dsh-novel-writer 6.5.0 → 6.5.5

package/README.en.md

@@ -39,6 +39,7 @@ Full version history: see [CHANGELOG.md](./CHANGELOG.md).
 | Your pain | What you get here |
 |---|---|
 | "This passage I just wrote doesn't sound like me" | **Six-metric style baseline**: μ±σ of the original measured per chapter across syntactic complexity / modifier density / abstraction / action density / hedging / gap index; a new chapter is compared dimension by dimension and flagged ⚠ when out of band |
+| "The AI keeps writing, and a few chapters in it sounds like the AI, not like my novel" | **Baseline scope (whole book / selected chapters)**: the baseline no longer has to be sampled from the whole book — tick the chapters you treat as the yardstick (say, the first 30 you wrote yourself) and the style check, chapter brief, fix plan, style report and new-chapter baseline will use **only** those chapters as the baseline; the chapter being checked is always excluded. **The plugin never decides "you have drifted" — it tells you "how far this chapter sits from the yardstick you chose"** (answers issue #1, "style drift") |
 | "The AI says my style changed but can't say where" | **Style check**: similarity score + a deviation list (which sentence types increased, how far sentence length drifted, whether the dominant emotion changed) |
 | "I have to poke six or seven tools before writing, and I still miss one" | **Chapter brief**: 16 inputs in one call; whatever can't be found is explained in `degraded` — **missing an input is the number one source of style drift and setting contradictions** |
 | "I can see the chapter has problems but not which one to fix first" | **Fix plan**: to-dos sorted by severity → effort → line, each with line-anchored text, current vs. target values, a reference passage and a rewrite direction; re-checkable in three states after editing |
@@ -83,6 +84,8 @@ Then ask, in order (this is one complete chapter):
 3. **"check chapter 7"** → `novel_style_check` returns similarity, a deviation list and the six-metric comparison
 4. **"turn chapter 7 into a fix list"** → `novel_fix_plan` returns the prioritised to-dos; run `verify` afterwards to re-check the three states
 
+> Want the baseline to follow **the chapters you wrote yourself** instead of the whole book? In the panel's **Style baseline** view, set **Baseline scope** to **selected chapters** and tick them: the check, the chapter brief, the fix plan and the style report all switch to those chapters as the baseline (**per book**; the default **whole book** is the old behaviour — leave it alone and nothing changes). Creating a chapter (`novel_new_chapter`) uses the same scope for the μ summary and the anchor pack it attaches.
+
 To just look at the baseline first, ask **"run novel_style_report on my novel"** and you get:
 
 ```text
@@ -104,7 +107,7 @@ To just look at the baseline first, ask **"run novel_style_report on my novel"**
   - [mu5kjla…] 琥珀色齿轮怀表的来历与停摆的指针（high｜第 1 章埋下，已过 6 章）
 【相关人物】- 林昭：守码头的女人，父亲失踪后回到旧宅
 【世界观用语】欧式中世纪沿海城邦；点烛不烧香｜禁用：上香、烧香、时辰、老夫
-【风格基线】complexity μ=2.42  modifierDensity μ=22.2  abstractDensity μ=10.06 …
+【风格基线】complexity μ=2.42  modifierDensity μ=22.2  abstractDensity μ=10.06 …（取样范围：全书除本章外的 7 章）
 【原著锚段·照这个味道写】[对话] …  [心理] …
 【上一章自检】第 7 章六维对照：全部维度在容差带内 ✓
 【本章禁用清单】- 禁词：上香（建议改用：点烛）
@@ -173,7 +176,7 @@ fixAnchors：3 条原著锚段（对话 / 心理 / 描写各一条，供逐句
 7. **Worldview & pragmatics detection**: auto cultural-baseline detection with confidence; speechStyle title/honorifics/rituals/tone norms; genre & theme + webnovel signals.
 8. **Writing toolkit**: plot tracking / five settings tables (characters·locations·items·timeline·worldview) / chapter summaries / continuity audit / batch import / style check / continuation writing.
 9. **Per-tool UI toggles**: a 「写作助手功能」 ("Writing Assistant") entry in the left sidebar plus a main-column panel (**official DSH 0.2.0 seats**; hosts older than 0.2.0 fall back automatically) — master switch + grouped tool toggles + feature toggles + prompt mode/scene/**lean workflow**, plain-language labels, data-dir usage & semantic-engine status. All colors go through `--dsw-alias-*` semantic tokens (three themes in sync with the host), controls use the official UI primitives, DeepSeek-blue accents, a glow hero header with section kickers, and equal-width pages across every level.<br>**The UI goes further (Plan A)**: source split + build chain (assert-failed builds never write, `--check` verifies source==artifact), 6 unified components with a single row pattern (`PreferenceRow`), group cards across every switch page, 19+5 line icons, a slider as auxiliary input on the tolerance rows, and page-level push-pop transition animations (zero state / zero hook).
-10. **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 panel's **Six-metric baseline** view (**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. **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 panel's **Six-metric baseline** view (**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.<br>**Baseline scope (v6.5.5, configured per book)**: the baseline does not have to be the whole book — in the panel's **Style baseline** view choose **whole book** (default, = the old behaviour) or **selected chapters**, then tick the chapters you treat as the yardstick; the baseline then uses **only** those (it means "only these", not "prefer these and top up when short"). This exists for "the AI keeps writing and ends up sounding like the AI": from chapter 2 onward the model is really reading its own previous chapter for reference, and **a baseline sampled from the whole book drifts along with it** (the longer the book, the higher the AI share). Nail the yardstick to the chapters you wrote yourself and the plugin can answer "**relative to those chapters**, how far has this chapter moved" — it never judges whether you have drifted, which is not decidable in the first place. The chapter being checked is **always excluded** (otherwise it would be comparing itself with itself); when fewer than 3 chapters are ticked the recommended tolerance widens to **at least ±25%**, so even a single chapter can serve as the baseline; chapters that cannot be used (renamed / deleted, or **too short to measure**) are skipped and reported honestly, and if not one of them can be used it falls back to **whole book** and states the reason in the returned baseline description. The scope is stored per book (`<library root>/.novel-writer/baseline/<book>.json`), so books do not affect each other.
 11. **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.
 12. **Original mode & creation files**: fill in creation settings in the panel's **Original mode** view (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.
 13. **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

@@ -38,6 +38,7 @@
 | 你的困扰 | 这里给的答案 |
 |---|---|
 | "我续写的这段，读起来不像我自己写的" | **文笔六维基线**：从句法复杂度 / 修饰密度 / 抽象度 / 动作密度 / 不确定性 / 留白指数六个维度算出原著的 μ±σ，新章逐维对照，出带标 ⚠ |
+| "AI 接着往下写，几章之后文风越来越像 AI 自己，不像原著了" | **基线范围（按全书 / 按部分）**：基线不必再从全书取样——勾出你认作标杆的那几章（比如你亲手写的前 30 章），自检 / 开写包 / 改稿台 / 风格画像 / 建章都**只用**这几章当基准，正在检查的那一章永远排除在外。**插件不判定「你漂移了」，只告诉你「相对你选的标杆，这章偏了多少」**（回应 issue #1「风格漂移」） |
 | "AI 说文风变了，但说不清哪儿变了" | **风格自检**：相似度 + 偏差清单（哪类句式多了、句长偏了多少、主导情绪有没有换） |
 | "动笔前要翻六七个工具，还老漏读一样" | **开写包**：一次调用取齐 16 项材料；取不到的在 `degraded` 里说明缺什么、怎么补——**漏读正是文风漂移与设定矛盾的头号来源** |
 | "知道这章有问题，但不知道先改哪句" | **改稿台**：按「严重度 → 难度 → 行号」排好待办，每条给原句行号 / 当前值与目标值 / 原著锚段 / 改写方向；改完可三态复测 |
@@ -80,6 +81,8 @@ mkdir -p novels/我的小说     # 把章节文件放进去（第01章.md、第0
 3. **「自检第 7 章」** → `novel_style_check` 给出相似度、偏差清单与六维对照
 4. **「把第 7 章排成改稿清单」** → `novel_fix_plan` 给出按优先级排好的待办；改完再 `verify` 复测三态
 
+> 想让基线对着**你亲手写的那些章**、而不是全书？面板「风格基线」→ **基线范围** 选 **按部分** 并勾章即可：上面这条链上的自检、开写包、改稿台与风格画像都会改用这几章当基线（**按书生效**；默认「按全书」＝旧行为，不动它就跟以前完全一样）。建章（`novel_new_chapter`）时附的 μ 摘要与锚段用的是同一份范围。
+
 只想先看看家底，就说 **「用 novel_style_report 给我的小说做一次风格画像」**：
 
 ```text
@@ -100,7 +103,7 @@ mkdir -p novels/我的小说     # 把章节文件放进去（第01章.md、第0
   - [mu5kjlb…] 海图上红铅笔圈的礁区坐标（high｜第 3 章埋下，已过 4 章）
 【相关人物】- 林昭：守码头的女人，父亲失踪后回到旧宅　- 沈砚：随船出海的人
 【世界观用语】欧式中世纪沿海城邦；点烛不烧香｜禁用：上香、烧香、时辰、老夫
-【风格基线】complexity μ=2.42  modifierDensity μ=22.2  abstractDensity μ=10.06  actionDensity μ=122.45 …
+【风格基线】complexity μ=2.42  modifierDensity μ=22.2  abstractDensity μ=10.06  actionDensity μ=122.45 …（取样范围：全书除本章外的 7 章）
 【原著锚段·照这个味道写】[对话] …  [心理] …
 【上一章自检】第 7 章六维对照：全部维度在容差带内 ✓
 【本章禁用清单】- 禁词：上香（建议改用：点烛）　- 禁词：时辰（建议改用：钟点）
@@ -169,7 +172,7 @@ fixAnchors：3 条原著锚段（对话 / 心理 / 描写各一条，供逐句
 7. **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。
 8. **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。
 9. **全工具 UI 开关**：左栏「写作助手功能」入口 + 主栏面板（**DSH 0.2.0 官方席位**；老宿主 < 0.2.0 自动回退旧挂载）——总开关 + 工具开关分组 + 功能开关 + 提示词档位/场景/**精简工作流**，大白话文案，显示数据目录占用与语义引擎状态。颜色全走 `--dsw-alias-*` 语义 token（三主题与官方同源）、控件用官方 UI primitives、品牌蓝强调、光晕头部与分区 kicker、各级页面等宽。<br>**界面再进一步（方案 A）**：源码分块 + 构建链装配（构建失败不写盘、`--check` 校验源与产物一致）、6 个统一组件与单一行模式（`PreferenceRow`）、分组卡片贯穿全部开关页、19+5 枚线图标、阈值 slider 辅助输入、页面级 push-pop 过渡动画（零 state / 零 hook）。
-10. **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 ⚠）；面板「六维基线」视图可自定义每维 ±% 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. **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 ⚠）；面板「六维基线」视图可自定义每维 ±% 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。<br>**基线范围（v6.5.5，按书选择）**：基线不必永远是全书——面板「风格基线」里选**按全书**（默认，等于旧行为）/ **按部分**，再用章节选择器勾出你认作标杆的章，基线就**只用**这几章（是「只用」，不是「不够再补」）。这是给「AI 续写越写越像 AI 自己」准备的：从第 2 章续写起，模型参考的其实是它自己写的上一章，而**从全书取样的基线自己也在漂**（书越长 AI 占比越高）。把标杆钉死在你亲手写的那些章上，插件就能回答「**相对这几章**，新章偏了多少」——它不判断你有没有漂移，那一项本就不可判定。正在检查的那一章**永远排除**（否则等于拿自己跟自己比）；勾中的章少于 3 章时推荐容差自动放宽到**至少 ±25%**，所以单章也能当基线；取不到的章（已改名 / 已删除，或**正文过短无法测量**）会被跳过并如实计数，**一章都取不到**时自动回退「按全书」并在返回体里写明原因。范围按书保存（`<书库根>/.novel-writer/baseline/<书名>.json`），换书互不影响。
 11. **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**。
 12. **原创模式与创作资料**：面板「原创模式」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。
 13. **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/lib/brief.js

@@ -19,6 +19,7 @@ import {
   bookDir,
   buildStyleAnchorPackage,
   buildTolerance,
+  resolveBaselineScope,
   creationDir,
   CREATION_FILES,
   featureEnabled,
@@ -463,8 +464,24 @@ export async function buildChapterBrief(root, book, chapterArg, opts, exec) {
   // v5.0.0 修正：锚段/骨架必须取自**其它章**（与 novel_fix_plan 的锚段口径一致）。
   // 目标是已存在的章（改稿场景）时把本章自身排除——否则"照着原著范本改"可能照着正在改的这一段，
   // 把待修的病句当成标杆。chapter="next" 时目标章不存在，等价于不排除。
-  const anchorChapters = targetChapter ? chapters.filter((c) => c.file !== targetChapter.file) : chapters;
-  const anchorTexts = targetChapter ? texts.filter((t) => t.file !== targetChapter.file) : texts;
+  //
+  // v6.5.5：这一处的取样范围交给**基线范围**的单一裁决点（按全书 / 按用户选中的部分章）。
+  // 为什么锚段也要跟着走：这一段的用途就是给文风当锚（名字就叫「原著锚段」）。
+  // 若基线按部分、而锚段仍取全书，用户指定"以这几章为标杆"就没落到**最容易影响生成**的那一处：
+  // 书越长、AI 续写占比越高，锚段越会变成 AI 自己的文风——issue #1 的漂移正是这么来的。
+  const anchorScope = await resolveBaselineScope(baseRoot, safeBook, chapters, targetChapter ? targetChapter.file : "", exec);
+  const anchorFileSet = new Set(anchorScope.items.map((c) => c.file));
+  const anchorChapters = anchorScope.items;
+  const anchorTexts = texts.filter((t) => anchorFileSet.has(t.file));
+  // v6.5.5：这里的措辞**必须跟上裁决点给的原因**（独立验证代理抓到本行是全文唯一残留的「已失效」硬编码）：
+  // 旧句只要回退就写「自定义基线已失效（N 章改名或删除）」，而另一种原因是"选中的只有本章自身"——
+  // 那句话在那条路径上不仅原因说错，还把"按设计排除"量化成"0 章改名或删除"，是半句空话。
+  // degraded 是**渲染进 <content> 给模型看的**，口径必须与 baselineScope 一致，否则同一份开写包自相矛盾。
+  if (anchorScope.fellBack) {
+    degraded.push("原著锚段/六维基线：" + (anchorScope.reason === "only-target"
+      ? "选中的章只有本章自身（不能作为自己的基线），已回退按全书"
+      : "自定义基线已失效（" + anchorScope.missing + " 章已改名或删除），已回退按全书"));
+  }
   let anchors = [];
   let skeletons = [];
   try {
@@ -491,7 +508,9 @@ export async function buildChapterBrief(root, book, chapterArg, opts, exec) {
   }
   // v5.0.0：目标章已存在时把它排除出基线语料（与 novel_style_check 同口径）；"下一章"尚未存在，
   // 全部现有章都是它的参照系。
-  const baselineCorpus = targetChapter ? perChapter.filter((pc) => pc.file !== targetChapter.file) : perChapter.slice();
+  // v6.5.5：范围同样走基线范围的单一裁决点（按全书 / 按用户选中的部分章），与锚段、六维基线一致。
+  const baselineScope = await resolveBaselineScope(baseRoot, safeBook, perChapter, targetChapter ? targetChapter.file : "", exec);
+  const baselineCorpus = baselineScope.items;
   let baseline = null;
   if (baselineCorpus.length > 0) baseline = computeBaselineFromPerChapter(baselineCorpus);
   else if (perChapter.length === 0) degraded.push("六维基线不可用：全书没有正文达 40 字以上的章节（书太短或章节为空）");
@@ -501,7 +520,10 @@ export async function buildChapterBrief(root, book, chapterArg, opts, exec) {
   let lastVerdict = "";
   if (prevChapter) {
     const prevMetrics = perChapter.find((pc) => pc.file === prevChapter.file);
-    const others = perChapter.filter((pc) => pc.file !== prevChapter.file);
+    // v6.5.5：这里也走同一套基线范围。注意**排除的是上一章**（被判定的是它，不是目标章）——
+    // 范围（哪些章算标杆）只有一个，而"排除谁"取决于在判定哪一章，两者不要混为一谈。
+    const prevScope = await resolveBaselineScope(baseRoot, safeBook, perChapter, prevChapter.file, exec);
+    const others = prevScope.items;
     // v6.2.0 修正（缺陷 #65①）：原来写的是 `!prevMetrics || !prevMetrics.metrics`，而 perChapter 的每一项
     // 都由 style-metrics.computeBaseline 写成 `{ file, metrics }`（缓存也只存这个形状），后半句恒为 false。
     // 「上一章正文不足 40 字」的表现是 perChapter 里**没有这一项**（computeBaseline 会跳过短章），
@@ -859,6 +881,12 @@ export async function buildChapterBrief(root, book, chapterArg, opts, exec) {
     worldview,
     lexicon,
     baseline,
+    // v6.5.5：**披露基线范围**。锚段与六维基线都可能只取自用户选中的部分章
+    //（按部分 = 只用勾选的章节），不写出来，模型会把"你选的 6 章"当成"整本书"来校准语感——
+    // 那恰恰是这个功能想避免的事。fellBack（选择全部失效→回退全书）也已含在 scope 文案里。
+    // 由独立复核代理指出「开写包原本完全不披露范围」后补上（与 style_check/fix_plan 同口径）。
+    baselineScope: baselineScope.scope,
+    baselineMode: baselineScope.mode,
     anchors,
     skeletons,
     lastVerdict,