Moonlit Library — three credentials, one shipped repo

The brief

Moonlit Library is Dev Academy’s fullstack-collection challenge. The brief is open: build a small fullstack web app — anything you’d actually use — and bring it through the full lifecycle. Mine became a personal library: books I’ve read, books I want to read, with covers and ratings and notes. Express + Knex + SQLite on the server. React + Vite on the client. Live at moonlit-library.onrender.com.

The library framing was intentional. I read carefully, slowly, and I wanted somewhere to track that — somewhere designed for how I read, not how the average reading tracker assumes everyone reads.

Three credentials, one repo

One repo carries three Dev Academy submissions:

• CP02 — Software Quality. The testing + refactoring + code review track. Achieved 2026-06-13. The submission documents what “quality” means in this codebase: tests that cover real behaviour (not coverage theatre), refactoring that earns its place, commit messages that explain the why.
• WD03 — Forms. The form-handling track. Validation, error states, accessibility at the input layer. Submitted 2026-06-15.
• WD04 — Fullstack. The whole thing — Express + Knex + React + deploy. Submitted 2026-06-15.

What made one repo carry all three is incremental delivery. Each phase is a deployable feature with its own tests and its own commit story. Nothing was bolted on at the end to “satisfy CP02”; CP02 was the way the code was already being written, just documented for the submission.

Designed for how I read

I am dyslexic and AuDHD. Two practical consequences for any reading-shaped UI I build for myself:

• No italics in body text. Italics break my reading flow — the letterforms slant out of the muscle memory I use to read at speed. In Moonlit’s BookDetail, book titles do not render in italics. Emphasis is by colour and weight only.
• Real keyboard semantics on custom widgets. MoonRating is a five-moon star-rating widget. It is a radiogroup. Arrow keys move between options, Enter selects, focus is visible at every step. A screen reader walks it correctly without me writing a parallel announcement layer.

The reading width is capped at 66ch. Line height sits at 1.6. Lexend Variable is the body face — designed specifically for readers with dyslexia, with letterforms that are harder to confuse than the standard Inter / Roboto family.

None of this is a11y theatre. It is the rules I live by, applied to my own code so I can use my own product. The fact that it makes the product usable for other people is downstream of that.

When the API broke

I started Moonlit with Google Books for cover images and metadata. Two weeks in, the rate limits + the thumbnail policy started biting: the covers were missing or low-res for a meaningful fraction of books, and the fallback experience was worse than the success path. I switched to Open Library: no auth, no rate limit problems at my scale, better cover images, same metadata shape.

Half a day to rewrite the fetch layer. A small loss for a real win — the product now works for the books I actually have, not just the bestsellers Google indexes well.

A second pivot came at deploy. Render’s free tier runs Node, but my project is TypeScript with ESM, and the default node start command does not handle .ts on Node 22 without flags. I switched the start command to tsx. Three flags had to align — NODE_ENV, the build target, and the start runtime. The first deploy failed silently. The second worked. The third was the one I actually wanted.

These are not heroic stories. They are honest ones. Shipping under deadline pressure means making the call that gets you to a working product, then documenting the trade-off in the commit so the next person (often future-me) does not have to relitigate it.

What’s next

Moonlit’s Phase 1 accessibility sweep is open: BookDetail italics removal (already drafted), MoonRating screen-reader walk under actual NVDA (planned). After that, the case-study version of this page lives on — the project’s behaviour evolves but the writeup stays anchored to where it was on 2026-06-16, with the assessments achieved and the live URL working.

Three Dev Academy credentials, one shipped repo, one deployment that survives without me touching it. That is the lesson I am taking out of Moonlit Library.