为什么值得关注

它结合了无需登录的免费绘图工具、可移植的开放格式 (OKF) 和 AI 驱动的洞察生成，使数据建模变得易于访问且协作性强，同时避免供应商锁定。

适合谁使用

数据工程师和分析师
数据架构师和建模人员
构建数据产品的产品经理
BI 团队原型设计数据市场

典型使用场景

以可视方式绘制数据表和关系
通过 AI 探索数据模型可以回答的业务问题
通过 URL 共享模型图（无需后端）
将模型导出为 OWOX Data Marts 草稿表

项目优势

完全免费且匿名；设计、模板和共享无需登录
字段级 ERD 视图，显示列、键和连接详情
开放知识格式 (OKF) 通过 Markdown/YAML 文件确保可移植性
AI 生成的洞察问题仅基于模式元数据（不发送行级数据）

使用前须知

推送到 OWOX 需要 OWOX API 密钥（登录模态框）
AI 问题生成需要在服务器环境中设置 GEMINI_API_KEY
OWOX API 密钥存储在浏览器 localStorage 中，存在 XSS 风险（通过 CSP 缓解）

README 快速开始

OWOX Model Canvas

Draw a data model on a free, anonymous canvas — then push it into OWOX Data Marts in one click.

▶ Open the live app: model.owox.com — free, no sign-in.

OWOX Model Canvas is a Miro-like canvas for data people. Sketch your data marts (tables, views, SQL or connector outputs) as nodes and joinable relationships as edges, start from ready-made templates, see the business questions a model unlocks (AI), share it as a link, and — when you're ready — push the whole thing into OWOX Data Marts as drafts. The canvas is 100% free and works without an account; an OWOX API key is needed only to Push.

Why use it

Free & anonymous — design, template, AI-explore and share without signing in.
Templates for real domains — E-commerce, SaaS, Marketplace, Marketing / Lead-gen, Mobile / Gaming, Finance, Healthcare, plus public datasets (Bitcoin, Stack Overflow).
Field-level ERD view — see columns, primary keys and join keys, not just boxes.
Insight Questions (AI) — the business questions a model can answer, generated from schema metadata only (no row-level data leaves your browser).
Build it with AI — copy a short brief into Claude, ChatGPT or Gemini and it generates an importable model. In-app guide: /ai-instructions.html.
Shareable links — the whole model travels inside the URL; anyone can open your exact diagram, no backend.
One-click push to OWOX — turn the diagram into draft Data Marts + joinable relationships.

Open Knowledge Format (OKF) — your model stays portable

The canvas reads and writes Open Knowledge Format — a vendor-neutral way to describe a data model as a folder of Markdown files with YAML frontmatter: one document per data mart, a schema table, and a ## Joins section for relationships. It's human-readable and agent-friendly — people can review it in a pull request and LLMs can author it.

Author + export, as a pair. Design visually here and Export OKF to get a bundle that round-trips — re-importing restores the model. OWOX publishes real bundles at OWOX/models; this canvas is the visual authoring/export front-end for that same format.
No lock-in. Be

项目描述

OWOX Model Canvas — visual editor for data models in the Open Knowledge Format (OKF)

owox-model-canvas