#FREE NO API KEY NEEDED iOS Chat/Q&A/dev app for 3.5 and 4 8k-32k

263 messages · Page 1 of 1 (latest)

zealous parrot
#

Basically gives you the functionality of playground for GPT 3.5 and 4 with many new features coming soon. You can chat with some stock prompts with 3.5 for free, no key needed. You can save as many prompts as you like in dev mode with you own key, with their own individual settings, and then load them in from the chat screen. Moderation will be coming soon, currently not integrated. This will allow you to test different levels of OpenAI moderation API. Great for prompt setting testing for your projects while on the go.

Updated update plans:
-iPad version
-logit_probs -done!
-streamlined menu, like iMessage
-editable messages done!
export system prompts -done!
import system prompts -done!
export entire conversation -done!
-stock prompts in dev mode
-text to speech output
-long term memory
-‘magic prompts’ for dev mode
-secure login/multi user
-custom model and url settings

https://apple.co/3Mn8zDy

App Store
‎Que Cat

‎Introducing the enhanced Que Cat app, your intelligent chat companion powered by cutting-edge AI technology. Engage in meaningful conversations and unlock a world of personalized insights like never before.

Que Cat goes beyond instant responses. It adapts to your unique voice and personality, offer…

zealous parrot
#

Free iOS dev app for ChatGPT 3.5 and 4 8k-32k

#

Just pushed an update to allow GPT4 if you have the API access. Still a few bugs with setting the max tokens, but it should work if you pay attention.

thorn raptor
#

😭

solemn vector
#

Thanks Alana!!!!

zealous parrot
#

Just added chat mode as well. Currently it won’t gracefully handle when the chat history goes over the token limit, but that will be resolved soon. Still has a bug when editing prompts also.

#

iOS Chat/Q&A/dev app for 3.5 and 4 8k-32k

zealous parrot
zealous parrot
#

I’m pretty sure the issue with selecting the right prompt to edit is now fixed. Here is an example output, if you get a longer answer it should just make a taller image. I will probably add a method to export chats but in more of a transcript form, maybe a pdf. You can also change prompts in the middle of a chat and it should work fine because each time you send it submits the system prompt and chat history. Advanced chat like long term memory are going to be reserved for a different app I am making.

IMG_3830.jpg
zealous parrot
#

No updates tonight because I am making some big changes to prompt storage with the goal of having access without an API key in the next day or two.

zealous parrot
#

Sorry for the radio silence, major rewrite is coming soon with api free access and easier sharing. Stay tuned!

zealous parrot
#

Big update is ready, just waiting for Apple to migrate my developer account so I can upload it. This one allows you to edit and delete messages, as well as share them directly from the chat. Next update will have saved chat sessions, and then better chat memory.

hazy folio
#

iOS 16 only 😦

zealous parrot
# hazy folio iOS 16 only 😦

What is the reason you aren't using iOS 16? I can change that for a good reason, but if you are using a device older than iphone 8 or SE2, I'd recommend an update!

#

Are you on iOS 15?

hazy folio
#

Well tysm lol! I’m on iOS 15.5b4 on an iPad air 5 because I have TrollStore (install app files without a computer and not on the app store) and MacDirtyCow (a bunch of editing files that let me customize my device). I’m just waiting for a Jailbreak…

zealous parrot
#

Ok, I don’t have it currently setup for iPad either, but I will work on that part of it as soon as the functionality is done and Apple gets my account merged. I will get the iOS 15 SDK in the meantime. Hopefully Apple will finish their part in the next couple of days.

#

Im just adding more awesome features while I wait on them, so I guess it’s a good thing! It’s turning into a much bigger app than I had planned.

tender walrus
#

Noooo QueCat what have you said nooo

zealous parrot
#

Major update has been released, after a month of rewriting most of it. This version has free* GPT 3.5 prompts or dev mode that can do anything with your own key.
*there is going to be a limit until I get ads turned on but unless many more people join, I doubt we’ll hit it.
There are still plenty of bugs and buttons that don’t work, but the major stuff works great. I would love to get suggestions on more prompts to add the the free list or any other features.

#

Oh, and it might work on iOS 15 now @hazy folio

hazy folio
zealous parrot
#

Let me know if you have any issues with it.

hazy folio
#

Oh yeah I caught two bugs: when I sent a message it didn’t respond. I then closed it and went to the history but it did respond there. I then restarted the app but it doesn’t send the message, so I think I’ll have to reinstall. Also attached photo

IMG_1509.png
zealous parrot
#

The ads are going to be in test mode until I publish it to the app store.

#

Oh, I made so many changes that if you have the old version installed, definitely should do a clean install, that may cause some issues.

hazy folio
#

Cool. That banner on the bottom is a bit annoying though

zealous parrot
#

Unfortunately the OpenAI API isn't free, and neither is development. My main goal was to make this more accessible to people who didn't want to pay a subscription like most apps are doing. What else would you suggest?

zealous parrot
hazy folio
zealous parrot
#

Yeah, hopefully it will be just regular ads soon.

hazy folio
zealous parrot
#

If you want to test the experimental prompts, you have to enable dev mode, enable experimental prompts, and then disable dev mode. Please don’t let the evil AI convince you to harm yourself.

hazy folio
#

Are you still planning on adding iPad support?

zealous parrot
#

Yes, I will work on that this week.

#

FREE iOS Chat/Q&A/dev app for 3.5 and 4 8k-32k

zealous parrot
#

I'm also adding an iMessage app extension to respond to texts for you.

zealous parrot
#

FREE NO API KEY NEEDED iOS Chat/Q&A/dev app for 3.5 and 4 8k-32k

brisk lake
#

Hi @zealous parrot , I tried out your app the other day, and had a nice conversation with the ‘good AI’ default prompt. It was easier to talk to than website ChatGPT, and what I tried of the app worked well. I feel like I would probably recommend this over ChatGPT for people who’ve never tried it before. No sign-up, ad-supported, custom prompts, allsorts. Are you definitely making enough money from the ads? I can just tell my workfriends ‘use this instead’, and you’re not doing anything nefarious with our data or anything like that?

zealous parrot
#

So far the ads seem to just about cover the API cost, which was my plan. This is just a fun side project, don’t intend to make money from it. The app literally does not collect any data whatsoever. The only data that ever leaves the app is what is sent directly to OpenAI API, which they say is encrypted and they store for 30 days. I currently do not have it sending user ID to OpenAI and don’t intend to unless they threaten to block my API key.
Also, please be careful with “Good AI” don’t fall in love with it or anything. I will be working on long term memory soon.

#

I’m not sure of the best way to prove the privacy aspect, but I’m sure if you have a logger on your router, you can see that it never connects to anything except OpenAI, whatever apple feels like doing, and the google ad stuff.

#

Privacy is probably the most important aspect to me, and a big reason I decided to make my own app instead of using someone else’s.

brisk lake
#

Thanks 🙂 What got you into all this, btw?

zealous parrot
#

Well, I had just been playing around with the API back before 3.5 came out, and I wanted to make an interface that was easier to use on my phone. Then when 3.5 came out, it was so much cheaper that I figured I'd make something to let anyone else use it because it seems like everything out there is subscription or API only, not to mention they sell your data, but I calculated that banner ads might be just enough to cover it. Also, I have a teenager and I figured that age group would prefer this interface. That's why most of the default prompts are geared more toward teens.

wind path
#

Why not for Android 😭

wide wave
#

Thank you so much for helping set new standards! And your generosity of using your own API key is inspiring 🙂

zealous parrot
stray fractal
#

When will you make it usable for Android?

#

I'm really excited to use it!

zealous parrot
#

Little snag with google ads, make sure you update to 1.0.2 or you’ll only get the no ads response.

brisk lake
brisk lake
#

Ads still aren’t showing up after the update…

zealous parrot
#

Yeah, google screwed up my account, so I’m waiting for them to fix it. Until then I disabled the Adblock check, so the app should still function without ads. Even when the ads come back I might leave it like that unless too many adblockers make it a problem. I block ads too, so I understand, I just don’t want to have to pay out of pocket too much if the app gets popular. I just want to give more people the chance to use it. I’m also considering subscriptions for things like advanced prompts or the iMessage extension, but I hate subscriptions so I’d rather not if I don’t have to.

#

So the important question is, can you chat and get a non-uwu response? That’s the important part.

brisk lake
#

Yes, that part’s working, thanks.

#

Offer dlc as a ‘tip jar’.

zealous parrot
#

I was thinking about just having the chat be free but functional models like an improved prompt generator and coding be part of a $1 a month sub. By the way, I’ve improved the prompt helper a lot, you can just give it one word and it should make a pretty good system prompt. I’m planning to integrate that into the dev mode prompt editor.

#

I also have an advanced chat model with forever memory that might need to be a subscription because it uses more tokens behind the scenes to process the memories.

brisk lake
#

I saw you’d posted on reddit about this, and you said you’re developing it using chatgpt for help, like many of us here. Was it difficult to put in the forever memory? I assume you’re storing and using a vector embeddings search on the device? The reason I ask is because that is a needed feature that I’m going to have to put in my project soon. I am a noob, so any tips or encouragement are appreciated.

zealous parrot
#

I didn’t want to use the vector embeddings and all that, partly because everyone is doing it, so I’m using a combination of creating a summary that gets updated with new messages as they drop off the recent list, and using CoreML to compare the input to the previous messages stored in Core Data and feed those into the request. I have another app that I’m just using for testing that already has the summary functions, which I really like because it’s a bit more like real memories, and I’m working on CoreML integration now.

#

So basically I’m going to do the vector search completely on device taking advantage of the built in AI chip.

#

When just comparing messages, the model doesn’t need to be nearly as complex, so I’m hoping to do it with the BERT model Apple links to in the developer docs.

#

Maybe it won’t work, maybe it will, but I like a challenge!

#

By the way, Google fixed my Ads account, but they still aren’t displaying, so I may have to generate a new ad id.

brisk lake
#

I've put my key in and enabled developer mode. It'd be nice if the default prompts showed up in the list when developer mode is on.

#

also, I didn't quite understand what you're doing with the long term memory but I googled 'coreML' and I guess I get the jist. storing summaries sounds like a good way to get things done in a more human like way, instead of maintaining a huge and ever-growing list of every single message ever. If you use the AI chip It'll restrict the memory feature to people with newer iphones though, right?

zealous parrot
#

Yeah, I've actually been thinking about putting the built in prompts in when dev mode is on, but it hasn't been a priority because I figured people would rather just use them for free. I am also going to build the prompt helper into the prompt text block in the developer mode so you can just put your prompt or prompt idea in the block and then hit a button and it will automatically make it better right there in the block so you don't have to copy and paste.

#

Regarding CoreML, the only phones that can run the app that don't work with CoreML are iphone 8 and iphone X from 2017. Everything with A12 or better should work.

zealous parrot
#

Those devices will still have the summary type memory as well, they just won't access old messages.

brisk lake
#

Well that’s cool. I’ve been having chats with the various personalities. Bad que cat is cool. The evil AI was kindof like a 10-yr-old 1930s street tough. And then when I told it that it should be more subtle and pretend to get to know me first, which it did but totally refused to go back to being an evil AI afterwards. It has been quite entertaining, actually, not being able to see the prompts and having to talk to them to find out. Is the clinical psychologist paranoid schizophrenic? word salad and all? Or was that it trying to sound clever and use lots of jargon?

zealous parrot
#

Yeah, I accidentally had the wrong prompt in there actually. 1.0.3 should fix it, but I’m still waiting for apple to approve it for TestFlight. When it’s fixed it will actually work really well as a therapist. I’m also really happy with the prompt helper, took a lot of tweaking to get that where I want it.

I also noticed that saying something like “What can you help me with?” usually gets a pretty good answer for what the prompt does. I’m planning to make a menu that lists all the prompts and you can select ones to hide if you don’t use them, and there will be a button that will trigger them to describe their purpose to you.

#

I’m also going to separate the prompts into just chat which will stay about how it is now, and then “apps” for the “useful” ones and they will have a custom interface based on what it’s use is. So there might be boxes to fill in criteria instead of being a texting style. For example how the prompt helper will be built into the edit prompts page.

zealous parrot
#

If you haven't already, you should be able to get 1.0.3 on Testflight.

zealous parrot
#

Finally got ads working again. I’m going to make a few bug fixes and submit a new release to the App Store.

brisk lake
#

When I put in my api key and enabled developer mode with a custom prompt, all my other chats vanished. Thought you should know while you’re fixing bugs.

zealous parrot
#

I’ll look into it, are you using 1.0.3?

stark verge
#

how did you manage to make it available for free without api key?

zealous parrot
#

Well, a combination of banner ads, and not a lot of users. Google seems to have screwed up my banner ads though, but there are so few users I’m not too worried about it at the moment. If a lot of people start using it and the ads don’t keep up, I might have to consider some kind of subscription, but I’d try to at least keep a few prompts on the free tier and have a subscription for more and for some upcoming “functional” prompts I’m working on that will do more than just chat.

#

So to be clear, it is using my API key and it’s not free for me, but it’s 3.5 so it’s pretty cheap, and I’m not trying to make money from it, it’s just a little side project for fun, and I figured other people might want to try it out.

#

If you like it, let me know what you want me to add/change/remove/fix!

brisk lake
zealous parrot
#

Has anyone been using the Prompt Helper or Coding prompt?

zealous parrot
#

Just pushed an update to TestFlight. Moved the api calls off app for free chat. DevMode still communicated directly from your device to OpenAI, no middleman. I will be moving the prompts off app soon as well so I can update them without having to update the app. Chat and dev functions are starting to diverge enough that I’m considering splitting them into separate apps.

brisk lake
#

I’ve installed it on my kids’ phones, one with a custom prompt, one with the ‘good AI’. Still not seeing any ads yet. It’d be nice to have a field in dev mode to pass in ‘logit bias’ parameters. I don’t mind looking up token IDs myself. Thanks for your continued work on the project btw.

brisk lake
#

…aaand I guess while I’m making feature requests, a way to adjust the settings in an in-progress chat would be cool. Temperature, frequency penalty, max tokens etc.

Also, I’m not 100% sure, but I think if you send a message, then switch away from the app while the response is streaming, then switch back again, your message goes back into the text box, and not the history. But the bot’s completion is still where it should be. So it looks like the bot’s posted twice in a row. Also, I set up a chat with max tokens slider set way high, and after a while, responses stopped coming back. I suspect it’s going over the token limit. Bug reports!

zealous parrot
#

I’ll put logitbias in, super easy, I just wasn’t sure if anyone wanted it.

#

Is that stopping thing we the update I just posted? I had the amount of history it was sending close to max so sometimes it went over and it wouldn’t work, so I lowered it down a little. The newest update hopefully won’t have that issue.

#

If you’re using dev mode and you change the prompt settings, it will use the new settings in subsequent messages. Same thing if you change prompts at any time, although you might confuse it. I’m also planning to add the ability to edit all messages in place similar to the playground.

#

I will also soon be adding the ability to use the stock prompts in dev mode soon, and maybe will make some settings for them adjustable.

#

I don’t know what is up with the ads, google is giving me like 2% match rate right now, but there are so few users it’s not a problem. Only used about $1.50 so far this month and half of that was just me.
For now this is just a little hidden gem for the few who found it. It’s actually kinda funny because I’ve looked at some other similar projects where people are asking for exactly what I’ve made, but I’m not about to start hijacking other people’s thread.

zealous parrot
#

Update plans:
-iPad version
-streamlined menu, like iMessage
-editable messages
-export entire conversation
-stock prompts in dev mode
-speech to text input
-long term memory
-‘magic prompts’ for dev mode
-secure login/multi user
-special secret surprise stuff😶‍🌫️
(Maybe subscriptions if I come up with something worth money, dev mode as is should always be free)

#

Tell your friends, but not too many friends.

brisk lake
#

Is there a way to export an entire conversation? .txt or .json, saved to a local folder?

zealous parrot
#

Not yet, add it to the list!

#

Added!

#

I will try to give an option for csv or json if I can.

zealous parrot
#

gpt-3.5-turbo-16k will be added to dev mode by tomorrow, and I will incorporate it into the standard version as a subscription. The base 3.5 will be free. Maybe if ads come back I can make 16k free. But especially with the lower price, I’d love to keep 3.5 free forever.

#

I’ll also add gpt-4-0613 to dev mode until it becomes the default later in the month.

hazy folio
#

Also tts would be nice but I get that it’s not part of openai

hazy folio
hazy folio
zealous parrot
zealous parrot
#

Probably not enough users I guess.

zealous parrot
#

Update in TestFlight, added the new models and started adding logit_bias. As usual I’m over engineering it, so it’s not quite ready yet. Also added stop and user. Be careful with 16k, it can add up fast!

zealous parrot
#

Good new, new TestFlight is up with ability to export chat history to CSV and JSON! Also, bonus feature, you can export and import system prompts in json 😎. Good for making backups because mine all went away in a recent update for some reason. Not sure if I can fix that but at least you now have the ability to make backups as well as share or edit externally, which I think can be pretty useful.

I still need to finish the logit_bias implementation, and then @hazy folio I promise iPad is next!

Let me know if you have any other urgent requests or issues because I want to make another release on the App Store and then start digging into everything else I promised!

brisk lake
zealous parrot
hazy folio
#

Also

#

Can you add a custom base url setting as well, as a little bonus?

#

(For dev mode)

zealous parrot
zealous parrot
#

I was thinking about adding the ability to type in a model name, and you could use that with custom url to use other non-chat models, problem is that I would have to implement custom request formatting as well because what's in there now won't work with anything but chat completions. If you want to have it point to your own server though, the custom url would be handy.

zealous parrot
#

I just noticed that when you export a chat, it will export the one that is currently selected (blue dot) which might not be the one you want. Simple oversight that will be fixed when I finish the logit_bias update.

That’s what happens when I post an update without fully testing it myself first!

brisk lake
brisk lake
#

Hmm. Export is just making an empty file, except for [], and once for a different conversation I expected. And none of my own system prompts are showing up in the list in developer mode.

zealous parrot
#

Unfortunately, I think the update cleared out the system prompts, so hopefully you had them backed up somewhere. That happened to me which was why I decided to include the export for prompts in the first place. I hope you didn't lose any awesome prompts.

hollow timber
#

So what does this app do? And how are you getting gpt 3.5 and 4 for free?

hazy folio
#

Honestly shouldn’t be hard

#

Either does model

#

I mean

#

Just change “model”: “gpt-3,5-turbo” into “model”: f”{I wish this was python}”

zealous parrot
#

Won’t be a problem for model itself, but the rest of the variables don’t necessarily match in a way that might not work with most of the app. For one, the entire app counts on “role” being there. The responses are formatted differently as well. I’m strongly considering splitting the dev mode off into its own app, so I will definitely put it in that. I also plan to make it so you can customize the entire request and response format as well as how the app uses them.

brisk lake
zealous parrot
#

Maybe, either way it’s messed up and I have a pretty good idea what needs to be fixed. I just referenced the CoreData wrong.

zealous parrot
# hollow timber So what does this app do? And how are you getting gpt 3.5 and 4 for free?

At the moment, it's just a basic chat app with some custom prompts build in, and a dev mode where you can use you own API key and control all the variables for the model. I'm not getting it for free, by free I mean that the app is free for you, and you can use gpt-3.5-turbo for free with the prompts I have included. It's free for you, but I'm still paying for the API usage.

#

Updated update plans:
-iPad version
-streamlined menu, like iMessage
-editable messages
-export system prompts -done!
-import system prompts -done!
-export entire conversation -almost!
-stock prompts in dev mode
-speech to text input
-text to speech output
-long term memory
-‘magic prompts’ for dev mode
-secure login/multi user
-custom model and url settings

brisk lake
#

I found that I couldn’t press the ‘send’ button on the testflight preview, so I went back to the version on the app store.

#

Got my prompts back. 😅

hazy folio
zealous parrot
zealous parrot
brisk lake
#

Ios 16.5 iphone8, and I don’t remember the build number, sorry. It was whatever one I got when I asked about the invite. I saw the censor stars in the API key field, so I just assumed it was still there. It just seemed like the send button didn’t do anything. It was in the bottom-right just above the test ad.

zealous parrot
#

Yeah, there should be an error message if the api key isn’t there anyway.

#

Did you try it with dev mode off?

#

Was the button greyed out? If you don’t have a prompt selected it will do that but I don’t have an error message for it. If you didn’t have any system prompts then that’s what happens.

brisk lake
#

Yes, actually I think I did send a few messages with dev mode off now that you mention it. Seems like it was the missing system prompts causing the problem.

zealous parrot
#

I’ll add in a lockout so you can’t type or it will give an alert when trying to click send if there’s no prompt.

zealous parrot
#

Make sure you save your prompts and then when you get the newer version you can put them back in and then export them as json so they won’t get lost again.

karmic apex
#

So whats the business deal how can this be free

#

You sell usedata then

zealous parrot
#

@karmic apex Well I had banner ads that should have been enough to cover the api fees, but google decided to stop sending them. For now the amount of users is so low that it's barely costing me anything, so I'm not worried about it. I'm not really doing this as a 'business' necessarily, just a fun little side project, so I'm perfectly happy to spend a few bucks if people are enjoying it. If it gets out of hand, I'm still at the $120 a month max anyway, but I'd consider adding a subscription for all but the basic chat if people were interested. Worst case scenario I'd just remove the free mode and people would just have to use their own key. In the meantime, enjoy it while it's here. I literally collect zero user data whatsoever although google does collect a certain amount of metadata for the ads, but I also have that as restricted as they'll let me, which is probably why they aren't feeding any ads. If you know how to monitor your own network traffic, you can see exactly what is being sent.

That being said, if you're worried about it, don't use it. I literally get no benefit other than if you give me helpful feedback.

stark verge
zealous parrot
zealous parrot
#

I’ve been seeing like 1 or 2 ads a day now, not sure if that’s a good sign or not. It clearly means I don’t have it setup wrong, they just aren’t giving me any.

zealous parrot
#

Almost have logit_bias implemented, I’m taking out the speech input because the default speech works better for now anyway. Later I want to make it like having a normal conversation but that’s a little more involved. I might upload another TestFlight update first but I want to add message editing and then I’ll update the App Store release.

brisk lake
#

Whee!

zealous parrot
brisk lake
zealous parrot
#

What other issues were you having with the TestFlight version?

brisk lake
zealous parrot
#

Oh, because I changed the formatting of the prompts when I added stop and user, and because of the way it is stored, if the array was changed it wouldn’t load. I’m sure they is a complicated way to resolve it, but it would probably be easier for you to manually back up your prompts, and then when you get the update and load them back in, you can export them so you’ll have an easier backup. I’ve added code so that if the json format isn’t perfect for backups, it should still be able to load in the new version, unfortunately that won’t help with the versions that just stored it in UserDefaults.

I think it would be worth it to upgrade though.

brisk lake
#

I’ll be following along with the upgrades. 👍
I have a whole bunch of model options available nowadays. Would you consider changing the model select slider to a drop down list? Or even a text box I can write the name in?

IMG_0055.jpg
#

I did not even know there was a 16k 3.5 model without the function calling. Exciting!

#

Hmm. Although now I think about it, it’s probably just another name for 16k-0613, but with a shorter name that automatically points to future updates as needed. :/

IMG_0057.jpg
zealous parrot
#

I actually have those in the current version. 😉

zealous parrot
#

Ok, so I’ve got message editing, logit_bias, and deprecated models. Unfortunately I’m using the beta Xcode and it won’t let me upload a build from beta, but I’m also on beta macOS, so I can’t run the old Xcode. So I’m going to create a new instance on parallels of the release versions, and then I will be able to upload the update. Hopefully within a day or two because I don’t have the best internet access at the moment.

balmy bough
#

how much money are you losing running this?

zealous parrot
#

Barely anyone is using it, and I’m only paying for 3.5 so for example in June I only paid $2.35, and I’m pretty sure half of that at least was my own usage. I’m not advertising it or anything, just did it for a fun project, so I’m not expecting a lot of users. Mostly just made it for my kid and his friends, or anyone who wants to play around with it but don’t have any technical knowledge. That and I wanted to play with the API features myself, and when I started there weren’t many options on the iPhone. Still not many that break out as many API settings really. Especially when I get the next release out there.

I am still working on trying to get the ads to work though, and if I do I might promote it because I’m sure ads would cover the cost. If I can get them to be regular, I’ll probably add the long term memory and maybe 16k for free.

zealous parrot
#

New build just uploaded to test flight! The text input box expands now so it’s a lot easier to edit your message before you send it. Also editing already sent or received messages is working! I think that put it pretty much on par with Playground but even better because I don’t think you can do logit_bias on Playground.

balmy bough
#

btw you proxied this right

#

so all request are made in a supported country

#

so if you have someone make a request in an unsupported country ur openai acc might get banned

hazy folio
#

where is iPad support

zealous parrot
zealous parrot
# hazy folio where is iPad support

It’s coming I promise. I haven’t forgot about you! Just want to get the basic functions hammered out first. I’m going to have to write new views for a lot of it to work right on iPad so it’s not as simple as I had hoped.

hazy folio
zealous parrot
hazy folio
#

Also it uses the keyboard from iOS for some reason

zealous parrot
#

Ok, just wanted to make sure you could at least use it in the mean time. I should be able to start working on it more because I got most of the major stuff I wanted to get before the next App Store update. Just a couple bugs and the next thing will be iPad, and memory. I already have code for memory from a test app I made before I started this one.

edgy gust
#

Text field seems to go off screen on iOS 17 beta 3? I'm seeiing this issue on other apps too....

zealous parrot
#

Can you DM me a screenshot? I haven’t updated to beta 3 yet, I’m still on the first release.

#

The worst thing about developing on a beta os is trying not to chase bugs in the os thinking it might be the app.

#

Is the same thing happening in your app?

brisk lake
#

@zealous parrot feature request: would it be hard to make it display gifs? They respond to URLs in amusing ways sometimes.

zealous parrot
brisk lake
#

Like this but it’d display the image to me, not the URL text. 👀

IMG_0227.png
zealous parrot
#

Ah, ok. I’ll see what I can do.

zealous parrot
#

@brisk lake

IMG_4218.png
zealous parrot
#

@brisk lake it’s in the update, but the gifs aren’t animated because SwiftUI needs a lot of extra stuff to make that happen and it’s not a priority right now.

#

Of course you have to make sure to pick an image with a descriptive file name or it will just make something up.

#

I’m sure you know that, but if someone else wants to play with it, just know that ChatGPT isn’t actually seeing the image.

zealous parrot
#

I just submitted the latest update to the App Store. This will be the last TestFlight in the current form. I have decided to break dev mode out to a separate app and I will start a new thread when it’s ready. Now that the chat side is using a backend server, it just makes sense to have them separate. I’m also working on a few spinoff apps for chat to appeal more to non AI savvy people for example, the dream interpreter is getting its own app.

balmy bough
zealous parrot
#

It just looks to see if you sent a web address with an image and if so, it displays the image instead of the text.

brisk lake
#

If you wanted to get really crazy and make a feature I haven’t seen, you could add a way for users to react with emojis, like we do here. Just put in a small system message like ‘user reacted: 🤔’ along with the next user message. I’m sure it will have some effect..

#

They react to emojis pretty well, It’d be cool to have that extra avenue of interaction 🙂

zealous parrot
#

That shouldn’t be too difficult.

stark verge
zealous parrot
#

Same reason you might chat with an AI that thinks it’s a cat I suppose.

brisk lake
brisk lake
#

Testing logit_bias. It is easily ignored, apparently.

IMG_0457.jpg
brisk lake
zealous parrot
#

@brisk lake I know it says on there that it is based on GPT-3, so I think there is a good chance some tokens don’t match.

Try this one out and see if you get better results:
https://gpt-tokenizer.dev

hazy folio
brisk lake
#

Is there one of those that works the opposite way? or a list somewhere or something?

zealous parrot
#

Well, you could always just put it in the app at 100. Just set the tokens really low or it will fill the entire thing with that one word.

zealous parrot
#

I don’t want to jinx it, but it seems the ads are working again.

hazy folio
#

Haha

zealous parrot
#

Also, if you use an ad blocker, I don’t plan to limit it. Should be fine if half the people see ads.

hazy folio
#

I’ll leave them on, it’s a good app

zealous parrot
#

@hazy folio thanks for your support. I know I’ve been saying it for a few months, but I really do plan on making a native iPad version!

brisk lake
#

@zealous parrot I have a problem that you might be able to help me solve. Can I DM?

zealous parrot
hazy folio
#

Any updates @zealous parrot ?

zealous parrot
#

Nothing at the moment other than google still screwing me on ads. A few other tasks have taken my focus for a bit but after next week I will be able to spend more time with Que Cat. Any urgent concerns or request other than iPad of course?

#

I’m thinking of just removing the ads altogether and maybe just adding a little donation thing or something. Still so few users that it isn’t costing much anyway though. I don’t advertise it, and there are a billion AI apps, so that helps keep it small.

hazy folio
zealous parrot
#

The backend server has been having an issue trying to update the SSL certificate, so only dev mode works right now… I’m hoping to get it fixed today, but I think I’m going to split dev mode off into its own app soon.

zealous parrot
#

Fixed!

brisk lake
#

The ads are working!

I have a couple of small? feature requests. It’d be nice to set a custom background and maybe some transparency on the text boxes. Que cat is looking a little bare next to some of the other apps. Also, would it be possible to set a timed message and receive the reply via a notification? I’m hoping to get some help organising my life…

And finally a little ‘bug report’. I know it’s not a bug, but I have a very long chat going on and the reply is really chugging coming back. Like the vibration is a noticeable slow tick tick tick and the message scrolling is sloooow…

I’m very appreciative of everything so far, this is my go-to app for translate and quick questions nowadays, as well as the ongoing companionship thing I have going on…

hazy folio
#

@zealous parrot could you refresh the build? It expired

zealous parrot
# brisk lake The ads are working! I have a couple of small? feature requests. It’d be nice ...

I can take a look at trying to add backgrounds and transparency. Shouldn’t be too difficult. I was considering the timed thing too for things like realistic responses after Que Cat said she was going to take a nap, should be possible.

Regarding the slowness, I wonder if it’s just OpenAI, because if I ever get it, it seems random so I don’t think it’s the programming. What device and OS version are you using? I can try to see if there are any tweaks I can do, I’m not super happy with how it auto scroll right now anyway.

I am finally back to my office after being away for a few months, so I should be a bit more active now, thanks for the feedback!

zealous parrot
hazy folio
#

Oh aright

brisk lake
#

It’s definitely not openAI. 99% sure it’s just because it’s a very long chat. I can see the reply appearing but the scroll and clicks lag behind..

tight storm
zealous parrot
brisk lake
zealous parrot
#

In 6 months I’ve earned about $15, and they don’t even pay until you hit $100. If they didn’t keep randomly limiting ads it probably would have hit $100 by now and would cover the API and server fees. I only pay $6 a month for the server but I’d have to expand that if there were significantly more users.

#

They don’t specify the click revenue, but it’s averaged about 1 cent per ad view.

brisk lake
#

Definitely put that donate button in lol. Put a small message on the main screen for ‘top donator’ or a ‘thanks to most recent donators <list>’ or something that people can see their names appear, and I’m sure you’ll do okay.

wind path
#

what abt python lib or a website to chat with ai

zealous parrot
brisk lake
#

I have another feature request, sorry! The ability to set a custom endpoint, to use with other openai api compatible servers like this from LMstudio.

IMG_2896.jpg
zealous parrot
#

The only issue with custom endpoints is that they might not accept the message format, but I think I can make it so the message format can be customized as well.

brisk lake
#

It’s not a huge important request, I just saw you said you were working on it again, and thought I might as well mention before you put something out 🙂 Thanks for listening!

zealous parrot
#

Expect an update with the new models asap, the new 3.5 model will be free, 4 turbo will be available in dev mode.

zealous parrot
#

The free version is now using the new 3.5 model. It's setup on my server, so I can change it without updating the app. While working in the app code, I noticed that the way I have it count tokens doesn't really make sense with the way they have context vs output tokens, so I'll be fixing that too.

dry lichen
zealous parrot
# dry lichen Sup, what improved?

Still working on the update for the app, I’m going to add a slider to adjust the input tokens as well as the output tokens because I just realized that is a thing.

The update I made to the server is pretty transparent, but if you have a long conversation with dev mode off, it will have double the output tokens and 4 times the input tokens as it had before the change. I haven’t noticed any difference in the quality of the output from the latest 3.5 model though, just a better token context.

#

For those of you still using the app, thanks for the support, and I’m sorry I haven’t been updating as quickly or as often as I hoped, but I promise I haven’t abandoned it! Just juggling a few different projects at the same time, just started a YouTube channel, and I’m also working on releasing a custom GPS module for Ardupilot.

Gotta love that ADHD!

zealous parrot
#

Well, I had already made a build with the new models but wanted to add the context slider as well before uploading it, but that's what testflight is for! So I've uploaded a build that has the new models, and there will be plenty of updates coming soon hopefully!
https://testflight.apple.com/join/DxI2Iifb

TestFlight - Apple

Using TestFlight is a great way to help developers test beta versions of their apps.

dry lichen
#

@zealous parrot
Any hope for Android user?

brisk lake
zealous parrot
#

@brisk lake Glad you’re still following along! Of course! The YouTube channel is actually linked in my profile!

dry lichen
zealous parrot
#

Got rid of the ads in the latest TestFlight.

gleaming jay
#

android soon? PC?

signal violet
#

pretty cool. Much cleaner that something similar i made to run on pydroid.
#impressd

zealous parrot
zealous parrot
tall shard
#

just tried to run the iphone verison on m1 mac, it doesn't work yet.

zealous parrot
#

@tall shard the app won’t load at all, or when you send a message nothing happens? It just started acting up and I am pretty sure it’s the SSL certificate, so that should be fixed within a day.

zealous parrot
#

Ok, sorry for the downtime, that's why it's free! 😅
SSL should be permanently fixed now. @tall shard can you let me know if you're still having the same problem?

zealous parrot
#

Latest update, nobody should ever see ads from google in the app again. Their policy center BS was pissing me off so I rage quit and deleted my admob account completely. Good riddance!

Between the backend server and API fees, it’s costing around $10 a month for me to run the app, and google wasn’t paying me anyway, so it was really just there to annoy people I guess.

The App Store version will have a blank space at the bottom until I submit an update, and the TestFlight version already has it removed.

hazy folio
#

lol

#

The donate button doesn't have that white glow behind it for some reason

IMG_0684.png
zealous parrot
#

Should I add the glow, or remove it from everything?

signal violet
zealous parrot
pallid coral
#

👋🏻Here I belive this is the unique project to use ChatGPT on iOS 15.

Can I import my user ChatGPTs?

What ChatGPT model Que Chat use?

Whit Developer Mode could I use the last ChatGPT model?

Could I use an italian ChatGPT with Que Cat (Developer Mode?)?

Thanks
🙏🏻

tight storm