You got me well. I am a user of translators but through their APIs, not interfaces, and something currently missing in the space are models capable of grasping the context and deliver very good translations. DeepL is the best on this but in my experience it fails 50% of the time. Google Translate only gets it right 20% of the time. My agent, like 90% of the time. That’s huge. And it actually costs a fraction of Google’s and DeepL’s prices.

That’s the 1-2 features you talk about for this project. I’d also like to be able to add (1) glosaries, (2) formal, informal and custom tone, and (3) infinite context windows. But seems less important than the rest.