Skip to main content
Gremorie

Overview

Framework-agnostic canonical knowledge — when to use a button vs link, label rules, hierarchy, feedback. The "what / when / why" that components reference.

Principles

Heuristics

Nielsen’s usability heuristics as a review checklist — the foundational rules behind every interface decision.

Visual hierarchy

Grouping, spacing and emphasis — Gestalt applied so the eye reads the right thing first.

Patterns

Components

When and how to use each primitive — button vs link, dialog vs drawer, form anatomy. The UX rationale, not the code.

Patterns

Reusable interaction patterns — destructive confirmation, navigation choices — and when each one fits.

States & feedback

States

Loading, empty and error: the three states every screen needs and most forget.

Feedback system

Toast vs modal vs banner — choosing the right feedback surface by the stakes of the message.

Writing

UX writing

Microcopy rules — button labels, error messages — so the words match the action.

The UX corpus is the canonical knowledge layer of Gremorie. Text, not code. Framework-agnostic. It answers the questions every component doc shouldn't have to re-explain:

  • When to use a button versus a link
  • Label rules (length, voice, imperative vs descriptive)
  • Visual hierarchy and grouping
  • Feedback system (toast vs modal vs banner) by stakes
  • Loading / empty / error states
  • Destructive action confirmation patterns

Not to be confused with the component catalogue. The catalogue is the code the registry distributes; this corpus is the text knowledge components reference and that the MCP serves via get_guidelines.

Article shape

Each article must include:

  • Frontmatter — title, description, related components, tags
  • TL;DR — one paragraph
  • The rule — the prescriptive answer
  • Why — rationale, references to research
  • How to apply — concrete examples (with do / don't snippets)
  • Counter-cases — when the rule doesn't apply
  • Sources — academic and industry references

How the MCP serves this corpus

The MCP get_guidelines(topic?) tool returns the matching corpus article as MDX, so language models can apply the rules when generating code. The corpus is the explanation layer that makes generated code idiomatic and not just syntactically correct.

On this page

Article shapeHow the MCP serves this corpus