dsh-novel-writer
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| DSH_NOVEL_WRITER_ROOT | No | Path to the novel library root directory. Overrides the current working directory. If not set, defaults to the current working directory unless --root is provided via command line. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| novel_booksA | 列出小说章节库中的全部作品（novels 文件夹下的子目录），含章节数与总字数。只读。 |
| novel_chaptersA | 列出某部作品的全部章节：章号、标题、字数、行数、更新时间。只读；若存在同章号多文件，列表末尾会给出撞号分组提示。 |
| novel_readA | 阅读某部作品的某个章节，返回带行号的正文（含字数统计）。可用 offset/limit 分段读取长章节。只读：不写任何文件、不改动任何数据。 |
| novel_keywordsC | 统计某部作品（或单个章节）中出现频率较高的关键词：中文相邻二字/三字词组与英文词（novel_keywords 输出不含单字）。 |
| novel_new_chapterB | 为某部作品创建新章节文件（默认自动取下一个章号）。可指定标题与初始正文。 |
| novel_chapter_briefA | 开写包：一次调用装配出「写这一章需要的一切」——上一章结尾原文（承接口）、本章大纲方向、相关人物卡、待回收伏笔（含埋设距离）、世界观用语规范、六维风格基线、原著锚段与句式骨架、本章禁用清单与开写清单。只读不写盘（仅复用既有分析缓存）。写新章前应先调它，而不是逐个工具去翻。 |
| novel_importA | 批量导入原稿件：扫描文件夹自动识别书名/章号并分组，可复制/移动到 novels/<书名>/ 分类存放（scan 预览 / apply 执行）。 |
| novel_sentence_analysisA | 句式模式分析：统计某部作品（或单章）的句式分布（陈述/环境/心理/对话/疑问/反问/感叹/祈使/省略留白九类）、句式排列规律（转移、高频模板、段首段尾、按章节的压缩节奏序列）、段落结构、句长分布、情感曲线、风格指纹与节奏建议，用于快速掌握作者的写作习惯、主观情感并参考其叙事节奏。【问「这本书是怎么写的」时用它】——它与 novel_style_check（问「我这一章写得像不像」）、novel_style_report（问「这本书是什么风格」）共用同一套引擎与缓存，三者是不同的问题、不是重复的工具。受 UI 开关控制，可先用 novel_sentence_config 查看状态。 |
| novel_sentence_configA | 查看或修改句式模式分析的开关状态：enabled（分析功能是否可用）、autoAnalyze（分析作品时是否主动使用）、systemPromptMode（系统提示词档位 off/brief/full）、promptScene（提示词场景 general/writing/revising/auditing/setup，决定完整档下注入哪一套流程）与 leanWorkflow（精简工作流：打开后只按需调用工具，省 token）。 |
| novel_style_checkA | 风格自检：对比某章节与全书其他章节的风格指纹（句式分布/句长/情绪），输出相似度与偏差清单。【问「我这一章写得像不像全书」时用它】——要可执行的改稿待办请接着用 novel_fix_plan，要全书风格画像用 novel_style_report，要看写作手法用 novel_sentence_analysis。 |
| novel_fix_planA | 改稿台：把风格诊断变成按优先级排好的待办清单——每条含原句位置（行号区间 + 原文）、当前值与目标值、可参照的原著锚段、以及改写方向。action=plan 生成清单 / verify 对当前正文复测（看哪些已回到带内）/ mark 标记单项（需 itemId + state）。只给方向不给句子：本插件不生成正文。 |
| novel_plotC | 伏笔/剧情线登记表：维护某部作品的伏笔与剧情钩子（open 待回收 / done 已回收）。 |
| novel_settingsB | 设定管理（五张表：人物/地点/道具/时间线/世界观用语规范），list/add/update/delete/scan 按 category 维护，detect 自动判断文化基准。 |
| novel_style_reportA | 风格画像报告：聚合 6 维测量数据（指纹/词汇/题材/情感/氛围 12 轴/语义距离）供 AI 判断风格。【问「这本书是什么风格」时用它】——它给的是全书画像而非单章判定（那是 novel_style_check），也不是写作手法拆解（那是 novel_sentence_analysis）。 |
| novel_summaryB | 章节摘要（模型生成、插件存储）：读摘要回忆剧情、避免重读。add/update/get/list/delete 按章节管理。 |
| novel_continuity_checkB | 连贯性审计：对照设定表扫描全书，输出矛盾候选（数字口径/人物缺场/别名/重复）供修正。 |
| novel_semantic_searchB | 语义检索（本地 embedding，0 token）：自然语言检索全书语义相关段落，无关键词也能命中。 |
| novel_outlineC | 维护原创小说的创作资料（novels/创作资料/<书名>/）：创作设定/主要人物/次要人物/剧情大纲/钩子记录/创作状态卡的初始化、读取与更新。 |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 18 tools
The three analysis tools (novel_sentence_analysis, novel_style_check, novel_style_report) are explicitly disambiguated in their descriptions, but novel_settings, novel_outline, and novel_plot have fuzzy boundaries—both settings and outline manage character/world data, and outline's hook records overlap with plot's foreshadowing tracker. An agent could plausibly route the same request to two different tools.
All tools share a consistent novel_ prefix in snake_case, but the second part mixes conventions: bare verbs (novel_read, novel_import), bare nouns (novel_books, novel_plot), noun+verb compounds (novel_style_check, novel_continuity_check), and noun+noun compounds (novel_sentence_analysis, novel_style_report). The prefix keeps it readable, but the pattern is not predictable enough to guess tool names.
18 tools is at the heavy end, but the server's scope is genuinely broad: reading, writing, importing, editorial analysis, style checking, revision planning, plot tracking, settings, summaries, continuity auditing, and semantic search. Each tool has a distinct job, though the analysis triad plus fix_plan could arguably be consolidated without much loss.
Reading, creating, analysis, and management surfaces are well covered, but there are notable gaps: no way to update or delete a chapter after creation, no book-level delete/rename, and novel_fix_plan explicitly refrains from generating text, leaving no tool that actually applies revisions to the manuscript. Settings and summaries have full CRUD, but the core chapter lifecycle is incomplete.