dsh-novel-writer 3.2.1 → 3.6.0

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 siweina
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.en.md

@@ -8,7 +8,7 @@ English | [**中文**](./README.md)
 [![GitHub release](https://img.shields.io/github/v/release/siweina/dsh-novel-writer.svg?style=flat-square)](https://github.com/siweina/dsh-novel-writer/releases)
 [![DSH plugin](https://img.shields.io/badge/DSH-plugin-4b8bbe.svg?style=flat-square)](https://github.com/deepseek-ai/deepseek-harness)
 
-A novel-writing assistant plugin for **DeepSeek Harness (DSH)** (v3.2.0): chapter library management, sentence-pattern analysis, emotion purification & quantification, **12-axis vibe spectrum**, **style portrait report**, **six-dimension style baseline band**, **writing sentinels (bridge/OOC/outline drift)**, plot & settings management, local semantic search (0 token), webnovel signal detection, **original mode with creation files (bible/dynamic outline/hook backfill)**, batch import, and AI-assisted continuation writing. Zero third-party dependencies on host. **Requires Node ≥ 22.3.**
+A novel-writing assistant plugin for **DeepSeek Harness (DSH)**: chapter library management, sentence-pattern analysis, emotion purification & quantification, **12-axis vibe spectrum**, **style portrait report**, **six-dimension style baseline band**, **writing sentinels (bridge/OOC/outline drift)**, plot & settings management, local semantic search (0 token), webnovel signal detection, **original mode with creation files (bible/dynamic outline/hook backfill)**, batch import, and AI-assisted continuation writing. The semantic engine depends on `onnxruntime-web` and `@huggingface/tokenizers` (auto-installed with the package). **Requires Node ≥ 22.3.**
 
 > **A note to non-Chinese users**: This plugin is designed specifically for Chinese-language novel analysis and writing — its core capabilities (sentence-pattern analysis, emotion quantification, imagery detection) and its built-in semantic model are all built and tuned for Chinese text. Fully supporting English or other languages alongside Chinese is beyond my current capability. I sincerely apologize for any inconvenience this may cause.
 
@@ -39,14 +39,14 @@ After installing, **restart the web app** to activate (host registers 16 tools +
 
 1. **Style portrait report** (`novel_style_report`): 6-dimension measurement — style fingerprint / high-frequency lexicon / genre-theme / emotion quantification / 12-axis vibe spectrum / semantic style distance. **Measurement-judgment separation**: the plugin only reports numbers, never labels; AI judgment can be saved back to `.novel-writer/style-reports/` for consistent continuation writing.
 2. **12-axis vibe spectrum**: nightmare / angst / heartwarming / fluff / tearjerker / dark / mystery / blaze / absurd / lonesome / aesthetic / sensual — with traceable evidence, 0 token.
-3. **Local semantic engine**: bge-small-zh Chinese model (23MB, shipped with the plugin) local CPU inference — `novel_semantic_search` finds semantically related passages with natural language (with chapter location), semantic style comparison, semantic implicit emotion; lazy loading + graceful fallback.
+3. **Local semantic engine**: bge-small-zh Chinese model (24MB, shipped with the plugin) local CPU inference — `novel_semantic_search` finds semantically related passages with natural language (with chapter location), semantic style comparison, semantic implicit emotion; lazy loading + graceful fallback.
 4. **Sentence-pattern analysis**: 9 categories, arrangement patterns, rhythm, emotion curve, style fingerprint + guidance, with cache & report export.
 5. **Emotion purification & quantification**: strong/weak emotion-word grading, pollution detection, caveat warning + AI re-verification; Valence sliding window → variance V / delta Δ / conflict index C + implicit imagery carriers.
 6. **Worldview & pragmatics detection**: auto cultural-baseline detection with confidence; speechStyle title/honorifics/rituals/tone norms; genre & theme + webnovel signals.
 7. **Writing toolkit**: plot tracking / five settings tables (characters·locations·items·timeline·worldview) / chapter summaries / continuity audit / batch import / style check / continuation writing.
 8. **Per-tool UI toggles**: "Writing Assistant" sidebar panel (master + grouped tool toggles + feature toggles), plain-language labels, data-dir usage & semantic-engine status display.
 9. **Style Baseline**: Six writing metrics (syntactic complexity / modifier density / abstraction / action density / hedging / gap index) + per-chapter μ±σ baseline band; `novel_style_report` outputs the band, `novel_style_check` compares new chapters (in-band ✓ / out-of-band ⚠); per-metric ±% tolerance configurable in the sidebar (**recommended = 1.5× σ of the book's chapter variance**, rounded, clamped to ±10%~100%; leave blank to use recommended) — free theme, writing style kept inside the band.
-10. **Writing sentinels (v3.2.0)**: `novel_continuity_check` extended — ①**bridge check** (`chapter`: time jumps / semantic distance / character continuity / hook handoff, with quoted evidence) ②**OOC check** (`ooc`: per-character emotion baseline deviation) ③**outline drift** (`outline`: direction vs body keyword overlap); **brief mode** for report tools.
+10. **Writing sentinels**: `novel_continuity_check` extended — ①**bridge check** (`chapter`: time jumps / semantic distance / character continuity / hook handoff, with quoted evidence) ②**OOC check** (`ooc`: per-character emotion baseline deviation) ③**outline drift** (`outline`: direction vs body keyword overlap); **brief mode** for report tools.
 11. **Original mode & creation files**: fill in creation settings in the sidebar (worldview/characters/forbidden/main conflict/genre/extras, blank = model decides, per-book profile library); novel_outline maintains creation files (bible/characters/outline/hooks/status), enforcing the bible → outline → hook chain with dynamic batches (10→20→30 chapters) to prevent plot jumps and OOC.
 12. **Experience & stats**: main panel **library stats** (per-book chapters/total chars/7-day active chars, 🔥 green), **🎬 demo** (built-in sample, no files, runs the 6-dim baseline), **📊 report history** (analysis/style-reports browsing); actionable error hints; slimmer tool descriptions.
 

package/README.md

@@ -8,7 +8,7 @@
 [![GitHub release](https://img.shields.io/github/v/release/siweina/dsh-novel-writer.svg?style=flat-square)](https://github.com/siweina/dsh-novel-writer/releases)
 [![DSH plugin](https://img.shields.io/badge/DSH-plugin-4b8bbe.svg?style=flat-square)](https://github.com/deepseek-ai/deepseek-harness)
 
-为 **DeepSeek Harness (DSH)** 打造的小说写作助手插件（v3.2.0）：章节库管理、句式分析、情感净化与量化、氛围光谱、**风格画像报告**、**文笔六维基线带**、**写作哨兵三件套（衔接/OOC/大纲走偏）**、伏笔设定管理、本地语义检索（0 token）、网文信号识别、**原创模式与创作资料管理（设定书/动态大纲/钩子回填）**、批量导入与 AI 续写辅助。宿主端零第三方依赖，浏览器端仅依赖 Web GUI 自带的 react。**要求 Node ≥ 22.3。**
+为 **DeepSeek Harness (DSH)** 打造的小说写作助手插件：章节库管理、句式分析、情感净化与量化、氛围光谱、**风格画像报告**、**文笔六维基线带**、**写作哨兵三件套（衔接/OOC/大纲走偏）**、伏笔设定管理、本地语义检索（0 token）、网文信号识别、**原创模式与创作资料管理（设定书/动态大纲/钩子回填）**、批量导入与 AI 续写辅助。语义引擎依赖 `onnxruntime-web` 与 `@huggingface/tokenizers`（随包自动安装），浏览器端仅依赖 Web GUI 自带的 react。**要求 Node ≥ 22.3。**
 
 > **致非中文用户**：本插件为中文小说分析写作而设计——句式、情感、意象等核心能力以及内置的语义模型，全部针对中文语料构建与调优。在深耕中文的同时兼顾英文等其他语言，确实超出了我目前的能力范围。若因此给您带来不便，我深感抱歉，恳请谅解。
 
@@ -39,14 +39,14 @@ dsh plugin --profile web add github:siweina/dsh-novel-writer#main
 
 1. **风格画像报告**（novel_style_report）：6 维测量报告——文风指纹 / 高频词汇 / 题材流派 / 情感量化 / 氛围光谱 12 轴 / 语义风格距离。**测量与判断分离**：插件只报数不贴标签，AI 判断可回传存盘（`.novel-writer/style-reports/`），续写保持风格一致。
 2. **氛围光谱 12 轴**：噩梦感 / 焦虑压抑 / 温馨治愈 / 甜宠日常 / 催泪虐心 / 黑暗残酷 / 悬疑神秘 / 热血激昂 / 荒诞无厘头 / 孤独疏离 / 文艺唯美 / 情欲暧昧——证据链可追溯，0 token。
-3. **本地语义引擎**：bge-small-zh 中文模型（23MB 随插件分发）本地 CPU 推理——`novel_semantic_search` 自然语言搜全书语义相关段落（带章节定位），语义级风格对比、语义隐性情感，懒加载 + 自动回退。
+3. **本地语义引擎**：bge-small-zh 中文模型（24MB 随插件分发）本地 CPU 推理——`novel_semantic_search` 自然语言搜全书语义相关段落（带章节定位），语义级风格对比、语义隐性情感，懒加载 + 自动回退。
 4. **句式模式分析**：九类句式分布、排列规律、句长节奏、情感曲线、风格指纹与节奏建议，带缓存与报告导出。
 5. **情感净化 + 量化**：强/弱情绪词分级、污染源检测、caveat 预警 + AI 复核；Valence 滑动窗口 → 方差 V / 斜率 Δ / 矛盾指数 C + 隐性意象载体。
 6. **世界观与语用检测**：文化基准自动判断（西/东/混合）+ 置信度；speechStyle 称谓/客套/仪式/语气规范；题材流派 + 网文信号。
 7. **写作辅助全家桶**：伏笔登记表 / 设定五张表（人物·地点·道具·时间线·世界观）/ 章节摘要 / 连贯性审计 / 批量导入 / 风格自检 / 续写辅助。
 8. **全工具 UI 开关**：侧边栏「写作助手功能」面板（总开关 + 工具开关分组 + 功能开关），大白话文案，显示数据目录占用与语义引擎状态。
 9. **风格基线**：文笔六维测量（句法复杂度/修饰密度/抽象度/动作密度/不确定性/留白指数）+ 按章节 μ±σ 基线带；`novel_style_report` 输出基线带，`novel_style_check` 对照新章偏差（带内 ✓ / 出带 ⚠）；侧边栏可自定义每维 ±% 容差（**推荐值 = 原著章节波动的 1.5 倍 σ**，自动取整、限 ±10%~100%；输入框留空即用推荐）——原创/续写时主题自由、写法保持在基线带内。
-10. **写作哨兵三件套（v3.2.0）**：`novel_continuity_check` 扩展——①**衔接检查**（chapter 参数：时间硬跳/语义距离/人物延续/钩子承接四路检测，带原文引用）②**OOC 检测**（ooc 参数：角色情绪基线偏离）③**大纲走偏**（outline 参数：方向行 vs 正文关键词重合）；报告工具支持 **brief 精简模式**。
+10. **写作哨兵三件套**：`novel_continuity_check` 扩展——①**衔接检查**（chapter 参数：时间硬跳/语义距离/人物延续/钩子承接四路检测，带原文引用）②**OOC 检测**（ooc 参数：角色情绪基线偏离）③**大纲走偏**（outline 参数：方向行 vs 正文关键词重合）；报告工具支持 **brief 精简模式**。
 11. **原创模式与创作资料**：侧边栏填写创作设定（世界观/角色/禁忌/主线/题材/额外要求，留空=模型自定，多书独立设定库）；novel_outline 维护创作资料（创作设定/人物/剧情大纲/钩子记录/创作状态卡），原创强制「设定书→大纲→钩子」链，动态批次（10→20→30 章）防剧情跳跃与角色 OOC。
 12. **体验与统计**：主面板**书库统计卡**（每本章数/总字数/近 7 天活跃字数，活跃🔥标绿）、**🎬 体验演示**（内置示例不落盘跑六维基线）、**📊 报告历史**（analysis/style-reports 列表浏览）；错误提示带解决步骤；工具说明压缩省 token。
 

package/cordis.patch.yml

@@ -1,4 +1,4 @@
-# dsh-novel-writer v2.0.0 bundle patch: 14 tools (incl. novel_semantic_search) + local embedding engine.
+# dsh-novel-writer v3.5.0 bundle patch: 16 tools (incl. novel_semantic_search) + local embedding engine.
 # dsh-novel-writer bundle patch: mounts the novel-writing assistant plugin
 # into the profile's loader tree. The entry's name is the plugin package
 # itself, resolved from the profile's node_modules.

package/lib/analysis.js

@@ -55,18 +55,19 @@ const STRONG_EMOTION_WORDS = Object.freeze({
   surprise: ["震惊", "目瞪口呆", "难以置信", "不可思议", "惊愕", "骇然", "震撼"]
 });
 const WEAK_EMOTION_WORDS = Object.freeze({
-  joy: ["兴奋", "满足", "愉快", "痛快", "爽快", "爽", "笑眯眯", "笑容", "微笑", "笑", "哈哈", "乐", "欢", "喜", "高兴", "开心", "喜悦"],
-  anger: ["生气", "恼火", "恼", "怒", "恨", "气愤", "气", "咬牙", "不满", "发火", "气冲冲"],
-  sorrow: ["难过", "伤心", "痛苦", "哭泣", "哭", "泪", "眼泪", "流泪", "哽咽", "抽泣", "叹息", "叹气", "失落", "悲"],
+  joy: ["兴奋", "满足", "愉快", "痛快", "爽快", "笑眯眯", "笑容", "微笑", "哈哈", "高兴", "开心", "喜悦"],
+  anger: ["生气", "恼火", "气愤", "咬牙", "不满", "发火", "气冲冲"],
+  sorrow: ["难过", "伤心", "痛苦", "哭泣", "眼泪", "流泪", "哽咽", "抽泣", "叹息", "叹气", "失落", ],
   fear: ["害怕", "惊慌", "不安", "紧张", "担心", "发抖", "哆嗦", "心慌", "忐忑", "心悸", "心虚", "惊惶"],
   surprise: ["惊讶", "意外", "吃惊", "诧异", "愕然", "愣住", "傻眼", "呆住", "惊奇", "惊呆"]
 });
+// v3.5.0 M1：情感词表构建去重（欣慰/惊惶等同表重复——双计残留）
 const EMOTION_WORDS = Object.freeze({
-  joy: [...STRONG_EMOTION_WORDS.joy, ...WEAK_EMOTION_WORDS.joy],
-  anger: [...STRONG_EMOTION_WORDS.anger, ...WEAK_EMOTION_WORDS.anger],
-  sorrow: [...STRONG_EMOTION_WORDS.sorrow, ...WEAK_EMOTION_WORDS.sorrow],
-  fear: [...STRONG_EMOTION_WORDS.fear, ...WEAK_EMOTION_WORDS.fear],
-  surprise: [...STRONG_EMOTION_WORDS.surprise, ...WEAK_EMOTION_WORDS.surprise]
+  joy: [...new Set([...STRONG_EMOTION_WORDS.joy, ...WEAK_EMOTION_WORDS.joy])],
+  anger: [...new Set([...STRONG_EMOTION_WORDS.anger, ...WEAK_EMOTION_WORDS.anger])],
+  sorrow: [...new Set([...STRONG_EMOTION_WORDS.sorrow, ...WEAK_EMOTION_WORDS.sorrow])],
+  fear: [...new Set([...STRONG_EMOTION_WORDS.fear, ...WEAK_EMOTION_WORDS.fear])],
+  surprise: [...new Set([...STRONG_EMOTION_WORDS.surprise, ...WEAK_EMOTION_WORDS.surprise])]
 });
 
 /** v1.5.0 情绪污染源词表：检测到高密度时降低情感可信度并提示 AI 复核。 */
@@ -131,6 +132,8 @@ export function splitBlocks(text) {
   let current = [];
   for (const line of lines) {
     const trimmed = line.trim();
+    // v3.5.0 #51：Markdown 标题（# 开头）不是句子——跳过（含空行场景）
+    if (/^#{1,6}\s/.test(trimmed)) continue;
     if (trimmed === "") {
       if (current.length > 0) {
         blocks.push(current.join("\n").trim());
@@ -154,6 +157,13 @@ export function splitSentences(block) {
   let buffer = "";
   for (let i = 0; i < block.length; i += 1) {
     const ch = block[i];
+    // v3.5.0 #50b：无标点换行也是句边界（逐行排版正确分句，整篇不再算 1 句）
+    if (ch === "\n" || ch === "\r") {
+      const lineSentence = buffer.trim();
+      if (lineSentence !== "") sentences.push(lineSentence);
+      buffer = "";
+      continue;
+    }
     buffer += ch;
     if (TERMINATORS.includes(ch)) {
       while (i + 1 < block.length && TERMINATORS.includes(block[i + 1])) {
@@ -267,7 +277,7 @@ export function classifySentence(raw) {
 /** 单字情感词的多字搭配排除（"叹气"不算"气"、"笑道"不算"笑"）。 */
 const EMOTION_EXCLUDE = {
   joy: { "笑": ["笑道", "说笑", "玩笑", "搞笑", "笑话", "苦笑", "冷笑", "嘲笑", "嬉笑", "傻笑", "微笑", "笑死", "笑呵呵", "笑眯眯"] },
-  anger: { "气": ["叹气", "运气", "客气", "一口气", "服气", "脾气", "争气", "节气", "淘气", "景气", "暖气", "口气", "名气", "勇气", "底气", "俗气", "风气", "香气", "水气", "湿气", "潮气", "和气", "生气勃勃"] },
+  anger: { "气": ["叹气", "运气", "客气", "一口气", "服气", "脾气", "争气", "节气", "淘气", "景气", "暖气", "口气", "名气", "勇气", "底气", "俗气", "风气", "香气", "水气", "湿气", "潮气", "和气", "生气勃勃"], "生气": ["生气勃勃"] },
   sorrow: { "悲": ["慈悲"], "哭": ["哭丧"] },
   fear: { "怕": ["哪怕", "只怕", "生怕"] },
   surprise: { "愣": ["愣住"] }
@@ -296,10 +306,12 @@ function emotionOf(text) {
         // 单字词的搭配排除（如"叹气"中的"气"）
         const exclude = EMOTION_EXCLUDE[emotion]?.[word];
         if (exclude !== void 0) {
+          // v3.5.0 #54：窗口按最长排除词取（"生气勃勃"=4 字，旧窗口永远命不中）
+          const maxLen = Math.max(4, ...exclude.map(function (x) { return x.length; }));
           const two = text.slice(Math.max(0, idx - 1), idx + 2);
           const prevTwo = text.slice(Math.max(0, idx - 1), idx + 1);
           const nextTwo = text.slice(idx, idx + 2);
-          const prevFour = text.slice(Math.max(0, idx - 2), idx + 2);
+          const prevFour = text.slice(Math.max(0, idx - maxLen), idx + maxLen);
           let excluded = false;
           for (const bad of exclude) {
             if (two.includes(bad) || prevTwo === bad || nextTwo === bad || prevFour.includes(bad)) {
@@ -433,22 +445,20 @@ export function buildGuidance(ratios, topPatterns, chapterSequences) {
  */
 const VALENCE_WORDS = Object.freeze({
   "欣喜": 0.9, "狂喜": 0.9, "欢天喜地": 0.9, "雀跃": 0.7, "高兴": 0.7, "开心": 0.7, "快乐": 0.7,
-  "喜悦": 0.7, "愉快": 0.5, "欢喜": 0.7, "兴奋": 0.7, "愉悦": 0.5, "欢快": 0.7, "乐": 0.5,
-  "欢": 0.5, "喜": 0.5, "笑": 0.3, "微笑": 0.3, "笑容": 0.3, "笑眯眯": 0.3, "哈哈": 0.3,
-  "欣慰": 0.5, "满足": 0.3, "痛快": 0.5, "爽快": 0.3, "甜": 0.5, "甜蜜": 0.5, "幸福": 0.9,
+  "喜悦": 0.7, "愉快": 0.5, "欢喜": 0.7, "兴奋": 0.7, "愉悦": 0.5, "欢快": 0.7, "微笑": 0.3, "笑容": 0.3, "笑眯眯": 0.3, "哈哈": 0.3,
+  "欣慰": 0.5, "满足": 0.3, "痛快": 0.5, "爽快": 0.3, "甜蜜": 0.5, "幸福": 0.9,
   "美满": 0.9, "温馨": 0.5, "温暖": 0.3, "踏实": 0.1, "平静": 0.1, "释然": 0.4, "解脱": 0.4,
-  "喜欢": 0.7, "喜爱": 0.7, "欣赏": 0.5, "爱": 0.9, "疼爱": 0.7, "宠爱": 0.7, "仰慕": 0.7,
+  "喜欢": 0.7, "喜爱": 0.7, "欣赏": 0.5, "疼爱": 0.7, "宠爱": 0.7, "仰慕": 0.7,
   "尊敬": 0.5, "敬仰": 0.7, "崇拜": 0.7, "心动": 0.5, "眷恋": 0.7, "依恋": 0.7, "思念": 0.3,
   "心疼": 0.3, "怜爱": 0.5, "温柔": 0.3, "珍惜": 0.5, "信赖": 0.5, "感恩": 0.7,
   "愤怒": -0.9, "暴怒": -0.9, "怒发冲冠": -0.9, "火冒三丈": -0.9, "恼羞成怒": -0.9,
-  "生气": -0.7, "恼火": -0.7, "气愤": -0.7, "怒": -0.7, "恨": -0.9, "怨恨": -0.9, "憎恨": -0.9,
-  "厌恶": -0.7, "憎恶": -0.9, "不满": -0.5, "恼": -0.5, "气": -0.3, "发火": -0.7, "咬牙": -0.3,
+  "生气": -0.7, "恼火": -0.7, "气愤": -0.7, "怨恨": -0.9, "憎恨": -0.9,
+  "厌恶": -0.7, "憎恶": -0.9, "不满": -0.5, "发火": -0.7, "咬牙": -0.3,
   "怒意": -0.7, "怒火": -0.9, "愤恨": -0.9, "恼羞": -0.7, "气冲冲": -0.7, "咬牙切齿": -0.5,
   "悲伤": -0.7, "悲痛": -0.9, "悲痛欲绝": -0.9, "悲哀": -0.7, "哀伤": -0.7, "难过": -0.5,
-  "伤心": -0.7, "痛苦": -0.7, "心碎": -0.9, "绝望": -0.9, "哭泣": -0.7, "哭": -0.5, "泪": -0.3,
-  "眼泪": -0.3, "流泪": -0.5, "哽咽": -0.5, "抽泣": -0.5, "叹息": -0.3, "叹气": -0.3,
+  "伤心": -0.7, "痛苦": -0.7, "心碎": -0.9, "绝望": -0.9, "哭泣": -0.7, "眼泪": -0.3, "流泪": -0.5, "哽咽": -0.5, "抽泣": -0.5, "叹息": -0.3, "叹气": -0.3,
   "惆怅": -0.5, "失落": -0.5, "忧伤": -0.5, "黯然": -0.5, "心酸": -0.7, "辛酸": -0.7,
-  "悲": -0.7, "凄凉": -0.7, "苦涩": -0.7, "苦闷": -0.5, "沮丧": -0.7, "消沉": -0.7,
+  "凄凉": -0.7, "苦涩": -0.7, "苦闷": -0.5, "沮丧": -0.7, "消沉": -0.7,
   "落寞": -0.5, "孤寂": -0.5, "郁闷": -0.3, "低落": -0.3, "压抑": -0.5, "心灰意冷": -0.9,
   "恐惧": -0.9, "害怕": -0.7, "惊慌": -0.7, "不安": -0.3, "紧张": -0.3, "担心": -0.3,
   "畏惧": -0.7, "惊恐": -0.9, "胆怯": -0.5, "发抖": -0.3, "哆嗦": -0.3, "心慌": -0.5,
@@ -592,9 +602,11 @@ export function implicitEmotionScan(text, semResolver = null) {
     else { posHits += n * valence; if (false) fragileHits += n; }
     carrierCounts.set(label + ":" + word, (carrierCounts.get(label + ":" + word) ?? 0) + n);
   };
-  // ① 单方向表（原有）
+  // ① 单方向表（原有）——v3.5.0 M2：跳过与多方向表重叠的词（雨/黄昏/烛火等），防同一出现计 2-3 次（语境裁决交给 ②）
+  const ambWords = new Set(Object.keys(AMBIGUOUS_CARRIERS));
   for (const carrier of IMPLICIT_CARRIERS) {
     for (const word of carrier.words) {
+      if (ambWords.has(word)) continue;
       const n = text.split(word).length - 1;
       if (n > 0) {
         see(carrier.label, word, n, carrier.valence);
@@ -682,11 +694,13 @@ export function valenceSeries(text, winChars = 100) {
 
 /** v1.6.0 三指标：方差 V + 相邻撕裂 V_adj + 斜率 Δ（最小二乘+鲁棒版）+ 矛盾指数 C。 */
 export function valenceStats(text) {
-  const { series, posWords, negWords } = valenceSeries(text);
+  // v3.5.0 #57：一次 valenceSeries 取全部（旧版窗口统计二次调用，大书白扫一遍词表）
+  const { series, posWords, negWords, windowPosNeg } = valenceSeries(text);
   const n = series.length;
   if (n === 0) return { windows: 0 };
   const mean = series.reduce((x, y) => x + y, 0) / n;
-  const variance = series.reduce((s, x) => s + (x - mean) ** 2, 0) / n;
+  // v3.6.0：样本方差口径 /(n-1)（n=1 时无方差=0）
+  const variance = n > 1 ? series.reduce((s, x) => s + (x - mean) ** 2, 0) / (n - 1) : 0;
   let adjSum = 0;
   for (let i = 1; i < n; i += 1) adjSum += Math.abs(series[i] - series[i - 1]);
   const adjVariance = n > 1 ? adjSum / (n - 1) : 0;
@@ -701,7 +715,7 @@ export function valenceStats(text) {
   const third = Math.max(1, Math.floor(n / 3));
   const headMean = series.slice(0, third).reduce((x, y) => x + y, 0) / third;
   const tailMean = series.slice(-third).reduce((x, y) => x + y, 0) / third;
-  const { windowPosNeg } = valenceSeries(text);
+  // v3.5.0 #57：复用上面的 windowPosNeg（不再二次扫描）
   const totalWords = posWords + negWords;
   const posRatio = totalWords === 0 ? 0 : posWords / totalWords;
   const negRatio = totalWords === 0 ? 0 : negWords / totalWords;
@@ -855,8 +869,11 @@ export function emotionalQuantification(text, perChapter, blocks, semResolver =
  * @returns 结构化分析结果（与 novel_sentence_analysis 输出 schema 一致）。
  */
 export function analyzeText(text, options = {}) {
+  // v3.5.0 M6：入口容错（null/undefined 不崩溃）
+  text = String(text ?? "");
   const top = Number.isInteger(options.top) ? options.top : 8;
-  const maxSentences = Number.isInteger(options.maxSentences) ? options.maxSentences : 20000;
+  // v3.5.0 #63：maxSentences 钳制 ≥1（0/负数会让 slice(0,-5) 产生错误语义）
+  const maxSentences = Math.max(1, Number.isInteger(options.maxSentences) ? options.maxSentences : 20000);
   const curveSegments = Number.isInteger(options.curveSegments) ? Math.min(Math.max(options.curveSegments, 1), 50) : 20;
   const blocks = splitBlocks(text);
   const sentences = [];
@@ -1097,7 +1114,21 @@ export function analyzeText(text, options = {}) {
   // 主观性指数（启发式 0-100）
   const psychRatio = totalSentences === 0 ? 0 : counts.psychology / totalSentences;
   const exclaimRatio = totalSentences === 0 ? 0 : counts.exclamation / totalSentences;
-  const firstPersonCount = FIRST_PERSON_WORDS.reduce((sum, word) => sum + (text.split(word).length - 1), 0);
+  // v3.6.0：第一人称最长匹配优先（"我们"不拆成 我+我们 双计）
+  const firstPersonCount = (function () {
+    let fpSum = 0;
+    const fpSorted = FIRST_PERSON_WORDS.slice().sort((x, y) => y.length - x.length);
+    let fpText = text;
+    for (const w of fpSorted) {
+      if (w.length < 2) continue;
+      const parts = fpText.split(w);
+      fpSum += parts.length - 1;
+      fpText = parts.join("\u0000".repeat(w.length));
+    }
+    // 剩余孤立单字"我"
+    fpSum += (fpText.match(/我/g) || []).length;
+    return fpSum;
+  })();
   const firstPersonDensity = totalChars === 0 ? 0 : (firstPersonCount / totalChars) * 1000;
   const subjectivityIndex = Math.min(100, Math.round(
     psychRatio * 50 + exclaimRatio * 60 + Math.min(emotionDensity * 2.5, 25) + Math.min(firstPersonDensity * 1.2, 20)
@@ -1191,6 +1222,8 @@ function emotionCurve(blockMeta, maxSegments) {
       intensity: chars === 0 ? 0 : round((total / chars) * 1000, 2)
     });
   }
+  // v3.6.0 复审修正：不裁剪——强度 0 段是"平缓段"（跨章节曲线对齐需要固定段数契约）；
+  // segmentCount=min(blocks, maxSegments) 已保证每段至少 1 块，无空 slice
   return curve;
 }
 
@@ -1269,7 +1302,8 @@ const DENSITY_SENSE = {
 /** 细节密度统计（v0.8.0）：动作链/物件名词/感官词，按千字归一。 */
 export function densityOf(text) {
   const sentences = splitSentences(text);
-  const totalChars = text.replace(/\s/g, "").length;
+  // v3.5.0 #62：分母口径统一（含空白，与 style-metrics 的 per1000 一致，跨模块数值可比）
+  const totalChars = String(text).length;
   let actionVerbs = 0;
   let actionChainSentences = 0;
   let objectHits = 0;
@@ -1316,7 +1350,7 @@ function loadDutir() {
   }
   return DUTIR;
 }
-const DUTIR_TO_EMOTION = { 乐: "joy", 好: "joy", 怒: "anger", 哀: "sorrow", 惧: "fear", 恶: "sorrow", 惊: "surprise" };
+const DUTIR_TO_EMOTION = { 乐: "joy", 好: "joy", 怒: "anger", 哀: "sorrow", 惧: "fear", 恶: "anger", 惊: "surprise" }; // v3.5.0 H4：恶(厌恶) 归 anger 而非 sorrow
 const dutirLookup = new Map();
 function dutirEmotion(word) {
   if (dutirLookup.size === 0) {
@@ -1324,7 +1358,8 @@ function dutirEmotion(word) {
     for (const [cat, words] of Object.entries(dutir)) {
       const emo = DUTIR_TO_EMOTION[cat];
       if (!emo) continue;
-      for (const w of words) if (typeof w === "string" && w.length >= 2) dutirLookup.set(w, emo);
+      // v3.5.0 H4：跨类冲突词首见保留（后写不再覆盖——"开心"不会被 恶 类覆盖成 sorrow）
+      for (const w of words) if (typeof w === "string" && w.length >= 2 && !dutirLookup.has(w)) dutirLookup.set(w, emo);
     }
   }
   return dutirLookup.get(word);