Danweel
My feedback
5 results found
-
347 votes
An error occurred while saving the comment Danweel
supported this idea
·
-
312 votesDanweel
supported this idea
·
-
304 votes
An error occurred while saving the comment Danweel
commented
Would love some minor interactions with the cat, and optional cat personality written up. Especially since something like that wouldn't deeply cut into any serious development time and improves 'likability' that I think Lumo has other agents beat on.
I really appreciate the avoidance of a strange human staring at me, Lumo's cartoon representation and art is far more comfortable than the uncanny or frankly unpleasant fake "human" presence other AI companions and chats force (often using generic-looking AI generated art that isn't very warm or interesting). Lumo's 'presence' is great - the fact that people want to interact with it (and not in a weird way) seems to indicate a pretty positive impression!
Danweel
supported this idea
·
-
372 votes
An error occurred while saving the comment Danweel
commented
The RAG currently focuses on user files for the AI to look at - that's great, and I find that crucially useful, but it would be nice/interesting/engaging to have a RAG for the "personality" and "memory" of the Lumo agent, separate from the files it looks at to review. There would have to be a very clear delineation between injections for behavior and memories and files it's objectively supposed to discuss with the user. This could muddy responses unless it's able to clearly divide out the two intents. Might take some doing, but I think users would appreciate it, especially coming from Local models where that sort of control is fairly standard (but difficult to reliably manage).
Danweel
supported this idea
·
-
7 votes
An error occurred while saving the comment Danweel
commented
Agree, this is Important-to-Critical for organisation and accounting for user error and the changing of minds about topics or to branch out a thought into it's own concept.
Danweel
supported this idea
·
Because one has to define the agent for each Project, I came to think of the Project as being my instructions for that instance of Lumo.
I then had the instinct to create different folders in Drive for the different projects of mine I wanted that instance of Lumo to use.
The workflow seemed to be create Lumo instructions > work with _that_ Lumo on certain projects _types_.
A good example is a Lumo for learning code and a Lumo for written communications and grammar - editing is very different than teaching in terms of tone, personality, brevity and the types of asides it might offer.
I find myself making user projects and then having to copy/paste my Lumo into the different Projects that share similar Lumo needs. Not a huge problem, but feels a bit counter intuitive from a user process perspective.