Incorrect prompt in the target text field, in Hindi to English translation questions
In the Hindi to English translation questions, I'm seeing that the prompt text in the target text field says "अंग्रजी (केवल देवनागरी लिपि) में लिखें". This is obviously wrong, as the English translation is not supposed to be written in the Devnagri script, but in the Roman script. This needs to be fixed. I would have uploaded a screen-shot, but doesn't look like I can do that in these comments.
(The English to Hindi translation question prompt text is correct however, as it says "हिंदी (केवल देवनागरी लिपि) में लिखें".)
We are trying to see how we can convey these two important message to the users:
You need to type Hindi only in Devanagari.
The list of resources for Devanagari input.
Today, you should be seeing an updated message "अंग्रेज़ी में लिखें। (हिंदी लिखते समय केवल देवनागरी लिपि में ही लिखें, एंड्राइड पर गूगल प्ले से". So our message is getting truncated. We will try to squeeze in the important text in the available space and you will see a revised string tomorrow. These strings get updated once a day.
Another thing: Panini keyboard application seems to be inserting a line along with your input. Please see how you can avoid it, because along with the English text, your answer will not be graded as correct.
Here is a sample answer we received for "Sorry, not me":
"क्षमा करें, मैं नहीं। Typed with Panini Keypad"
It is all correct except for the inserted text "Typed with Panini Keypad".
Only if that condition is met, should the submitted text be sent to the server for sentence validation. (I'm assuming that the sentence validation is happening on the server-side, from the behavior that I've seen).
When Duolingo detects an incorrect language used, it does provide a warming (client side validation). We have requested exactly what you suggested - to not accept the a Hindi answer with roman characters - but as you can imagine, the dev team has a lot on the plate. These stings that you see are a "stop gap" arrangement.