Skip to content

StoryMemory Studio v1.0.0

Latest
Latest

Choose a tag to compare

View all tags
@Jackeyhate9 Jackeyhate9 released this 20 May 04:08
· 9 commits to main since this release
v1.0.0
f803ee8

StoryMemory Studio v1.0.0

Local-first AI writing control center for long-form fiction and IP development.

本地优先的长篇小说与 IP 创作中控台。

StoryMemory Studio is built for web novel authors, short drama writers, comic adaptation creators, and anyone who needs an AI writing system that remembers characters, facts, foreshadows, timelines, style, and forbidden settings across long projects.

StoryMemory Studio 面向网文作者、短剧编剧、漫画改编创作者。它不是简单的一键写小说工具，而是把结构化记忆、长上下文 Prompt 编排、章节生成、一致性检查、AI 腔治理和 IP 改编放进同一个本地桌面工具里。

Download / 下载

Download StoryMemoryStudio.exe from the release assets below.

下载下方 Release Assets 中的 StoryMemoryStudio.exe，双击即可启动本地前端。

Highlights / 核心亮点

  • Local-first SQLite Story Memory database
  • DeepSeek-ready long-context prompt orchestration
  • GLM-5.2 / Z.ai provider preset
  • Ollama local model path for private/offline writing
  • OpenAI-compatible API support
  • Context budgets: lite 32k, standard 128k, deepseek_long 800k, full_audit 1M
  • Create Novel Wizard for starting from zero
  • Chapter import, generation, editing, versioning, and docx export
  • Style Profiler and humanizer-zh integration
  • AI Tone Detector and Novelization Rewriter
  • Foreshadow, timeline, character arc, pacing, and platform adaptation tools
  • IP adaptation matrix for comic, short drama, video storyboard, Xiaohongshu, posters, and character cards

Long-Context Memory / 长文本记忆

StoryMemory Studio does not blindly paste the entire novel into a model. It stores facts in SQLite first, then Context Builder assembles a structured prompt by priority:

  • S: current chapter goal, outline, current cast, current location, relevant foreshadows, previous chapter
  • A: hard world rules, timeline nodes, forbidden contradictions, confirmed facts
  • B/C/D: long-line summaries, unresolved questions, style profiles, platform rules, compressed archives

DeepSeek-style long context is used for global reasoning and consistency. SQLite remains the editable source of truth.

系统不会把全文粗暴塞进上下文，而是先把人物、伏笔、时间线、章节事实写入 SQLite，再由 Context Builder 按优先级组装 Prompt。DeepSeek 百万 token 级长上下文适合全局理解和一致性审查，但事实来源仍然是可编辑、可追踪的本地数据库。

First Launch / 首次启动

  1. Double-click StoryMemoryStudio.exe.
  2. Wait 10-30 seconds.
  3. The browser should open automatically.
  4. If it does not open, check start_log.txt next to the exe.

首次启动可能需要等待 10-30 秒。如果浏览器没有自动打开，请查看 exe 同级目录下的 start_log.txt。

Data Safety / 数据安全

User data is stored locally:

  • database: data/
  • exports: exports/
  • settings: .env

默认不会上传用户作品数据。数据库、导出文件和配置均保存在 exe 同级目录。只有当用户主动调用 DeepSeek / GLM / OpenAI-compatible 等云端模型时，当前上下文才会发送到对应服务。Ollama 路径则走本地模型。

Website / 项目页

Note / 注意

This MVP executable is built with PyInstaller and is not code-signed yet. Windows or antivirus software may display a warning for unsigned executables.

当前 MVP exe 由 PyInstaller 打包，尚未做代码签名。Windows 或杀毒软件可能提示未知发布者。

Assets 3
Loading