Skip to content

AI-UI Handbook

What is this? AI-UI is a CLI that measures the gap between what your docs promise and what your UI delivers. It crawls your running app, parses your documentation, and tells you which features are invisible. The core pipeline is deterministic — optional local AI commands extend it with semantic matching and patch generation.

Who is it for? Frontend developers and design teams who want to keep their UI honest. If your README says “ambient soundscapes” but no button leads there, AI-UI finds that.

Status: Stable (v1.1.6) — 20 commands, 880 tests

TL;DR: Point it at your docs and your dev server. It tells you what’s buried, what’s missing, and what to fix. ai-ui stage0 runs the deterministic pipeline. ai-ui ai-hands generates patches.


The problem: SPAs accumulate features faster than they accumulate navigation paths to those features. Docs promise capabilities that users can’t find. Nobody notices until a support ticket arrives.

What AI-UI does better: It doesn’t guess. It builds a trigger graph from real browser interactions, fuzzy-matches features to triggers deterministically, and produces actionable verdicts: promote, keep, demote, merge. Then optional AI commands (Brain, Eyes, Hands) close the remaining gaps with semantic matching, visual enrichment, and ranked code patches.

What it doesn’t do: It doesn’t redesign your app. It measures discoverability and tells you where the gaps are. The AI commands propose fixes but never apply them — all output is for human review.

If you only remember one thing about this repo, remember: it closes the loop between documentation promises and UI reality.

This is not for: visual regression testing, accessibility auditing, or performance profiling. Those are different tools.


Back to landing page