01 / 04Case study / 01 of 04
Web and Android / Commerce
Design + full-stack engineering
In one line
Manga, found.
Fans sell manga and anime merch to each other, and a map shows the nearest shop that has the volume you want.
01 / The problem
01 / The problem
A whole world of manga. No single place to find it.
Manga fans had to search bookstore after bookstore to see what was available locally. Titles were scattered across retailers, often listed under inconsistent or incomplete names.
Second-hand volumes and merch had nowhere to go, and there was no way to tell which shop near you actually had a book. Finding one meant asking around all over again.
02 / The solution
02 / The solution
A marketplace where fans trade with each other, and a map that knows which shop has it.
Fans list second-hand manga, figures, clothing and vinyl in a moderated marketplace and reach each other on Telegram.
Behind the map, Hiki reads every store's shelves each day and enriches each title from anime databases, so readers find the exact edition and the nearest shop that has it, in their own language.
How it flows
Runs daily on Cloud Scheduler
Daily
01 / Collect
aiohttp scrapers with rate limiting and exponential backoff
02 / Enrich
AniList and Shikimori, with Jikan and Kitsu as fallbacks
03 / Store
FastAPI with async MongoDB via Motor
04 / Serve
Next.js on Cloud Run, live at hiki.world
03 / Key features
03 / Key features
Each feature answers a problem readers had, from finding a title to finding where to buy it, in a shop or second-hand. Together they turn a search across several stores into one.
F.01
Moderated
Fans sell second-hand manga, anime figures, clothing and vinyl, with the condition and up to five photos. An admin reviews every listing before it goes live; buyers reach sellers on Telegram.
F.02
Synced nightly
Search an exact edition by title, volume or ISBN and pick a city: the map shows the nearest Librarius and Cărturești shops that have it, with address, phone and opening hours. Moldova today, Romania next.
F.03
Full-text search
Filter by genre, tag, stock and store. Full-text search works in every language.
F.04
Built with next-intl and locale-aware routing, so every reader browses the catalog in their own language, titles included.
F.05
4 sources
AniList supplies scores, genres, tags and covers. Shikimori translates Cyrillic titles, with Jikan and Kitsu as fallbacks. Even mangled product names resolve.
F.06
Daily
Scrapers read every store's shelves each day: names, stock and images. New arrivals and titles back in stock are flagged.
04 / Built with
04 / Built with
Next.js
React
TypeScript
Tailwind CSS
shadcn/ui05 / Results
05 / Results
Manga fans in Moldova have one place to trade with each other and to find the nearest shop that has their next volume.
Every listing is reviewed before it goes live, and the map is synced with the shops every night. Romania is next.
Live
Next product
Case study / 02 of 04
Open source / 2026
SEO audits inside your AI editor. 28 rules, a full site crawl, no API keys. Runs locally.
SEO audits inside your AI editor. 28 rules, full site crawl, no API keys.
Node.js · TypeScript · MCP
StartStart / 30-minute call
No corporate theater. Just results.
Prefer email?
nicholas@larch.ee
Kotlin
Jetpack Compose
Firebase
Python
FastAPI
MongoDB
Google Cloud