Wehi's Boutique: headless Shopify, and a custom AI stylist.

Wehi's Boutique is family-owned by Marcel Rodrigues and her daughter Ani. Two stores on Hawaiʻi Island: the original on Mamane Street in Honokaʻa, opened in 2015, and a Hilo flagship on Furneaux Lane. They have drawn their own aloha-wear prints in-house since 2015. The Hilo flagship has a SipNPretty café inside it, the only boutique on the island with a café inside. They serve locals, the Hawaiian diaspora, and visitors, and they sell into Tokyo.

A brand like that does not need a stock template. It needs a store that carries the prints, the story behind each one, and the way Marcel and Ani actually merchandise. So the build is not a Shopify theme. It is a genuine headless storefront.

What headless Shopify development actually means here

Headless means Shopify still runs the catalog, the cart, and the checkout in the background, and a separate custom application is the store the shopper sees. Wehi's storefront is built on Shopify Hydrogen with React Router 7, which is Hydrogen's server-rendering framework after Remix, plus React 19, TypeScript, and Vite. It is hosted on Shopify Oxygen, which runs on a Workers runtime. Styling is Tailwind CSS on top of a custom brand token system in cream, clay, ink, and gold.

The whole thing is about 19,500 lines across roughly 107 source files and 35 routes. That is real application code, not a theme with some settings changed.

Data flows through three Shopify APIs. The GraphQL Storefront API powers products, collections, cart, and search. A custom app on the Admin API runs about 48 operations scripts for bulk tagging, metaobjects, catalog audits, and drop setup. The Customer Account API handles login, orders, and addresses. Checkout is native Shopify checkout. We did not hand-roll payments, and you should not want anyone to. Money stays on the platform built to handle it.

Pick My Wehi, a custom AI stylist that shops the live catalog

The centerpiece is a custom AI shopping experience called Pick My Wehi. It is a stylist, not a visual try-on and not AR. It is about a thousand lines of resource-endpoint code plus a Claude integration.

The flow is identity-first. A shopper answers five short questions: how she wants to feel, her color mood, her shape and size, the moment or occasion, and her budget with any hard-nos. The system maps her moment to a live Shopify collection, then pulls about 30 candidate products with their variants and custom metafields through the Storefront API. It hard-filters on budget and rules, then scores and ranks the rest on color mood, coverage, and shape.

That ranked pool goes to Claude, which picks three takes on her: one easy, one elevated, one statement. Each pick comes with a written reason. Every pick is then resolved back to the real live product, so the price and photo are always current inventory, never stale. It adds a completing piece so a look is never half an outfit, and every look is one-tap add to cart.

Here is what that looks like. Say she is headed to a lūʻau. She taps effortless, warm tones, her size, the occasion, her budget. Pick My Wehi reads that, pulls the aloha pieces that actually fit, and hands back three looks she can wear on Sunday. One easy, one elevated, one statement, each with a reason, each ready for the bag. No scrolling twelve pages of dresses hoping to guess.

If the AI call ever fails, there is a deterministic local fallback, so the stylist never breaks in front of a shopper. Prompt caching keeps it cheap and fast. That is the difference between a demo and something a store can run every day.

A live-data merchandising layer a theme cannot do

The stylist gets the attention, but the merchandising layer is what makes the store feel current. All of this is custom, and all of it reads real Shopify data.

There is a timed drops system with a countdown, a notify-me capture, and a drop-lock so a product cannot be bought before it goes live. There is a Most Loved row driven by real Shopify best-selling sales data, not a guess about what should sell. Garment-type filtering shows live counts, and search is predictive with typeahead. Product pages carry JSON-LD product schema, and email capture runs through Klaviyo.

A custom wehi_print metaobject documents each Hawaiian print's cultural meaning in English and partial Japanese, so the story behind a pattern travels with the product. The storefront is bilingual with a Japanese path at /ja for the Tokyo market, and free shipping kicks in over $150 on-island.

There is also a separate, password-protected Next.js content studio that helps Ani generate on-brand captions and trend briefs. It sits next to the store rather than inside it, so the shopping experience stays clean.

Custom e-commerce for boutiques and apparel brands

Here is the honest scope. There is no visual try-on and no AR. Payments run on standard Shopify checkout. This is real custom application code, not a template theme. That is the point. A boutique with its own prints, its own café, and a market in two languages does not fit in a theme, and it should not have to.

If you are a boutique or apparel brand and you have outgrown a stock Shopify theme, this is the kind of work we do. You can scope a custom build with us, or see more of the storefronts and systems on the work page. When you are ready, start a project and we will map the build to what your store actually needs.