Circassian AI API
RUEN

Translation and speech
for Circassian languages

Kabardian and Adyghe in your product: translation both ways with Russian, English and Turkish, a dictionary, language detection and speech in a male or female voice. Plain REST, one key, no SDK required.

Documentation OpenAPI schema Service status

What it does

Translation

ru, en, tr ↔ kbd (Kabardian), ady (Adyghe). The source language may be omitted.

In batches

Up to 50 texts per request — faster and cheaper than one by one. Results come back in the order you sent them.

Dictionary and detection

Dictionary entries for a single word by dialect, and language detection without translating.

Speech synthesis

Male and female voices — the same ones used by the bot and the website. Voice cloning from a sample and custom voice descriptions.

Long texts

Long text turns into a job: the connection never breaks and the result is fetched by link.

Errors you can act on

One shape for every operation and a machine-readable type: you can tell what to fix from what to simply retry.

What a request looks like

curl https://api.circassian.ai/v1/translate \
  -H "Authorization: Bearer $CIRCASSIAN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Hello, how are you?", "source_lang": "en", "target_lang": "kbd"}'

{"text": "ФӀэхъус, дауэ ущыт?", "source_lang": "en", "target_lang": "kbd"}

Examples in Python, JavaScript, PHP and Go are in the documentation.

Getting started

  1. Ask for a key. There is no signup: keys are issued by hand. Write to the service owner, describe what you are building and your expected volume.
  2. Check it. A single call to /v1/me shows your limits and costs you nothing.
  3. Start building. Everything else is in the documentation: operations, limits, errors and long-text handling.

About your data

Every call records its time, key, operation and volume — that is what usage accounting runs on. Request and response content is kept for 30 days to investigate quality complaints, and can be switched off per key on request. Reference audio for voice cloning is never stored. Details are in the “Data” section of the documentation.