UI Component Similarity Toolkit

This project implements the complete pipeline outlined in the research plan for measuring similarity between web UI components. It includes utilities for discovering and cloning open‑source repositories, scanning those repositories for React/Angular/Vue components, extracting multi‑layer feature representations (DOM structure, CSS styling and behavioural information), building a corpus, splitting the corpus into training and query sets, constructing a proximity index over the components, and running similarity queries.

The code is organised into several modules:

• core – data models for component signatures and their constituent feature classes, together with metric implementations that compute distances between signatures.
• extractor – framework‑agnostic and framework‑specific feature extractors that map raw source files to ComponentSignature objects.
• scanner – repository scanners that traverse cloned repositories and produce ComponentSourceRef objects describing where each component’s template, logic and style live on disk. Separate scanners exist for React, Angular and Vue projects.
• corpus – data structures for binding source references to extracted signatures (ComponentRecord) and grouping them into a corpus (ComponentCorpus) for later indexing and experimentation. The corpus layer also provides convenience functions for train/query splits.
• index – a permutation‑based proximity index (PermutationIndex) that organises components in metric space and supports fast approximate nearest neighbour queries.

The top‑level Main.kt script demonstrates the full pipeline on a user‑specified directory of cloned repositories. It scans each repository, extracts component signatures, builds a corpus, constructs a permutation index using randomly chosen pivots, and then performs similarity queries on the query split. Although network operations (e.g. cloning from GitHub) are not performed directly in this codebase, the architecture anticipates an external discovery/cloning step to populate /data/repos/… before running the scanner.

Note: This project uses Kotlin for its implementation. The code files provided here represent a self‑contained library and CLI, but compilation and execution are not performed within this environment. Researchers can download the source, set up a Gradle project with the Kotlin standard library, and run the CLI on a corpus of repositories on their own machines.