Intelligence Applied — Structured creative simulation through AI-assisted novel creation.
We see intelligence as a two-way process:
- Intelligence In — absorbing, organizing, and storing knowledge (study, research, reading)
- Intelligence Out — applying knowledge to construct coherent, functioning systems (creation, simulation, design)
Some read novel creation as art. Some read it as self-exploration. We read it as the application side of intelligence.
Novel creation is one of the most demanding forms of intelligence application. It requires you to:
- Define a world with consistent rules (system architecture)
- Create agents with internal logic and constraints (entity modeling)
- Design interactions that produce emergent outcomes (relationship dynamics)
- Maintain consistency across thousands of decisions (state management)
- Iterate based on feedback (testing and refinement)
That's not just art — that's engineering a simulation. The fact that the output is prose doesn't change the underlying cognitive process.
LuminTree-Intel is our tool for exploring this applied side of intelligence. A standalone web app with a 4-panel layout for structured worldbuilding, plot design, character profiling, chapter management, and AI-assisted writing.
The tree isn't organized by writing craft categories. It's organized by system components — an intelligence architecture for creative simulation:
📖 Core Concept → The thesis: genre, theme, logline, core conflict
🌍 Worldview Setting → The system: world rules, geography, magic, history, factions
📐 Plot Framework → The causal logic: story arcs, turning points, subplots
👤 Character Profiles → The agents: bios, motivations, relationships, behavioral boundaries
📑 Chapter Structure → The execution plan: outlines, scenes, pacing, POV assignments
🗂 Writing Materials → The interface layer: style directives, references, research notes
- 4-panel resizable layout — Tree (structure) · Editor + AI Assistant (refinement) · Chapters (execution) · AI Writer (generation)
- Category-aware AI assistant — Each category has a specialized AI prompt. Ask about characters → character expert. Ask about plot → plot expert.
- AI Writer — Dedicated panel that reads all your settings and strictly follows your Writing Material Library style directives. No drift, no breaking character.
- Use as Draft — One click to push AI output into the chapter editor
- Voice input — Chinese/English speech-to-text for all AI chat inputs. Double-click 🎤 to switch language.
- Smart Import — Feed it any file (JSON, TXT, MD) — even messy formats — and AI auto-classifies content into the correct categories
- Compile & Export — One-click compile all chapters into Markdown or PDF with automatic chapter pagination
- Two-way chapter sync — Chapter Structure tree and Chapters panel always stay linked
- Drag & drop — Rearrange tree node relationships by dragging
- Persistent AI history — Conversations saved per node and per chapter
- Clear Memory — Reset AI Writer conversation per chapter for a fresh start
- Full JSON export/import — Complete project backup including tree, chapters, and all AI chat history
- Local-first — All data in localStorage, nothing uploaded, your story stays yours
- Bring your own API key — Presets for DashScope (Qwen) and OpenAI, or any compatible endpoint
- Clone this repo
- Open
index.htmlin your browser (orpython3 -m http.server 8080for voice input support) - Click ⚙️ to configure your AI API
- Start building your simulation
Click ⚙️:
| Field | Example |
|---|---|
| Preset | DashScope (Qwen) / OpenAI / Custom |
| API Endpoint | https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions |
| API Key | Your API key |
| Model | qwen-plus, gpt-4o-mini, etc. |
Any OpenAI-compatible endpoint works.
- Standalone Web App (HTML / CSS / JS) — zero backend, zero dependencies
- localStorage for persistence
- Marked · Mermaid · KaTeX (all bundled locally)
情报的应用侧 —— 通过 AI 辅助小说创作进行结构化创意仿真。
我们把情报看作一个双向过程：
- 情报输入 —— 吸收、组织、存储知识（学习、研究、阅读）
- 情报输出 —— 运用知识构建连贯、自洽的系统（创作、仿真、设计）
有人把小说创作看作艺术，有人看作自我探索，而我们把它看作情报的应用侧。
小说创作是最具挑战性的情报应用形式之一，它要求你：
- 定义一个规则自洽的世界（系统架构）
- 创建具有内在逻辑和约束的角色（实体建模）
- 设计产生涌现结果的交互（关系动力学）
- 在成千上万个决策中保持一致性（状态管理）
- 根据反馈迭代优化（测试与改进）
这不仅仅是艺术——这是在工程化地构建一个仿真系统。输出恰好是文字，并不改变背后的认知过程。
LuminTree-Intel 是我们探索情报应用侧的工具。一款独立 Web 应用，采用 4 面板布局，提供结构化的世界观构建、情节设计、角色建模、章节管理和 AI 辅助写作。
这棵树不是按写作技巧分类的，而是按系统组件组织的——一个面向创意仿真的情报架构：
📖 核心设定 → 论题：类型、主题、核心冲突、目标读者
🌍 世界观设定 → 系统：世界规则、地理、魔法体系、历史、势力
📐 情节框架 → 因果逻辑：故事弧线、转折点、支线剧情
👤 角色档案 → 智能体：人物传记、动机、关系网、行为边界
📑 章节结构 → 执行计划：章节大纲、场景分解、节奏、视角
🗂 写作素材 → 接口层：风格指令、参考片段、研究笔记
- 4 面板可调布局 —— 树结构 · 编辑器+AI助手 · 章节 · AI写手
- 分类感知 AI 助手 —— 每个分类有专属提示词，问角色问题就是角色专家，问情节就是情节专家
- AI 写手 —— 读取所有设定，严格遵循写作素材库中的风格指令，不跑偏，不出戏
- 一键用作草稿 —— AI 输出满意？一键推送到编辑器
- 语音输入 —— 中英文语音转文字，双击 🎤 切换语言
- 智能导入 —— JSON、TXT、MD 甚至乱格式，AI 自动分类到正确位置
- 编译导出 —— 一键编译全书为 Markdown 或 PDF，每章自动分页
- 双向章节同步 —— 章节结构树与章节面板始终保持链接
- 拖拽排序 —— 拖拽调整树节点的父子关系
- 持久化 AI 记录 —— 按节点和章节保存对话
- 清除记忆 —— 按章节重置 AI 写手对话
- 完整 JSON 导入/导出 —— 包含树、章节和所有 AI 聊天记录
- 数据全在本地 —— localStorage 存储，不上传任何服务器
- 自带 API Key —— 支持通义千问、OpenAI 或任何兼容接口
- 克隆本仓库
- 浏览器打开
index.html（或python3 -m http.server 8080以支持语音输入） - 点击 ⚙️ 配置 AI API
- 开始构建你的仿真系统
- 独立 Web 应用（HTML / CSS / JS）—— 零后端，零依赖
- localStorage 持久化存储
- Marked · Mermaid · KaTeX（均本地打包）