Launching today
SIYAK
Context-aware Arabic translation API
2 followers
Context-aware Arabic translation API
2 followers
Context-aware localisation product Drop-in SDKs for Flutter, React, Next.js and Node.js, and one JSON endpoint for any other backend.




Free
Launch Team / Built With

Atlassian LoomHit record, hand off to AI
Promoted

Every app with Arabic and English hits the same wall: one word is not one meaning.
I felt it in every multilingual project. On the backend: resource files, or a database table of keys with one value per language — and if the end user enters the data themselves (a product listing, say), you make them type it twice, in Arabic and in English. Tiring for them, error-prone for you. On the frontend: JSON files or a translation package, and the output read like literal machine translation.
The worst part: the word Report on a complaint button and Report above a downloads list are two different Arabic words. إبلاغ, or التقرير. No engine knows which screen you are standing on, so it picks one and applies it everywhere — and the same string looks different in the next build.
I looked for a fix. Most tools are a monthly subscription in dollars, and Arabic isn't treated as a locale at all: Google Cloud Translation ships ar and ar-SA, Azure ships ar.
So we built siyak (سياق).
It is not a translator — it is a localisation layer for frontend, backend and mobile:
• You send the string with the screen it lives on, and the context decides the word.
• A translated string is cached and never re-translated: the same request keeps returning the same answer in every build.
• Pin your own terms in the glossary. If the model ignores one, the row is flagged and shown to you instead of shipped quietly.
• An Inbox to approve, reject or edit — what a human approves is what gets served.
• Export your strings as JSON, CSV or ARB for Flutter.
• Up to 20 new strings in a single batch call when you fill a project from scratch, or the same strings into up to eight languages at once.
• siyak-js on npm, siyak_flutter on pub.dev, and a plain API for any backend.
Any BCP-47 tag works — en, tr, fr, zh-Hans, ar-EG — up to eight languages in one call. But here is the honest split: Arabic is the language we review and stand behind. Every other language comes back from the model and waits in your Inbox for you. And ar-EG still returns Modern Standard Arabic with the right term per screen — the Egyptian colloquial voice is not built yet.
Try it now at siyak.cloud
#i18n #localization #Flutter #Nextjs #Arabic