The company also released Chitrarth 1, a vision-language model built on top of Krutrim 1, capable of understanding images and documents