Inconsistent Language and Inaccessible Source References in Multi-turn Conversations

1. Language Inconsistency:

During multi-turn conversations, I've noticed that the AI sometimes responds in a language different from the one the user is using. This inconsistency disrupts the flow of the conversation and can lead to confusion. It would be greatly beneficial if the AI could maintain language consistency throughout the entire dialogue, matching the user's chosen language.

2. Inaccessible Source References:

I've observed that when the AI cites information sources, it only provides numerical references (e.g., [1], [2]) without offering a quick way to access these sources. This makes it challenging for users to verify or further explore the referenced information. It would be extremely helpful if you could implement a feature that allows users to click on these reference numbers, providing direct links to the corresponding sources.