Specs

A living inventory of what Levantrain is built on: open-source libraries, AI tools used to build and run it, and external data sources. Expect this to fill in as the site grows.

Last updated: July 2026.

Foundation

TanStack Start + Router
MIT
SSR-friendly React app shell and file-based routing
React 19
MIT
UI runtime
Vite
MIT
Build tool and dev server
TypeScript
Apache-2.0
Typed codebase — strict type-checking is being rolled out (some files still opt out)
Tailwind CSS v4
MIT
Styling via semantic tokens in src/styles.css
TanStack Query
MIT
Server-state cache and loaders
shadcn/ui + Radix primitives
MIT
Accessible component building blocks
Lucide icons
ISC
Icon set used throughout the UI
react-hook-form + Zod
MIT
Form state and schema validation
Tiptap
MIT
Rich-text editor for blog posts in the admin
DOMPurify
MPL-2.0 / Apache-2.0
Sanitizes rich-text HTML before render
Leaflet + react-leaflet
BSD-2-Clause / MIT
Interactive maps of stations and routes

Backend & infrastructure

Supabase (Postgres, Auth, Storage, RLS)
Apache-2.0 + hosted
Database, email/password auth, row-level security, media storage, and Edge Functions (blog OG images, RSS feed)
Cloudflare Workers runtime
commercial
Serverless edge runtime for SSR and server functions
Bun
MIT
Package manager and JavaScript runtime

External data sources

OpenStreetMap
ODbL
Base map tiles for the interactive station/route maps
Curated primary sources
various
Historical station and route data compiled by hand from archival maps, timetables, and secondary literature — see /resources for the reading list

AI tools

Lovable
commercial
AI-assisted product development environment
Claude Code (Anthropic)
commercial
AI software architect and pair-programmer — planning, data modeling, implementation, and code review

Quality & tooling

Vitest
MIT
Unit and component test runner
Playwright
Apache-2.0
End-to-end test runner (auth, blog, and admin flows)
ESLint + Prettier
MIT
Lint and format