August 1, 2024 / 👛 AI clusterization, Text on a connector line, tons of other things

[mdubakov]

Today we have a lot of updates in various areas. Fibery is getting better every week.

AI clusterization: find repeating patterns in sources and create target entities (experimental)

Now you can run Clusterization for Sources in Highlights area. AI clusterization will analyze all data in Source databases and find repeating patterns. Then target entities will be created and Highlights automatically linked to these entities. For example, here we analyze all Discourse posts and create Features and Bugs from them.

Here are some use cases:

• Finding themes and problems in your interviews and usability tests.
• Process feedback and find Insights or Bugs.
• Fill your backlog from the start. By default you have empty backlog in Fibery, so you can run clusterization to find features in all feedback sources and have some backlog automatically.

Clusterization takes time (20-60 mins), so when results are ready, you will get a notification, and a special View will be created with the results. For example, these are the results for Fibery Community forum clusterization:

Try to run clusterization on some fresh data and see how it works for yourself.

Check AI clusterization user guide for more details.

Whiteboard: Text on a connector line

Now you can add text on connector lines, change its color, move it, etc. It simplifies diagrams creation. Check the video to see how it works.

Fibery AI improvements and changes

Today all accounts that used OpenAI key are moved to Fibery Basic AI plan. We no longer support OpenAI keys. If you want to fully utilize Fibery AI, please consider upgrading to Fibery AI Pro pack.

AI Assistant (that you will find in the left menu) can handle any workspace size now, before it did not work for workspaces with many databases. It was transferred to fine-tuned gpt-4o-mini model with the large context window.

Other changes:

• Formulas AI assistant switched to GPT-4. So you will have better results.
• Now you can disable most AI features in your Workspace with the new Fibery AI switcher in Settings → AI. When turned off, the following will be hidden: Semantic search, Rich text agent, AI Assistant, AI in automations, Highlights suggestions, Transcribe for files. Space configuration features like New Space and AI for formula will remain visible.
• Fibery AI limits for chatbots increased from 200 per user to unlimited.
• Default AI model in automations is gpt-4o-mini now.

Rich Text: Preserve context/anchor when converting links to mentions

Previously, when you tried to convert a Fibery link with an anchor to mention, an anchor was lost. Now, it is preserved, and you can use it in the following cases:

• Anchor to the header in documents and entities rich text field is preserved.
• Copy link to Whiteboard selection mention will contain a proper position.
• Mirrored View mention with a defined context will work.

Allow to set any database field via URL for a Form View

Sometimes it is needed to set hidden fields values via URL, but it worked only for fields included into a Form. Now you can enable Allow to set hidden fields via URL setting for a Form and set any database field via URL (well, except rich text field).

Airtable integration/import improvements

• Support for user information has been added in Airtable-synced fields. From now on, Collaborator (Single Select), Last Modified By (Single Select), and Multiple Collaborators (Multi-Select) can be synced using Airtble user’s email addresses.
• Airtable integration supports Relation fields between Airtable databases.
• Support the Duration (as a number in ms) and Sync source fields (as plain text) in Airtable integration.

Improvements

• Previously, naming a Field or automation was mandatory, and the “Add” action was blocked if name left blank. Now, all Fields, automations, and AI commands will have a default name unless a custom one is set by the user. As a result — some clicks saved.

• Left Menu: Renaming a View from Navigation Menu became easier: simply double-click on the name to rename it. Simple as that.

• Left Menu: We’ve polished Navigation menu resizing and now it automatically collapses if it’s width is less than symbols you can read on it.

• We’ve updated the confirmation modals for adding or editing Fields. If you discard changes, the messaging is now better adapted to the context of the Field and its modifications.

• Comments: now, even if you remove the text that was used as an anchor for an inline comment in a rich text, the comment will still be available. You can find these comments in the comment sidebar under the “unlinked” tab. This ensures no important feedback gets lost.

• Whiteboard: Show a limited number (5+) of items in collections on cards. For large collections it was awful to see all items, now it is much better.

20+ Fixed Bugs

Documents and Rich Text fields

• Ctrl+Z in documents and rich text fields does not restore comments
• Firefox: inline comment popup gets closed if I try to select a user via mouse in mention popup or edit comment
• Inline comments disappear when converting paragraph to a checklist
• Inline comments: send button disappears if a long url is pasted
• Safari: scrollbars in the inline comments look bad
• Dark theme: text with inline comment is unreadable when in focus

Whiteboard:

• Text formatting in a sticky note reverts to default on refresh
• In Multi-entity insert selected DBs don’t get unselected if you select another space in dropdown
• Objects get added to whiteboard from the second click after you cancel image adding
• Relation lines don’t display on whiteboard until you select the card if cards were added via multi-tool and vertical layout was selected
• Undo toast doesn’t appear if delete relation line
• Missing line after converting shape to entity
• 10 sec delay to display ‘entity not found’ when a user selects an already deleted entity
• Maximum call stack size exceeded in console when clearing Section (Frame) Header
• Incorrect 0 opacity for Shapes
• Unable to insert View as a Card
• Floating menu has excessive menu separators
• Incorrect 0 opacity for text
• Text Box size doesn’t follow Font and Font Size change when it is just selected and not in edit mode

Other fixes:

• Dark theme: no row selection in the search popup
• No icon placeholder is broken on views
• Support hide when empty for avatars collection unit

[Mircea_Braescu]

@mdubakov

Now you can add text to connector lines

That’s great news

[grex22]

@mdubakov

Dark theme: text with inline comment is unreadable when in focus

Literally just YESTERDAY discovered this bug (which was super annoying for our particular use case) and less than 24 hours later it’s fixed. You guys are INCREDIBLE.

I know I’m drawing attention to perhaps the least impressive item in this week’s announcement, but still it was a huge fix for me and my crew

I can’t wait to play around with clusterization too… I can’t even wrap my brain around what that might mean for my spaces (yet)!

[Yuri_BC]

@mdubakov

AI clusterization: find repeating patterns in sources and create target entities (experimental)

I think it will result in more focused results, and less resource intensive, if there is a step in the clustering action where the user is asked to select the source and target entities. Because, the hight database is a generic tool and I have the max (6) source and target datbases connected. I dont want AI to try to cluster that.

The drawback of additional highlight separate highlights is that they cannot use the same source databases. This is pretty problematic. I can explain why but I think people already encountered the same tension.

But back to the topic: selecting clustering databases would help optimize the process.

EDIT: SORRY, the selection of clustering database is already in place, thank you.

[ccollins]

@mdubakov

• Anchor to the header in documents and entities rich text field is preserved.
• Copy link to Whiteboard selection mention will contain a proper position.

Amazing updates! Very excited for this in particular.

One note, if you have two different anchor mentions to the same document, you can click on both links repeatedly and the right panel will jump up/down to the respective anchor point. But this does not work with two different position mentions to the same whiteboard. Clicking the second whiteboard mention just stays on the first position.

[mdubakov]

@Yuri_BC

selecting clustering databases would help optimize the process.

Maybe I miss something, but this is exactly how it works now…

[YvetteLans]

Awesome improvements

@mdubakov

Allow to set any database field via URL for a Form View

Sometimes it is needed to set hidden fields values via URL , but it worked only for fields included into a Form. Now you can enable Allow to set hidden fields via URL setting for a Form and set any database field via URL (well, except rich text field).

This is also great! Although it can be a bit hard for ‘normal’ users.

I want to create a simple interface where

• User fills in the basic URL of the form view
• Then chooses category + product
• And if applicable they can even extend the prefilled form with a specific URL per client or client project.

Currently I need to fetch UUID via script.

Would it be possible to provide UUID via formulas, like public ID?

Then the user can create 1 form that can be used for multiple use cases, with prefilled entities. They just share the ‘adjusted URL’ then

[Chr1sG]

@YvetteLans

Currently I need to fetch UUID via script.

You can get the UUID by alt clicking on the piblic id in entity view.

[njyo]

@mdubakov

Now you can add text on connector lines

TEXT ON CONNECTOR LINES!!!

/doing the Elaine dance

Thank you! Thank you! Thank you!

… had to test it!

[mdubakov]

@Chr1sG

You can get the UUID by alt clicking on the piblic id in entity view.

[Illusory]

Is it possible to set form fields from a URL if they are a relation?

e.g. Issue (Form) allows use to create an issue. On this form, we can select related “Bug” and “User” database entities. When we click the button from an external tool, we want to auto-set “User” and “Bug”.

[mdubakov]

Yes, but you need to provide UUID

https://the.fibery.io/@public/User_Guide/Guide/Set-Initial-Field-Values-via-URL-Parameter-on-Form-View-246

to-one references
you need to provide a full UUID here, can be copied using Alt+click on public id badge on entity view
state=ecebeec0-e34f-11ed-a326-e9cc78af8dd6

[YvetteLans]

@Chr1sG

You can get the UUID by alt clicking on the piblic id in entity view.

I know

But I think I didn’t explained the use case well enough.

• The goal is that the end user can share a URL with prefilled category, product, contact and client project
• If they need to fill in 4 UUID’s to achieve that, it’s a pain in the ass
• If you look at my solution, they can simply select the items they want to prefill
• Because then I can populate the URL (and fill in all those UUID’s in that URL)
• So that a user only needs to share the ‘end URL’ with the person that needs to fill in the form

So in my solution when my client wants to create 20 specific form URL’s for 20 customers, they simply need to add an entity in the table and select the right customer → the URL will be generated via formula.

And you can even automatically email that specific URL to the applicable contact.

In your solution my client needs to open each and every contact, copy UUID, paste it somewhere in a field, etc. etc.

And then to imagine if they want to prefill 4 fields (like form category, product, contact and client project)

[njyo]

I did notice this that if you delete a drawing and then undo, the text disappears
Hope that will get fixed in the next weeks.

[Yuri_BC]

@mdubakov

Maybe I miss something, but this is exactly how it works now…

Yes, I did not check it well enough. Selection of clustering databases is indeed in place.
Sorry!

[Yuri_BC]

AI Clusterization:

Will it be possible in the future to have more customizable clustering settings or rules?

For example, I have a generic database ‘Page’ which has a field ‘Page Type’, which purpose is to optimize the system by reducing the number of databases.
However, by choosing this approach, I currently miss out on narrowing down the clusterization to the field level of a source database and/or target database.

[mdubakov]

Short AI Clusterization demo

Michael Dubakov

[nvee]

I have this same use case. What I did was create a new field called UUID (copied), then I run automation with the script action to copy over the uuid to the new field:

const fibery = context.getService('fibery');

for (const entity of args.currentEntities) {
    await fibery.updateEntity(entity.type, entity.id, {'UUID (copied)': entity.id});
}

This new field can be used in your formulas. A bit of a workaround, but it works.

[bear]

Good update!