Skip to main content

Product vision

Audience

This document is for contributors and maintainers (the arbiter of trade-offs). If you want “how to use the product,” start with Start here.

This document answers one question: what this project aims to become, and why. It is the highest arbiter for all feature trade-offs. Changes require maintainer consensus.

One-line positioning​

Git + Figma for database design — the open-source source of truth for data structures in the AI era — bringing version management and real-time collaboration to database modeling; projectJSON and the version system let humans and AI agents read and write the same source of truth.

Positioning upgrade rationale: ADR-0012 (accepted · option B). Public API/MCP: ADR-0013 (📋 planned).

Why this positioning​

Ecological niches for database modeling tools:

  • Minimal tools (dbdiagram.io): pretty and easy, but no collaboration, no versions, closed source
  • General drawing (draw.io): free and flexible, but no database semantics
  • Desktop-only (PDManer, Navicat): feature-heavy, but no collaboration, dated UX

“Collaborative database design” is the only empty niche, and this project already has assets others lack: automatic versioning with diff on every change, three-tier permissions with approval flow, multi-datasource forward/reverse engineering. We lengthen our strengths instead of competing in others’ niches.

Three layers of user value (the standard for judging every requirement)​

  1. First-time visitors: amazed within 30 seconds — drives spread
  2. Power users: modeling work is indispensable; switching cost is very high — drives retention (locked by versions + collaboration)
  3. Potential contributors: code feels well-crafted; willing to open PRs — drives ecosystem (modern stack + engineering quality)

Every requirement review asks: which layer does it serve? If none, we don’t build it.

North Star metric​

Weekly count of active modeling projects that produce a version save (must be a non-empty diff version save; empty bookmark saves are allowed but don’t count toward North Star — see ADR-0022).

Star count is a lagging vanity metric; this metric is real product vitality. Guardrail metrics: CI gate strength, core module test coverage, first-screen bundle size, core journey duration (see Performance budget). Current quarter North Star levers: Roadmap — Next quarter: three things only.

What we don’t build​

  • No mobile / desktop clients — modeling is a desktop browser scenario; early distraction is fatal
  • No closed-source cloud / commercial distraction — win open-source reputation first; monetization comes after winning
  • No AI hype — no conversational black-box magic; AI delivered as “open API/MCP + version source of truth + auditable” platform; agents are first-class citizens of projectJSON
  • No dbdiagram clone — minimal-tool niche is taken; we differentiate on collaboration, versions, and open source of truth
  • No further investment in retired stacks — stopping loss is half of engineering decisions

Experience design principles​

See Design principles. Every interaction change must pass the six principles.

The designer is one main battlefield (alongside Home / landing as shell layers): relation diagram readability and “shareable” aesthetics, auto-layout quality, are product levers alongside versions/collaboration; this quarter’s weight: ADR-0016.

Technical roadmap highlights​

  • Designer canvas migrates to ReactFlow (Strangler strategy, parallel then cutover)
  • Backend Spring Boot stabilizes on 2.7 first; 3.x + JDK 17 is a separate milestone — the two big bangs don’t run in parallel
  • UI converges on Ant Design; JSON processing unified on Jackson
  • Core data format projectJSON is schema-versioned; never break in place (see data-format)