TempestArchive

The idea

Most fan wikis assume you have finished everything. Tempest Archive starts from the opposite premise: you choose how far you have read or watched, and the whole site reshapes itself around that. A character page for someone you have not met yet simply stays out of the way.

It is also written, not scraped. Each story record summarises cause and consequence in the archive's own words, which keeps it useful for readers and clear of the source material.

The challenge

The content is a graph, not a list. A skill has owners and evolutions; a character has affiliations, forms and appearances; an event links all of them. Keeping those relationships consistent, and then filtering the entire graph by spoiler level without breaking links, drove most of the data-model decisions.

The light novel and the older web novel also diverge. Treating them as separate continuities from the start avoided a class of contradictions that would have been painful to untangle later.

The build

A React application built with Vite, styled with Tailwind, with Motion for the interface transitions. The archive data ships as its own bundle so reference pages render immediately and search works client-side.

Every route is also prerendered to static HTML on Cloudflare, so a character or skill page has real content, a real title and real description before any JavaScript runs. Supabase handles accounts; bookmarks also work locally for readers who never sign in.

Analytics are opt-in, first-party and cookie-free. The prompt says exactly that, and defaults to off.

Real screens.

tempestarchive.com

tempestarchive.com

tempestarchive.com

How the archive is organised.

Six reference sections, all cross-linked, all filtered by the reader's spoiler level. Story records are the spine; everything else hangs off them.

What it actually does.

• 01

  Spoiler levels

  Anime, Moderate and Full. Every list, dossier and timeline entry respects the level the reader picks.

• 02

  Two continuities

  Light-novel and web-novel material are labelled separately and never merged into one false ending.

• 03

  Original writing

  Every summary is written from scratch. No novel prose or dialogue is reproduced.

• 04

  Linked records

  Story records name who was present, what each side wanted, which abilities decided the outcome and what returns later. Characters, skills, places and events link to each other.

• 05

  Indexable

  Every route is prerendered with its own title and description. The sitemap covers several hundred character, skill, location, faction and story pages.

• 06

  Accounts

  Sign-in with Supabase for a personal account, alongside local bookmarks that work without one.

Built with.

Identified from the production build: a Vite-built React application with Motion for interface transitions, Tailwind for styling, Supabase for accounts, served prerendered from Cloudflare.

Desktop and phone.

The phone layout is its own design, with a bottom tab bar for the sections readers reach for most, rather than the desktop navigation squeezed down.

tempestarchive.com