1.39 Bragi, the friendly AI here to help

Dear Kanka users, please welcome our first foray into modern AI and ChatGPT, project Bragi! Keep in mind that this feature is an experimental crazy idea we had, and we’re keen to collect feedback on how useful (or not) it is.

Bragi, generating character prompts for you

Everyone loves new buttons, right? So we’ve added on in the text editor, specifically on characters and posts that are on characters. This is the new “hat” icon next to “@”.

Accessing bragi

Clicking this brings up an interface to communicate with Bragi, our ChatGPT powered character generator! Provided a short prompt between 10 and 150 characters, Bragi will generate three (and very rarely four) paragraphs that describe a character.

Bragi's UI

These paragraphs can contain information about how they look, their upbringing, their work, education, situation growing up, fears, goals, successes, and other keywords we’ve built into our AI.

If you like what you see, the “Use” button adds the generated text into the entry.

Who can access this feature?

Because this is our first experiment with AI technology, and each generation costs Kanka money, we’re currently limiting this feature to Wyverns and Elemental subscribers. Wyverns get 15 tokens that refill once a month, while Elementals get 30. They can access the prompt in any campaign they are a part of, regardless if the campaign is (super)boosted or not.

Tokens refill to their max of 15 or 30 on the user’s monthly billing cycle, for both monthly and yearly subscribers. Unused tokens don’t roll over to the following month.

If you want to learn more, we cover more details in our documentation!

General improvements

We’re adding more security to Kanka to avoid losing data during server updates. Until now, a warning that Kanka is currently down would only be shown when creating or editing entities and posts. From now on, those alerts will also display for the following elements:

  • Editing the entity’s entry
  • Quest elements
  • Timeline eras and elements
  • Map layers, groups, and markers
  • Inventory
  • Reminders
maintenance notification alert

The abilities, assets, reminders etc subpage list of an entity now orders alphabetically, rather than rather randomly.

entity subpage list

Public campaigns get a new filter for worldbuilding prompt winners.

Attribute names now parse mentions. This also means that mentions in a ranged attribute now create a dropdown with entity names in the live update view.

attribute name mentions

Two-factor authentication for all Owlbears

Due to the successful implementation of 2FA with our Wvyerns and Elementals, we are rolling this feature out to all our Owlbears. This is to increase the pool of users, and continue to sort out any new issues that we encounter. We are still planning on giving all free accounts access to 2FA in the coming months, once we are fully satisfied with the systems we have in place for managing edge cases and issues.

To enable 2FA for your account, head over to your account settings.

API changes

The campaigns/id/users endpoint now supports the lastSync param, and a new campaigns/id/users/id endpoint now exists to gather information about a specific user.

Bug fixes

The member visibility limitation was ignored on map markers if a user was logged in and not a member of the campaign.

The “matches this or its descendants” filter option works again.

The CMD+SHIFT+S keyboard shortcut to save and edit works again.

That’s it for this update. It’s smaller in features, but we’re keen to collect feedback on project Bragi and are excited to see what the community does with it! If you missed the previous update, you can read it here.

As always, thank you to our amazing subscribers who make these frequent updates possible. Come join us and our community on our Discord server!

Photo by Aideal Hwa on Unsplash

5 Replies to “1.39 Bragi, the friendly AI here to help”

  1. Hey guys! This could be really cool, but I’m concerned about whether the AI datasets are ethically sourced. If the text is pulled from properly licensed sources (i.e. public domain, consenting authors) I don’t have a problem, but otherwise I may consider cancelling my subscription as I don’t want my money going towards art theft. Writing is art too!

    1. Yeah, strongly agree here. I think this is a bad direction for the site. On top of the ethics issues re: to the sourcing of data, there are tons of additional problems relating to racism and bias in models.

      An implementation I could get onboard with: Pay someone to make a bunch of random tables for you, and integrate them into the interface as subscriber perks.

      1. Hi, thanks for sharing your thoughts with us. I just want to empasise that this is just a little bit of fun, we’re not fundamentally reworking the way Kanka works, or the direction that we’re taking. We hadn’t considered all the points that have been raised concerning this, so we have some thinking to do.

        As for the second part, about having (human generated) integrated content, we’re already working on this. I don’t want to go too much into detail here, because we have been working on this for over 6 months, and have faced several delays, but we have been pouring a lot of resources to bring huge amounts of human generated content to Kanka . We hope to be able to divulge more information in the coming months!

Leave a Reply

Your email address will not be published. Required fields are marked *