Some aspects of UX + guidance around AI assistant

[lerone]

I am also very a) happy and b) impressed in general by the AI implementation.

there are several tangential threads, but nothing really congruent to my questions, and I didn´t want to hijack anything. so here some feedback:

1. scoping the assistant – more easily [kind of FR] : I would very much like to be the scoping of the AI in general (database – web – Wikipedia – …) to be controllable very much like the model – that is: from the right side pane itself. I think ‘hiding’ this with the preferences doesn´t really make sense as the scoping is as central (to me at least) as the choice of the model.

2. web-search & guidance: the implementation of a heterogenous field of LLMs and platform-logics etc. of course brings challenges, also to development as can be easily seen (see the discussion on OpenRouter AI). I understand that. but I think implementing all this especially in a way there web-search becomes tangential in the UI does warrant / call for some kind of guidelines for the average users as to how to achieve different scenarios given different models etc. that is a plea to somewhere in the documentation cover these scenarios (websearch on/off; different kinds of websearch; etc.) in the documentation, or some other more central guidance help

3. references from AI assitant answers to DB + again guidance: same goes for another important aspect: references to source documents in DB. – I was very(!) happy to see using o4 for example can produce links to the documents searched (a kind of index to DB content, as was mentioned earlie somewhere else in forum). this is really great, as these links seem persistant when e.g. one saves the results and so on (this reminded me of the great, underappreciated / -utilized feature of DT to allow timeline indexes to moves in the DB!). – again, as this is such a fundamental behavior re. to DT-character + -philosophy (intelligent document management), I would appreciate if there is something like a guide / overview as to how to achieve this with different models, differences, options, affordances etc.

Thx for considering!

[cgrunenberg]

@lerone

I would very much like to be the scoping of the AI in general (database – web – Wikipedia – …) to be controllable very much like the model – that is: from the right side pane itself. I think ‘hiding’ this with the preferences doesn´t really make sense as the scoping is as central (to me at least) as the choice of the model).

Assuming that the search options are enabled, then one possibility is actually to include the necessary instructions in the prompt, e.g. not to search at all or only to search on the web etc.

[lerone]

thx @cgrunenberg ! – … but, I think I might have been misunderstood in this one.
as that was actually the proposal: to make the ‘base setting’ more accessible in the UI and expose it better – just like you so nicely allow to switch models right from the assistant.
I think enabling/disabling websearch and Wikipedia in the right sidepane itself – currently solely possible from the more fundamental app settings – would make sense, as scope is as basic as choice of model.

[lerone]

anybody interested in this thread: @cgrunenberg hinted in a newer thread w/ same proposal that this (#1) is now under consideration.
– see here