model usage issues

[NeverBack2015]

After the initial installation, the embedding model will be downloaded automatically. I noticed that the official website provides the Jina model. Could you please let me know how to use the Jina model or other embedding models?
In the startup methods, models are divided into regular models and chat models. I understand that the "model" corresponds only to the completion function. The Q&A used in the chat box and the functions like "generate single-sided" and "code explanation" in the right-click menu all call the chat model, is that correct?
Why were two models distinguished at the beginning of the design? Currently, DeepSeek is very popular, and it has high scores in some metrics. However, when I used DeepSeek Coder v2 Lite as the chat model, it reported an error. The error log mentioned that it does not support k-shift. I am new to large models and lack sufficient knowledge, so I would like to understand what factors to consider when choosing other models as chat models

---

Please reply with a 👍 if you want this feature.