DeepL API integration

[MadayLyncoln]

How do I use the https://groq.com/ API? The free version. Is there a way to use Deepl?

[djethino]

you can use any LLM provider that use openai/v1 API.

for groq the url is : https://api.groq.com/openai/
(do not add /v1, it's automatic)
and you have to put your api key from https://console.groq.com/keys

for deepl, it's not a llm, it's a translator and for the moment I've not added the support for it, may be in a future version, I keep it in mind for the next version but they have their own API.

[ngoctienvt]

@djethino oh I tried groq and it got this error.

Warning:UnityGameTranslator] [AI] Translation error: One or more errors occurred. (No connection could be made because the target machine actively refused it. (localhost:5000))
[Warning:UnityGameTranslator] [AI] HTTP 400 BadRequest: {"error":{"message":"property 'think' is unsupported","type":"invalid_request_error"}}

[Warning:UnityGameTranslator] [AI] HTTP 400 BadRequest: {"error":{"message":"property 'think' is unsupported","type":"invalid_request_error"}}

[Warning:UnityGameTranslator] [AI] HTTP 400 BadRequest: {"error":{"message":"property 'think' is unsupported","type":"invalid_request_error"}}

[Warning:UnityGameTranslator] [AI] HTTP 400 BadRequest: {"error":{"message":"property 'think' is unsupported","type":"invalid_request_error"}}

[Warning:UnityGameTranslator] [AI] HTTP 400 BadRequest: {"error":{"message":"property 'think' is unsupported","type":"invalid_request_error"}}

[Warning:UnityGameTranslator] [AI] HTTP 400 BadRequest: {"error":{"message":"property 'think' is unsupported","type":"invalid_request_error"}}

[Warning:UnityGameTranslator] [AI] HTTP 400 BadRequest: {"error":{"message":"property 'think' is unsupported","type":"invalid_request_error"}}

[Warning:UnityGameTranslator] [AI] HTTP 400 BadRequest: {"error":{"message":"property 'think' is unsupported","type":"invalid_request_error"}}

[MadayLyncoln]

I was able to use Groq in the free version—it’s very fast—but after a minute or two, it stopped processing my requests. I’m not sure if this is a limitation of the free account or something else; I tested it with several LLM models and got the same result. Locally, with an RTX 5080 and Ollama running TranslateGemma:4b, it ran much better, but the problem is that it’s slow compared to the online APIs.

[djethino]

@ngoctienvt on the new version I do a better check, normally it will be good.

@MadayLyncoln I added Deepl and Google trad API.

I was able to use Groq in the free version—it’s very fast—but after a minute or two, it stopped processing my requests. I’m not sure if this is a limitation of the free account or something else; I tested it with several LLM models and got the same result. Locally, with an RTX 5080 and Ollama running TranslateGemma:4b, it ran much better, but the problem is that it’s slow compared to the online APIs.

for groq, the limits are described here for each models in free use : https://console.groq.com/docs/rate-limits
for ollama with cloud models, it's not very clear https://ollama.com/pricing
for openrouter : https://openrouter.ai/docs/api/reference/limits

for google & deepl translator api, the free tier is 500 000 chars.

@ngoctienvt & @MadayLyncoln now I check if there is a limit rate message from the server and do a retry, online models and traductor have a limit per minutes in general. you can config delay in the options. I putted 3sec by default if there is a 429 error and let the translation in pending for the retry.

new release : v0.9.57