2.0 roadmap #180
Comments
|
Note: switching to the new endpoint will cut off features like guest mode, and some endpoints/features might be removed since they are not available on the new interface. |
|
Hello there again, I've opened a new branch (2.0) to handle directly here: https://github.com/realcoloride/node_characterai/tree/2.0 Currently work in progress. I am busy working on another project at the same time, but I will try my best to keep up for a new update. |
|
Hi, do you think it will be possible to use/manage persona? |
|
What do you mean by persona? |
|
|
|
Hello everyone, I come forward to indicate the migration might be happening sooner than I thought. Announced today, the CharacterAI team have officially announced the departure of the old characterAI interface and endpoints (which will inevitably break the older versions of node_characterai). Source: https://www.reddit.com/r/CharacterAI/s/W0T6cZ3B9q Personally, I think this is a bad idea. I believe the new website has a worse interface and worse language capabilities and the model quality isn't the same and plenty of features are missing. So, I recommend users of node_characterai to prepare for these changes as the new rewrite of the package will come sooner than expected and everyone will probably have to migrate, and I will have to also handle the new package version faster. The new migration will also take new features in place, and they will probably be in newer versions of the 2.0+ branch. The changes will take immediate effect on September 10th. |
|
Sure. This will be probably added as a feature in node_characterai if possible. |
|
Hello, anyone else been having issues using the new interface? I am getting a lot of CORS related errors whilst using the website. |
I just checked, and I'm not having any issues currently. |
realcoloride
changed the title
|
Hello there! Just want to let you know all that I am actively working on 2.0 and it took me a long time to find the perfect design, but I will now aim for developer experience first. I just managed to get my first dms to send. Still got quite the stuff to go. Also, no puppeteer is used. Atleast for now. Again like usual, if you have feedback, feel free to let me know. |
|
Used puppeteer is hard enough to setup and make server resources too heavy, hope you can make it more lightweight in 2.0 ver, wish u the best for this! |
|
Hello,
Thank you for your support. I am using the Android's app's endpoints.
Meaning that if they plan on adding anti fetching measures like they do
with cloudflare, they would have to constantly update the Android app,
which would probably cause some problems for their massive mobile app user
base. It all relies on how long before I am forced to use puppeteer for
good.
For now, it has been doing fine. Some endpoints are not available without
cloudflare or specific things that I cannot fully get (like following an
user for example) but we'll eventually get to it.
Cheers
…On Sat, Sep 28, 2024 at 11:33 AM Nadila Vira ***@***.***> wrote:
Used puppeteer is hard enough to setup and make server resources too
heavy, hope you can make it more lightweight in 2.0 ver, wish u the best
for this!
—
Reply to this email directly, view it on GitHub
<#180 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AZ4WO5M5REBEE5X7RN3ZP2TZYZZVVAVCNFSM6AAAAABLJPKDL6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGOBQGU4DCNRVHE>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
Just a suggestion, you may want to look into curl_cffi / curl_impersonate |
It's not really necessary at the moment, but it is definitely something worth a note. Thank you for the suggestion. |
+ 1 for curl impersonate. You don't need a whole browser, in fact you just need any http library that can make its TLS fingerprint look like the browser's one. In short, the connection parameters of regular http libraries are different from those used by popular browsers and some websites use this to block requests from “not real” users. You can read more about this here: https://lwthiker.com/networks/2022/06/17/tls-fingerprinting.html |
TODO 2.0 (better) RoadmapEverything already done will not be shown here. Check the links by yourself to see what's done and what not. TODO means that I have not planned on looking forward to it yet. The classes below are in priority, but the extents of the classes shown below (because I use classes from now on and types) are not shown for the sake of simplicity, see the codebase for yourself. The following emojis have the following meaning:
|
Hello, after a long time and work, the I am proud to say the TTS/SST call feature is workingIt is experimental and I am using ffmpeg/ffplay magic to play back and record audio, but you can also use the passthrough streams that will give you back PCM audio or listen for audio and transcribe it back. You can also hang up interrupt the character when its speaking, mute yourself or know if the character is speaking with const call = await dm.call({
microphoneDevice: 'default',
useSpeakerForPlayback: true
});
call.on('userSpeechProgress', candidate => console.log("User speech candidate:", candidate));
call.on('userSpeechEnded', candidate => console.log("User ended speech candidate:", candidate));
call.on('characterBeganSpeaking', () => console.log("Character started speaking"));
call.on('characterEndedSpeaking', () => console.log("Character ended speaking"));For the microphone input, you can choose a microphone device by name or the default one (tested only on windows). Playback is optional on the default speaker. Warning: everything I say or I show is subject to change! So, here is a demo: p1ip92_1.mp4There are a lot of voice related things to finish like editing voices and they need heavy testing and more stability insurance, but I am writing the base foundation to do heavy QA later on. Cheers |
|
Hello, The main remaining objectives have been updated to be more accurate. |
|
Hello, Persona related features and management have been completed. |
Unavailable features list (temporary, subject to change)Below are all implemented but limited features (not available yet) because of the following reasons and their potential solutions. Everything below is subject to change.
|
Hello, I need a quick poll to be done here for those viewing the thread.I plan on making GroupChat functionalities later or now, which could take some time.Please vote: Note that Group Chats are a niche functionality that are not really used by most users and is only available on mobile. |
Hello, I began my
|
|
Hello again! I have published an experimental beta version of the package with Typescript and Javascript support for 2.0. If you wish to test the new 2.0 version in beta, as I will pass some real QA testing and fix any potential bugs, and to get your opinion on the new improved feel of the package. Warning: please do not use this in production, and use this preferably in a new project and wait for a stable release before migrating. npm install node_characterai@betaTo update if you want to check for updates (I will do regular beta updates until it is stable) npm update node_characterai
Also, please know that soon, I will merge the beta codebase with the main/master one. If you wish to keep that codebase somewhere or archive, feel free to fork it now. Otherwise, stay tuned for the stable release. |
|
Hello, I come to report that for some reason, when calling, I get weird crash dumps and crashes overall. It might be an issue related to The call feature is currently broken until further notice, but everything else is not. I will keep you posted of course. If you do get the same issues, please let me know. So far I can replicate it on windows. I will try to find a workaround for the issue. Cheers |
i have issue with this package too when installing this package, it says should i really need installing the latest version VS2022? (tbh i dont want it for something that i don't use), this happen when installing naudiodon package (use --loglevel silly) just wanna see your response that maybe facing this issue later, thank you! |
|
Hello, to answer your response, I am not really sure. I'm trying to get rid of the audio libraries in question as soon as possible and I apologize for any inconvenience caused. |
Hey again! I've been gone for a looong time. Seeing about getting everything updated, I'm getting an error on my RPi 3B+, it states : EDIT : After some more digging, seems like it's simply a missing dependency based on the RPi's CPU. LiveKit doesn't seem to have released something to support the CPU though. If anyone knows anything more, let me know! For now, seems like I need to downgrade for the time being. EDIT 2 : For those interested, it is indeed possible to simply lobotomize this fine man's code with a scalpel to remove all of the voice related code, and get it functioning properly(atleast, with the very basic messaging system) on an RPi. It seems to be working very well this way, as of tonight. |
|
Hello there,
Unfortunately this is a culmination of issues I can't really solve properly myself so I will have to find a compromise and make the voice features optional (as in, probably separating the packages). I deeply apologize for the inconveniences and I don't have the proper hardware (I currently only have Intel x64 based CPU machines with Windows or Linux) to extensively test every platform on. While I try to make my code as open as possible, libraries like this make it really hard for me to interact with core components. In such I started working on my own audio processing node library in C++ but it is taking me more time than I expected and I am not sure where it will lead at the current moment. This branch is still beta meaning I am trying to get things in order but seeing that the LiveKit team isn't really reactive to my issue and there are no workarounds and crashes are much more regular, it leaves me with either the decision of splitting the voice features as its own package with its own dependencies or having to find a build compromise. There's also less realistic goals in my opinion like rewriting my own LiveKit client which would take more time than I need to but these issues need to be redirected to the respective teams. I cannot handle all of the problems for them.
While you are free to do so I don't recommend relying on that because if I make updates or update the code you will have trouble organizing. Hopefully I will try to find the best solution possible for everyone. I had almost no free time for myself recently but I am still actively trying to tie the ropes together and make sure the best experience is ensured for the developers is possible, so, please feel free so suggest if you have ideas or comments aswell. My goal is and always was to make the simplest and best experience possible but relying on unstable or unreliable dependencies create a lot of problems I hate having to go through since I do not have the full control on them. EDIT: Any other feature than calling should be functional and OK. If not please report me the issues. Cheers |
Ah, you have nothing to apologize for, such is life when trying to make an API for something that doesn't seem to want one hahaha. In my personal opinion, I think it'd be best to simply split packages like you said. Different, more advanced functionalities(I.E. - voice) can be seperate and simply installed in tandem when/if the developer wants to use it.
Don't stress yourself! What you've done already is magical in my opinion, simply work on it when you can and want to, that's all anyone can ask for. So far, I haven't tested many features, I simply use the messaging part with DiscordJS to make my bot a little more interactive and fun for my server. They do seem to love it. Only noticable issue is that every few hours, it seems to stop working with CAI. I've tried both sending a message to a CAI bot to keep the account "active" every 3HRs and simply re-authing every 1.5HRs, but neither seems to have worked so far. I don't think its an issue with your code- I just need to figure out the nuances of CAI's(what I assume is) account inactivity disconnection. If you have any knowledge to share regarding that, please let me know! Other than that- have a great day friend, and keep up the outstanding work. |
Hello!
First of, thank you all for the support you guys bring to node_characterai.
I come here to talk about the fact I want to switch to the new endpoints and much more. I also feel limited by the codebase and I'd like to use something much more strongly typed and better written. Make a much more easier interface for beginners and a less confusing one as well.
In fact I have started to get very comfortable with programming in TypeScript and such I'd like to really make a better environment for node_characterai.
If you have suggestions for the rewrite please let me know.
Major (remaining) objectives roadmap:
Current branch: https://github.com/realcoloride/node_characterai/tree/2.0
Let me know if you guys are interested in a rewrite and or if you have ideas/feedback.
The text was updated successfully, but these errors were encountered: