There's models that has vision like Claude 3, GPT4 Turbo, Llava, FireLlava, etc on OpenRouter, but attachment selector can't select picture (greyed out). Please fix this.