It would be great to have the temp and top-p setting for ChatGPT - ideally as a per-column setting.
Accessing particular GPT assistants would also be super useful
Are there any plans to implement these settings? That would enable much more possibilities for people to interact with LLMs and might be a huge selling point for Baserow in terms of a broader spectrum of applications for AI. This would also make Baserow much more competitive vs such tools as GPTforWork (or Airtable, for that matter).
(especially given that Llama 3.1 is not legally available in the EU… )
Hey @dev-rd, these are all great suggestions and we have some plans already for implementing many already. I’ll add your comments to the GitLab issues myself as a reference!
Thanks! great news! @cwinhall
It would be great if you could link the gitlab issue ( I may have missed the relevant one in my search)
I’d love to know the ‘weight’ of the issue (as a roundabout way of asking when do you plan to implement it
Adding these advanced AI settings may really be an attractive selling point vs Airtable.
Right now, any larger project requiring AI use is simply not feasible financially with Airtable, unless one uses 3rd party integrators such as N8N/Make, which in turn add unnecessary complexity. Baserow having that built in would actually also remove a huge pain point for the “no code” user. Yes - diving into Make or even Zapier - is a pain for the average ‘no code’ user
any news on the AI features front? Any gitlab issue I can follow?
We’ve done some more tests for AI workflows.
Having the option of setting at least the “temperature” for the selected GPT models would significantly improve our workflow and eliminate a lot of pain points for us.
(you could name it randomness or creativity parameter for clarity for non-technical users, limit input to a number between 0 - 1, could even be in 0.1 increments, so that people do not go crazy with it)
We would be able to keep most of our workflow inside Baserow that way, without being forced to jump to other tools (which is problematic for large datasets - juggling these in and out of Baserow is far from convenient at the moment)
That, and fixing the bug with lost field references in the AI prompt upon database duplication/backup/snapshot would really make my day
With the new webhooks feature for manual triggers we will be able to route the more advanced OpenAI queries via Make/N8N, but the temperature option would REALLY help a lot as a Baserow native feature
btw - I’ve seen the other issue with batch generation for the AI field has already been fixed/sent as a merge request! Thanks for a super fast reaction on that! Really appreciate it!
I really like the current roadmap very useful features and fixes!
If you’ll be updating it further - I’d love to see the AI temp setting implemented in the coming weeks but I see that you have a lot of high priority gitlab issues already in the pipeline!
By the way - 1.28 - probably one of my favourite updates recently
