Pages not yet translated open that language’s home page.
T.04 / Technology
Back end
larch builds back ends in Python with FastAPI. Hiki’s whole back end is Python: a FastAPI service with async MongoDB, aiohttp scrapers and the pipeline that enriches every manga title.
Bring the idea. We’ll tell you how we’d build it and what it takes.
01 / Built with Python and FastAPI
Hiki’s catalog is only as good as its data. The Python back end collects it, cleans it up and serves it.
Web platform / 2025
01
FastAPI with async MongoDB serves the catalog. aiohttp scrapers collect it every day, and an enrichment pipeline resolves each title across four anime databases.
Web platform / 2025
Next.js · FastAPI · MongoDB · Google Cloud
02 / Related service
Fast, resilient platforms that grow with your business, not against it.
What you get
Hiki’s FastAPI back end serves a catalog that readers browse in several languages.
On Hiki we built both services: the Next.js front end, and the FastAPI back end with async MongoDB.
The back end runs on Cloud Run. Cloud Scheduler starts the daily scrapes, and Cloud Build deploys every push to main.
03 / How we use Python and FastAPI
Bookstore listings are noisy. Hiki’s Python back end turns them into a catalog that readers can filter and search.
Every day, Google Cloud Scheduler starts Hiki’s scrapers. They are built on aiohttp, with rate limiting and exponential backoff, and they collect product data from several bookstores: names, prices, discounts, stock status and images. Change detection flags new arrivals, price drops and stock changes, so readers never miss an update.
A raw product name from a bookstore is often noisy and inconsistent. Hiki’s enrichment pipeline orchestrates several APIs to find the manga behind it. AniList is the primary source for scores, genres, tags, synopsis and cover art. Shikimori translates titles for product names written in Cyrillic, and Jikan and Kitsu are fallbacks for the edge cases. The pipeline generates title variants, handles both Cyrillic and Latin characters, and resolves even badly mangled names to their canonical manga entries. Bookstores list no genres, so this step is what makes Hiki filterable: the 21 genres and 100+ tags that readers filter by come from AniList.
The API is FastAPI with async MongoDB through the Motor driver. MongoDB stores the catalog as flexible documents with compound indexes. The Next.js front end reads from this API and renders the catalog in every supported language.
The back end runs on Google Cloud Run, next to the front end, and Cloud Build deploys it on every push to main. Cloud Armor sits in front of it with geo-restrictions and an allowlist for crawlers.
04 / Other technologies
Hiki’s FastAPI back end feeds a Next.js front end written in TypeScript. See what else we built with them.
T.01 / Front end
Hiki’s multilingual catalog front end and the Scout web app, both on Next.js 15 with the App Router.
Built with it
T.03 / Language
The language under all four products, from the web front ends to the seo-mcp server.
Built with it
07 technologies
Seven technologies, each one running in a product we shipped.
Start / 30-minute call
No corporate theater. Just results.
Prefer email?
hello@larch.ee