feat(extraction): Hugo design-system support — Markdown, Templates, CSS/SCSS, YAML, JSON

[alexander-kastil]

What is Hugo?

Hugo is one of the most widely used open-source static site generators, built in Go. It powers documentation sites, corporate websites, developer blogs, and government portals across the industry. A Hugo project is organised around a distinct file tree: content (Markdown with front matter), layouts (Go html/template partials and base templates), assets (CSS/SCSS, JS), and data (YAML/JSON/TOML configuration and site data). Until now CodeGraph treated almost all of these as opaque blobs.

Why index CSS / SCSS?

Design systems live in CSS. When an AI agent needs to understand why a component looks the way it does, or wants to trace where a token like --color-primary is defined and consumed, it currently has no graph to walk — it can only grep. With full CSS/SCSS extraction:

• Design tokens (--custom-properties, $scss-variables) become first-class variable nodes — searchable, linkable, and traceable across files.
• Mixins and utility functions become function nodes, so an agent can find all callers of @include button-base() without a full-text search.
• @layer declarations become namespace nodes, making cascade-layer architecture visible in the graph.
• @use / @forward / @import produce real imports edges, so the SCSS module graph is walkable exactly like a TypeScript import graph.

This matters most for Hugo sites (which commonly have a bespoke SCSS design system), but the extractors work on any project that uses CSS or SCSS.

What this PR adds

New language support

Language	Extensions	What's extracted
Hugo Markdown	.md .markdown	Front matter fields (title, date, tags, …) as property nodes; headings as class/function nodes
Hugo Templates	.html (under layouts/ or themes/), .gohtml .tmpl .gotmpl	File as function node; {{ define }} blocks; {{ partial }} / {{ partialCached }} calls; {{ block }} slots; {{ template }} calls; $variable assignments
CSS	.css	CSS custom properties (--token), @keyframes, @layer, @import
SCSS / Sass	.scss .sass	Everything from CSS, plus $variables, @mixin, @function, %placeholder, @use / @forward, @include calls, @extend references
YAML	.yml .yaml	Upgraded from file-level tracking only → tree-sitter extraction of top-level mapping keys as variable nodes
JSON	.json	Tree-sitter extraction of top-level object keys as variable nodes

Hugo Template path gating

Plain .html files (e.g. email templates, test fixtures) are not processed. Only .html files whose path contains layouts/ or themes/ are routed to HugoTemplateExtractor. Files with explicit Hugo extensions (.gohtml, .tmpl, .gotmpl) are always processed regardless of path.

Infrastructure

• src/types.ts — six new Language union members: json, markdown, gotemplate, css, scss, sass
• src/extraction/grammars.ts — WASM grammar entries for YAML and JSON (tree-sitter); extension map entries for all new types; isLanguageSupported / isGrammarLoaded / getSupportedLanguages updated
• src/extraction/languages/index.ts — YAML and JSON extractors registered
• src/extraction/tree-sitter.ts — dispatch branches for the four new custom extractors

Test updates

• __tests__/extraction.test.ts — replaced two assertions that incorrectly expected .css and .json to resolve to unknown
• __tests__/security.test.ts — updated isSourceFile expectations now that CSS and Markdown are supported extensions

Test plan

• npm test passes
• npm run build succeeds (YAML + JSON WASM files copy into dist/)
• Index a Hugo project and verify Markdown front matter fields appear as nodes
• Index a Hugo project and verify {{ partial "nav.html" . }} produces a calls edge
• Index an SCSS design-system file and verify $primary-color appears as a variable node
• Confirm .html files outside layouts/ are not indexed as Hugo templates

🤖 Generated with Claude Code