DevBlog - Huge Refactor - 4/10/2025

Boy, this took awhile, and we're still just getting started.

Howdy folks. I come to you bringing good tidings and news from the dark, damp cave we store myself and the rest of the dev team inside. For about a month, the team and I have been cracking away at the largest, most ambitious backend and functionality shift we’ve embarked on since the migration from a static Firebase application to a Next.js application with a proper backend. That won’t mean much to most of you, but trust me, there has been a very good reason why we as a team have been so quiet (we’ve been busy).

What does that mean?

This means that in the future, it will be a lot easier and quicker for us to implement functionality changes. It also means that the chatting experience should be a lot more stable in terms of persistence and ease of use. We’ve gone ahead and fully redone the entire chat page, the entire chat log system, and our entire connection stack. This means you will need to use a converter (located at /chat/convert) to convert your old chat logs into fancy new ones! We’ve gone the route of allowing you to pick which ones you want to convert, but you can also do it in bulk if you prefer. It also means that now we have Featherless account linking and a new simple mode that allows you to pick a free queue model or a Featherless model (this requires an API key or an account link, either works), and it will create a connection with optimal instruct and parameters from the API. We, of course, allow users to override the defaults (now done per model/per connection for free queue) as well, and custom connections that don’t use any of our fancy systems are also available.

Featherless Account Linking

First up, let’s talk about the main event; Featherless AI account linking. Gone are the days of entering API keys on every device you own, and today we bring you the ability to link your Featherless Account & WyvernChat account. This lets you utilize all of the models on your plan in a streamlined view inside the chat page.

These models are automatically assigned the defaults provided by the Featherless API; this includes instruct templates and basic settings parameters.

Cloud AI Connections

You can link API keys to the backend; they will be encrypted and can be used across devices.

Cloud AI Connections? Why don’t I get to use them in Story or Scenario mode?

We haven’t got there yet. It’s coming, but we’ve elected to go for an iterative push where we overhaul one section at a time. You will still be using the connections saved per device for Story & Scenario mode.

Hang on. Where is the model list thing for providers that aren’t Featherless? It’s a text box?

Yeah, I know. I’m working on it, and we will get a better version similar to the last one available. I just did this one for now, so you guys can still use other providers while I finish the better implementation.

Uh, is the chat page missing older features?

Yep. They aren’t gone forever, though; they’re just being redone to be better. Impersonation, Chat Summaries, and… actually, I think that’s it for what’s missing are actively being worked on. Chat Summaries, in particular, are now going to be automatic shortly and will come with history. That’s right, if you get a bad summary, you can just choose to go back to the other one, or the one before that, or any in the same chat log!

Chat Page

Simple & Advanced modes are the best of both worlds. Allowing for those of us who want a simplistic easy experience to just pick models without worrying about bad samplers or confusing templates.

Using the same free queue models or with any model on featherless.

Or with completely custom connections inside of advanced mode.

Now, with editable instruct templates, connection-tied parameters, and many upgrades to the Chat UI. You can now edit your own characters inside the chat UI.

And with a much better command menu and many more commands in development.

With a brand new sidebar.

And the ability to select personas easily and in an obvious way.

With a new debug menu to view prompts directly in app, no matter the device.

And a new delete mode.

And the ability to reset to a specific point in a chat log.

Brand New Characters Page

AND NOW MY GREATEST CREATION-

Behold, the Character Random-inator!

Finally, there is an easy way to decide what character to use—without deciding yourself! This is perfect for folks like me who have trouble with decision paralysis.

Coupon Code For FeatherlessAI

To celebrate the launch of the new account link and better integration, we’ve decided to give everyone a fancy new coupon code!

WYVERN-REVAMP

This will be for 20% off FeatherlessAI and expires on April 15, 2025, at 12:00 PM PDT! Use your newfound model list wisely, and have fun!

Click Here to go straight to the checkout page with the coupon applied!

Wrapping Up

Well, I know some of you are going to be disappointed with this update, given some of the missing previous features. However, I think that we’ll be able to get the new, better versions out fast enough to satisfy folks. You should receive summaries, impersonation, and the token budget viewer shortly, and all of them will be in better form than before.

Here is a list of everything the team is working on for release within the next week or two:

• User Impersonation

• Advanced Automatic Chat Summaries

• Advanced Token Budget Tab

• Branch viewer (with visualized branching structure)

• Custom Provider model selector

• Group Chats

There will be a few other fixes and surprises I may have missed in my discussion here, and I’m sure you’ll find a few of them. I look forward to returning to a regular patch schedule, with updates rolling out whenever we get them fixed, just like before. This release has given me and the dev team a lot of anxiety, and I hope you all like the new changes as much as we do. Thanks for reading!