I am currently using OpenRouter with a free model as they provide a compatible OpenAI API. Problem is, as they are free models they often get the json format wrong, so the output has to be discarded.
But by looking at the logs, I see that after three tries the inference stops, and there are no more pending jobs of any type in the admin panel. My suggestion is then to add a rate-limiter setting in order to use such free options without DDOS'ing any site and to keep trying for unresolved tag.
In addition, it would be cool if in the admin section we could modify the base prompt and reset it to base value, as different models react very differently to the same templates and prompts
I am currently using OpenRouter with a free model as they provide a compatible OpenAI API. Problem is, as they are free models they often get the json format wrong, so the output has to be discarded.
But by looking at the logs, I see that after three tries the inference stops, and there are no more pending jobs of any type in the admin panel. My suggestion is then to add a rate-limiter setting in order to use such free options without DDOS'ing any site and to keep trying for unresolved tag.
In addition, it would be cool if in the admin section we could modify the base prompt and reset it to base value, as different models react very differently to the same templates and prompts