Shift + Comand = Shift to GPT4 model? | Voters

Shift + Comand = Shift to GPT4 model?

in progress

Vec

I currently set my default model as GPT3.5 0120. While some questions, it can't solve. 
Could it be set like, when using commands, if pressing a key(shift i.e), it would answer the current context with the second model(GPT4)?
Press Shift+Enter to use GPT4

February 10, 2024

Daniel Nguyen

I've made some good progress. Will be available in the next version.

https://share.cleanshot.com/w8XJFTjL

Vec

Daniel Nguyen Excited for that!

Vec

Guess the instant command system prompt without context would also apply the second model. Previously, when shifting different models, it would need mannual click several times.

If has this function would be one click ask two AI models on tough questions. (and hope the conversation window would be logged. )

Vec

Could we let the optional key be a flexible modifier as well, some user may prefer use 'shift' or 'command' key .

Daniel Nguyen

Vec: Yes it works the same for Instant Command. Though currently, only command with "Open in a new window" behavior is logged.

I'm not sure we wanted to log the "replace selection" behavior. Feel not useful for me. What do you think?

Daniel Nguyen

So here is the solution I decided to pursue:

Add a new "Alternative Profile" for each command. You can set custom set of configuration for this Alternative Profile beyond just the model. But probably in the first version, only model can be tweaked.
Within the popover, holding "Option" would switch to the "alternative" profile in the command list. You can then either click or press Return to go with this alternative profile.

Does this sound good? Vec 123hxsmyxh@gmail.com

Vec

Daniel Nguyen Appreciate! Would be nice.

My personal opinion though, plz stick with what you think convenient.

In most situation, it goes with default profile.
option + enter : use another profile
I hope to bunk assign all the commands with a profile instead of click for each instant demand.

Vec

Two ideas for option.

when not holding option, the profile would switch to original default model.

or 2. when press option once, the instant command would switch to model profile2. and showed on the command list to let user know.

and when pressing option again, it goes back to default model.

Daniel Nguyen

marked this post as

in progress

John H

The feature described here is one example of a higher-level concept / feature request: easier/faster modification of prompt settings.

For example, a user may:

Begin a chat using the default GPT-3.5 Turbo.
Wish to ask the same question again a different model (e.g., GPT-4).
Wish to ask the same question again using a different temperature (e.g., Temperature)

This is currently a painful process in BoltAI. Competitors like MindMap make this easier by exposing chat settings in a sidebar (see screenshot). MindMap also displays a settings summary under the input field (see screenshot). I would like to see BoltAI merge these two concepts along with an ability to reuse previous prompts as a templates to make generating new prompts with custom settings easier and faster.

Display current prompt settings below input field
Prompt settings displayed below input field are clickable and display a popover with details and/or additional settings (similar to BoltAI's current "Settings" popover in the titlebar)
Tastefully display prompt settings summary under each chat message in the history so users can easily identify settings used to generate responses (model, temperature, context, assistants/prompts, etc). To reduce visual noise, this information can be grayed out, displayed on hover or toggle, listed in "..." menu, etc.
Allows users to easily "reuse" a previous prompt's text and/or settings. For example, action menu items such as "Reuse Prompt" (text and settings), "Reuse Prompt Text", and "Reuse Prompt Settings" can be used to add prompt text to the input field and/or modify prompt settings but
not
send a message. Users could then quickly modify the parts they needed and send the message when ready.

Hope this helpful. Keep up the great work with BoltAI!

Daniel Nguyen

John H Wow. Thanks for your in-depth & valuable feedback. I'm aware of this issue and have been working to improve this. Although I think it's not applicable for the AI Command feature.
It's a challenging UX problem as the selling point of BoltAI is a clean and easy-to-use chat client. The majority of BoltAI customers don't tweak these advanced configurations.
But I hear you. I agree that it's not easy to customize these configurations at the moment if you're a super user. I will find a solution that would work well for both non-savvy users and super users.
John, you should join our Discord server so we can discuss more on these UX change there. I also provide faster support there https://boltai.com/discord
Cheers

Daniel Nguyen

marked this post as

planned

Yes this is in progress. I was thinking about using Tab, and it would open a new sub-menu where you can choose a different model, or a different behavior.

Some of my users also wanted to change the behavior at runtime (open in a new window instead of replacing etc.)

123hxsmyxh@gmail.com

Daniel Nguyen glad to hear. I think if even if go with tab key, I don't want to select twice. Could it add both shift+command or set if need double layer after pressing tab. Ideally it could be similar agile one time experience

Vec

Wonder if this is possible 😌 Some questions for instant commands just need to shift to GPT4 to answer.

Seems we just need to add a shortcut and could improve this function.