Here you can check were we are heading with Fibery AI for creators.
Our goal is to let people build solutions with minimal Fibery knowledge and just ask what a user wants in terms of domain knowledge. Here I tried to build Calories Tracker and evolve it from a very basic solution to a good enough solution only communicating with AI Agent and almost never touch UI (well, only for some views).
It is interesting to hear your feedback and whether you want to play with it.
Main cases:
Space creation and database creation
Automatic sample data filling
Formulas creation
Views creation
Domain evolution
Errors correction
This is FAR from final and with some flaws. It is close to some beta version that we can give to some deeply interested people.
dummy data for demos and for testing. I tried using the MCP, but it took way too long.
Renaming fields / views / folders (can it do folders?) in bulk. I was just asked to use title case instead of sentence case. Changing everything by hand will be a pain…
Data migration. When changing processes, this could be amazing! We can make new entities from field, and move data in bulk between databases very nicely.
Getting input on where existing workspaces could be optimised. A little “Builder Buddy” to bounce ideas with (it gets lonely).
I think its biggest limitations:
Automations: Due to the “DIY” nature of fibery, you’re gonna need to use automations to have a working workspace. Think: document templates, project templates, recurring tasks, and more.
Lack of understanding of fibery quirks. Like “Unify field names” and types, a power user would know what to ask, but a new user would not know this is an option. I would love it if it would know that this is something to do, even explain to user why it did it in this way.
A big one: the lack of branching. I would NOT feel comfortable to tell the ai to do a migration, just for it to fail and delete or overwrite data that wasnt migrated… I recall a topic on this but can’t find it now… Some way to make a “snapshot” of a Workspace, Space, or Database, and be able to recover if the AI decides to mess it up… Alternitavely we could duplicate the space, then make changes. The issue is that this could mess up historical data reports, comments, and the the “Created on” and “Created by” fields.
Ideas for future:
Improvement ideas for admins / creators: “This field is present in the relation, make it a lookup?”, “This field is never used, delete?”, “These fields are populated half the time, and these other fields are populated other time, shall we split into two different databases?”
I am actively developing AI agents and I think my beef with the Fibery Smart Agent manipulating workspace space data, is dat I have seen that at some point any (semi) autonomous AI will have unexpected impact on current space and database architectures, and I have no local data to back them up - I think the fibery team is in denial of that priority.
Again points to potential data loss, lack of user data ownership, local backups, data safety, let alone privacy. I will post something soon about this in my eyes serious issue of missing data structures export ability.
Yeah, automations will follow. From my experience new users rarely use them in the first several days, so for us it is less important part so far.
I think it is not Fibery knowledge, you could type this command in different ways, and agent indeed can advice already what to do based on your questions. It is interesting to see limitations for sure, we did not test it thoroughly
Now you can do it via Experimental Lab → AI Smart Agent: Workspace & Data Editing (Early Preview). Enable it and in Smart Agent click on Build mode. Note that it is early preview and be careful with actions
