DLX

DLX

Why was this project renamed?

In July 2026, we received a trademark notice forwarded by GitHub Trust & Safety, submitted on behalf of DeepL SE. The notice stated that this project’s former name, “DeepLX”, contained the registered trademark “DeepL” and might cause confusion about whether the project is authorized or endorsed by DeepL SE.

It never was, and it never claimed to be. To resolve the matter and remove any possible confusion, we renamed the repository to DLX and removed DeepL branding from the project. To state it plainly one more time: this project is not an official DeepL project and has no relationship with DeepL SE whatsoever.

DLX is a self-hosted translation API server written in Go. It exposes a simple HTTP API on port 1188.

Usage

Docker

Docker Hub and GHCR image names remain deeplx (only the GitHub repository was renamed to DLX):

docker run -d -p 1188:1188 ghcr.io/owo-network/deeplx:latest
# or: docker run -d -p 1188:1188 missuo/deeplx:latest

Or use the provided compose.yaml:

docker compose up -d

Binary

Download the binary for your platform from Releases and run it (artifact names remain deeplx_*):

./deeplx

Translate

text accepts either a single string or an array of strings. A string is answered with a string in data; an array is answered with a position-aligned array of the same length, and the whole batch travels in a single upstream request.

# Single text
curl -X POST http://localhost:1188/translate \
  -H "Content-Type: application/json" \
  -d '{"text": "Hello, world!", "source_lang": "EN", "target_lang": "ZH"}'

{"alternatives":null,"code":200,"data":"你好，世界！","id":123,"method":"Free","source_lang":"EN","target_lang":"ZH"}

# Batch: one entry per text, answers come back in the same order
curl -X POST http://localhost:1188/translate \
  -H "Content-Type: application/json" \
  -d '{"text": ["Hello, world!", "Good morning"], "source_lang": "EN", "target_lang": "ZH"}'

{"alternatives":null,"code":200,"data":["你好，世界！","早上好"],"id":123,"method":"Free","source_lang":"EN","target_lang":"ZH"}

Batch limits and failure modes:

• The anonymous limit is 1500 characters for the whole request, counted in UTF-16 code units — what JavaScript’s String.prototype.length reports, so a CJK character costs 1 and an astral one (emoji, historic scripts) costs 2 — across every text in the array, not 1500 per text. A batch over that total is rejected with 413 and text exceeds maximum length: N characters (anonymous oneshot limit is 1500), so a large batch has to be chunked by total length rather than by segment count. There is no separate cap on the number of texts.
• A blank text — empty, or nothing but whitespace — is answered with itself and never sent, so a blank line in a segmented document keeps its place instead of failing the request. "text": [] is answered with {"data": []}; a body with no text field at all, or "text": null, is rejected with 400 {"code":400,"message":"Invalid request payload"}.
• A batch is all-or-nothing: if DeepL returns a different number of translations or leaves one of them empty, the request fails with 503 {"code":503,"message":"Translation failed"} instead of answering with placeholders or a shorter, misaligned array.

Discussion Group

Telegram Group

Acknowledgements

Contributors

Activity

License

MIT