Architecture

Overview

RepoMetro is built as a monorepo with npm workspaces. The core analysis engine is shared with the GitHub Action — no analysis logic is duplicated.

Packages

@repometro/core

The heart of RepoMetro. Contains:

• Scanner (scanner/) — File system traversal, binary detection, line counting, .gitignore parsing
• Analyzers (analyzers/) — Modular analysis components:
  • filesystem.ts — Builds the file tree
  • languages.ts — Detects languages by extension (extensible registry)
  • statistics.ts — Calculates file/dir stats, largest files, extension counts
  • tests.ts — Detects test files and testing frameworks
  • dependencies.ts — Parses dependency manifests
  • git.ts — Analyzes git history and PR changes
  • important-files.ts — Detects README, LICENSE, configs, GitHub files
  • security.ts — Filters sensitive files
• Generators (generators/) — Output formatters:
  • json.ts — JSON output
  • markdown.ts — Markdown report
  • svg.ts — SVG visualizations

@repometro/action

GitHub Action entry point. Uses @actions/core and @actions/github for GitHub integration, including PR comments.

Data Flow

Repository files
    ↓
  Scanner (traverse, filter, classify)
    ↓
  Analyzers (languages, stats, tests, deps, git)
    ↓
  RepoMetroData (the JSON structure)
    ↓
  Generators (JSON, Markdown, SVG)
    ↓
  Artifacts / PR comments

Design Principles

1. No AI — Everything is deterministic
2. No external APIs — Analysis runs entirely from local files
3. No file contents — Only metadata is analyzed
4. Extensible — New analyzers implement the Analyzer interface
5. Fast — Streaming traversal, size limits, no unnecessary file reads
6. Zero-cost — Runs in GitHub Actions, no paid services