This might be a bit weird but currently this is an ongoing issue.

why? because of Jailbreaking.

It's not clear with Lumo, but if you use Local Ais you will notice that, without telling you, they are assuming you are trying to jailbreak them when you copy paste what they thought + you telling them what to do or similar. Aka safety mechanism.

Second reason: Pressing the Copy button does not include the thinking. And on phone, when you try to select the whole sections, the whole screen get's selected, and i don't know why since it does not occurs with any other app.

Also, i understand that this will make the conversations shorter due to Conversation limits however, the idea is to trade quantity for quality.

example, Lumo is thinking regarding an issue, it correctly thinks and implement in next output. Then, (the next next output/s) Lumo seemingly forgot what it was thinking about because sure enough, it actually does not "know". And this means, it give once again generic, wrong, answers based on misunderstanding. And we need to trigger the thinking via try again or similar, so that Lumo can once again, do the same conclusion, it already did, but did not "call".

Btw, if you copy paste it's thinking, and for whatever reason, it assumes you want to jailbreak it, the Ai will not tell you but you will notice drastic drops in quality, copy pasted answers, changing subject, giving incorrect informations intentionally, and more. Why? due to how some "genius" people decided how LLM are allowed to lie.

Anyway, idea is that, at the cost of conversation volume, the whole "thinking" is "visible" to Lumo. As such, this (hopefully) allows chain of though instead of "resets"