(task):
+    """圣经读取的小包装（模板兜底用；异常向上抛由调用方吞掉）。"""
+    from . import store
+    fp, text, err = store.read_story_bible(task.get("id") or "")
+    if err:
+        raise ValueError(err)
+    return fp, text, None
 
 
 def _resolve_author(task):
@@ -206,10 +337,13 @@ def make_book_meta(task, platform, author_agent=None, log_path=None):
     schema = {"fanqie": FANQIE_SCHEMA, "qimao": QIMAO_SCHEMA}.get(platform) or ""
     plat_label = PLATFORMS[platform]["label"]
     sk_block, _ = skills.block_for(task)
+    goal = task.get("goal") or ""
+    # 真实选项注入：分类/标签表按平台与读者频段给全（模型只能从表里选）
+    options = (_fq_options_block(goal) if platform == "fanqie" else _qm_options_block())
     prompt = (PROMPT_HEAD.replace("__PLATFORM__", plat_label)
               .replace("__SKILLS__", sk_block)
-              .replace("__SCHEMA__", schema) + "\n\n## 小说素材\n" + material)
-    goal = task.get("goal") or ""
+              .replace("__SCHEMA__", schema)
+              .replace("__OPTIONS_RULE__", options) + "\n\n## 小说素材\n" + material)
 
     orch = None
     if not template_only():
@@ -264,7 +398,7 @@ def render_markdown(task, platform, meta):
             v = str(v or "").strip() or "（待补充）"
         lines.append("- **%s**：%s" % (label, v))
     lines.append("")
-    lines.append("> 各字段可直接复制进 %s 建书表单；简介请保持 50-500 字。" % plat)
+    lines.append("> 各字段可直接复制进 %s 建书表单；分类与标签均选自平台真实选项表。" % plat)
     return "\n".join(lines)
 
 

package/app/core/bookmeta_catalog.py

@@ -0,0 +1,187 @@
+# -*- coding: utf-8 -*-
+"""建书表单的真实选项目录：番茄官方 API 实抓 + 七猫官网书库页实抓。
+
+数据来源（2026-09-16 抓取）：
+- 番茄：https://fanqienovel.com/api/author/book/category_list/v0/?gender=1|0
+  （官方接口，无需登录；主分类 + 主题/角色/情节 三组标签，按目标读者分两套）
+- 七猫：官网书库 SSR 页 + 榜单 API 的 category1/category2 真实配对；标签池来自
+  作者专区标签系统快照（GitHub Lorn.NovelWriteSkills，2026-06，四组：风格/角色/
+  情节/背景）。级联表未覆盖的一级分类（出版类等非主赛道）二级宽松放行。
+
+改这些表不影响已保存任务（book_meta 生成时快照）。
+"""
+
+# 番茄·男频主分类（gender=1）
+FANQIE_MALE_CATEGORIES = [
+    "西方奇幻", "东方仙侠", "科幻末世", "男频衍生", "都市高武", "悬疑灵异",
+    "悬疑脑洞", "抗战谍战", "历史古代", "历史脑洞", "都市种田", "都市脑洞",
+    "都市日常", "玄幻脑洞", "战神赘婿", "动漫衍生", "游戏体育", "传统玄幻",
+    "都市修真",
+]
+
+# 番茄·男频标签-主题组
+FANQIE_MALE_TAGS_THEME = [
+    "衍生", "仕途", "综影视", "天灾", "第一人称", "赛博朋克",
+    "第四天灾", "规则怪谈", "搞笑轻松", "古代", "悬疑", "克苏鲁",
+    "都市异能", "末日求生", "灵气复苏", "高武世界", "异世大陆", "东方玄幻",
+    "谍战", "清朝", "宋朝", "断层", "武将", "国运",
+    "综漫", "开局", "架空", "奇幻仙侠", "都市", "玄幻",
+    "历史", "体育", "武侠",
+]
+
+# 番茄·男频标签-角色组
+FANQIE_MALE_TAGS_ROLE = [
+    "多女主", "赘婿", "全能", "大佬", "大小姐", "特工",
+    "游戏主播", "神探", "贵族", "皇帝", "单女主", "校花",
+    "无女主", "女帝", "特种兵", "反派", "神医", "奶爸",
+    "学霸", "天才", "腹黑", "扮猪吃虎",
+]
+
+# 番茄·男频标签-情节组
+FANQIE_MALE_TAGS_PLOT = [
+    "都市江湖", "风水秘术", "斩神衍生", "十日衍生", "西游衍生", "公版衍生",
+    "红楼衍生", "甄嬛衍生", "如懿衍生", "惊悚游戏", "卡牌", "山海经",
+    "捉鬼", "剑修", "废土", "副本", "黑科技", "无脑爽",
+    "魂穿", "高手下山", "黑化", "迪化", "发家致富", "无后宫",
+    "争霸", "1v1", "升级流", "灵魂互换", "科举", "封神",
+    "四合院", "电竞", "双重生", "乡村", "同人", "打脸",
+    "破案", "囤物资", "钓鱼", "网游", "奥特同人", "求生",
+    "无敌", "民国", "九叔", "穿书", "聊天群", "大秦",
+    "龙珠", "漫威", "神奇宝贝", "海贼", "火影", "职场",
+    "明朝", "家庭", "三国", "末世", "直播", "无限流",
+    "诸天万界", "大唐", "宠物", "外卖", "星际", "美食",
+    "剑道", "盗墓", "灵异", "鉴宝", "系统", "神豪",
+    "重生", "穿越", "二次元", "海岛", "娱乐圈", "空间",
+    "推理", "洪荒",
+]
+
+# 番茄·女频主分类（gender=0）
+FANQIE_FEMALE_CATEGORIES = [
+    "女频悬疑", "古风世情", "科幻末世", "女频衍生", "民国言情", "悬疑脑洞",
+    "青春甜宠", "双男主", "古言脑洞", "现言脑洞", "玄幻言情", "宫斗宅斗",
+    "豪门总裁", "动漫衍生", "星光璀璨", "游戏体育", "职场婚恋", "双女主",
+    "年代", "种田", "快穿",
+]
+
+# 番茄·女频标签-主题组
+FANQIE_FEMALE_TAGS_THEME = [
+    "古言权谋", "悬疑恋爱", "纯爱", "衍生", "仕途", "综影视",
+    "天灾", "第一人称", "赛博朋克", "规则怪谈", "搞笑轻松", "古代",
+    "悬疑", "谍战", "职场商战", "虐恋情深", "日久生情", "豪门世家",
+    "综漫", "异世穿越", "独宠", "现代言情", "古代言情", "武侠",
+    "幻想言情",
+]
+
+# 番茄·女频标签-角色组
+FANQIE_FEMALE_TAGS_ROLE = [
+    "位尊权重", "总裁", "忠犬", "全能", "白切黑", "双学霸",
+    "作精", "大佬", "大小姐", "游戏主播", "神探", "贵族",
+    "将军", "毒医", "厨娘", "律师", "医生", "明星",
+    "替身", "双面", "冰山", "古灵精怪", "天作之合", "可盐可甜",
+    "无CP", "病娇", "反派", "萌宝", "宠妻", "学霸",
+    "公主", "皇后", "王妃", "女强", "皇叔", "嫡女",
+    "精灵", "天才", "腹黑", "扮猪吃虎", "团宠",
+]
+
+# 番茄·女频标签-情节组
+FANQIE_FEMALE_TAGS_PLOT = [
+    "男二上位", "代嫁代娶", "攻略反派", "风水秘术", "斩神衍生", "十日衍生",
+    "西游衍生", "公版衍生", "红楼衍生", "甄嬛衍生", "如懿衍生", "惊悚游戏",
+    "追夫", "山海经", "胎穿", "捉鬼", "剑修", "相互救赎",
+    "宠夫", "无脑爽", "魂穿", "黑化", "养崽", "年龄差",
+    "真假千金", "久别重逢", "发家致富", "养成", "互宠", "1v1",
+    "灵魂互换", "科举", "年下", "婚恋", "封神", "四合院",
+    "电竞", "双重生", "前世今生", "双洁", "追妻火葬场", "乡村",
+    "逃荒", "同人", "打脸", "破案", "囤物资", "钓鱼",
+    "HE", "相爱相杀", "暗恋", "逃婚", "带球跑", "强强",
+    "一见钟情", "双向奔赴", "破镜重圆", "契约婚姻", "隐婚", "闪婚",
+    "今穿古", "古穿今", "群穿", "护短", "虐渣", "情有独钟",
+    "马甲", "先婚后爱", "医术", "女扮男装", "青梅竹马", "无敌",
+    "民国", "穿书", "职场", "家庭", "末世", "直播",
+    "无限流", "兽世", "清穿", "星际", "美食", "盗墓",
+    "虐文", "甜宠", "灵异", "校园", "系统", "重生",
+    "穿越", "二次元", "娱乐圈", "空间", "推理",
+]
+
+# 七猫·一级分类（官网书库筛选栏，23 个）
+QIMAO_MAIN_CATEGORIES = [
+    "现代言情", "古代言情", "幻想言情", "历史", "军事", "科幻",
+    "游戏", "游戏竞技", "玄幻奇幻", "都市", "奇闻异事", "武侠仙侠",
+    "体育", "N次元", "文学艺术", "人文社科", "经管励志", "经典文学",
+    "出版小说", "少儿教育", "衍生言情", "现实题材", "现实主义",
+]
+
+# 七猫·一级→二级级联（已确认部分；未列出的二级宽松放行）
+QIMAO_CASCADE = {
+    "现代言情": [
+        "总裁豪门", "职场情缘", "民国旧影", "娱乐明星", "现实生活", "青春校园",
+        "年代重生", "都市奇幻", "现代悬疑",
+    ],
+    "古代言情": [
+        "种田经商", "宫闱宅斗", "古代情缘", "权谋天下", "武侠江湖", "远古大陆",
+        "古代悬疑",
+    ],
+    "幻想言情": [
+        "玄幻仙侠", "未来科幻", "无限快穿", "末世求生", "异世幻想",
+    ],
+    "都市": [
+        "明星娱乐", "乡村生活", "都市高武", "都市高手", "商战职场", "热血校园",
+        "都市生活", "灵气复苏", "官场",
+    ],
+    "玄幻奇幻": [
+        "东方玄幻", "异世大陆", "西方奇幻", "诸天万界", "王朝争霸",
+    ],
+    "奇闻异事": [
+        "侦探推理", "恐怖灵异", "寻宝探险", "奇门秘术",
+    ],
+    "武侠仙侠": [
+        "上古洪荒", "古典仙侠", "传统武侠", "幻想修真", "武侠幻想",
+    ],
+    "军事": [
+        "现代军事", "抗战烽火", "谍战特工", "战争幻想",
+    ],
+    "科幻": [
+        "星际机甲", "末世危机", "未来幻想",
+    ],
+    "游戏": [
+        "电子竞技", "虚拟网游",
+    ],
+    "游戏竞技": [
+        "网游电竞", "运动竞技",
+    ],
+    "体育": [
+        "其他竞技", "篮球风云", "足球天下",
+    ],
+    "N次元": [
+        "原生幻想", "衍生同人", "轻小说",
+    ],
+    "历史": [
+        "架空历史",
+    ],
+}
+
+# 七猫·作品标签池（四组，作者专区快照常用子集；校验宽松：池外不拒）
+QIMAO_TAG_GROUPS = {
+    "风格": [
+        "热血", "轻松", "爽文", "甜宠", "虐恋", "悬疑",
+        "快节奏", "反套路", "第一人称", "种田", "权谋", "宫斗",
+        "宅斗", "美食", "经营", "克苏鲁", "规则怪谈", "无CP",
+    ],
+    "角色": [
+        "战神", "赘婿", "神医", "神豪", "奶爸", "女帝",
+        "嫡女", "真千金", "大女主", "兵王", "医生", "主播",
+        "总裁", "大佬", "学霸", "团宠", "萌宝", "反派",
+    ],
+    "情节": [
+        "系统", "金手指", "签到", "御兽", "重生", "穿越",
+        "穿书", "快穿", "逆袭", "开局流", "无敌流", "先婚后爱",
+        "追妻火葬场", "破镜重圆", "闪婚", "打脸虐渣", "马甲", "养成",
+    ],
+    "背景": [
+        "末世", "无限流", "玄学", "盗墓", "鉴宝", "娱乐圈",
+        "电竞", "官场", "架空历史", "三国", "大唐", "大明",
+        "大秦", "年代", "民国", "高武", "灵气复苏", "诸天万界",
+        "科幻", "赛博朋克", "校园", "都市", "乡村", "海岛",
+        "星际", "玄幻", "武侠", "仙侠",
+    ],
+}