Easyling release notes January, 2025

Following our usual pattern, during this January, we focused on behind-the-scenes improvements. We had a lot of them too! The highlights include that we added support for newer OpenAI models and we added cookie support to the script for CloudFlare publishing. This allows for higher quality machine translations and helps with the end user experience on some sites. Thirdly, we improved the translator engine under the hood. Read the full article for details.

New OpenAI models

OpenAI models can be used as a translation engine in Easyling. With their great understanding of language(s), they are great both for pre-translation and suggestions in the Workbench. This month, we added support for the newer reasoning models o1-preview and o1-mini. These models can “think” for longer to ensure that the responses are more accurate. This can significantly boost translation accuracy. You can test these new models via the OpenAI GPT section of the dashboard and you can read more about these models in OpenAI’s documentation.

Cookies with CloudFlare

CloudFlare is one way to publish projects in subdirectories, like example.com/es/. We provide sample code snippets for the CloudFlare Workers. We updated these sample snippets to include cookie handling. In addition to that, we tweaked the formatting of the code itself to make it a bit more readable. Going forward, with this updated script cookies will be passed on between the browser and the source server.

If you already published your project with CloudFlare, this update doesn’t affect you and there isn’t anything you need to do right now. When publishing additional languages, however, we recommend that you use the updated script as displayed on the Dashboard.

JavaScript-based translator improvements

We added two main improvements to this engine this month.

First, we added an additional layer of cache to the translator script. This will reduce the loading time significantly, ensuring that the translations stay as fast as they can be, improving the quality of the experience on the translated site.

Secondly, we added an additional check for “foreign” translation systems. The system is now better at detecting them. In this case, our translator will stop translating and let the other system do its thing. The benefit of this is that it eliminates the possibility of picking up content translated by the other system as source content in Easyling.

Trace ID

This is a dream for engineers and our support team too. This month, we solved the problem of tracking (or tracing) requests in proxy mode from the browser, through Easyling and the original website’s server. Both in the browser and in server logs, you can look for an X-TranslationProxy-Trace-Id header. Use its value for correlating the logs with the requests. Additionally, if you let us know the value, our support team can provide further information by analysing our logs too.

Finally, there is an X-TranslationProxy-Custom-Trace-Id that you could add to your own requests. With this, you can specify a custom Trace ID that will be logged and forwarded to your server, so that you can trace the path the request takes on your end. Note that the longest value you can pass in is 128 characters long.

Miscellaneous

This month, we had a few smaller improvements too:

• We added support for the translation proxy in the recently-released variant 3 of the language selector. Now it works well regardless of configuration.
• The API for XTM changed so we updated our integration. This ensures continued compatibility.
• The API for DeepL changed so we updated our integration. This ensures continued compatibility.

AI disclaimer: This document was written by a human without involving LLMs.