kwaroran / risuai Goto Github PK
View Code? Open in Web Editor NEWMake your own story. Software for LLM roleplaying and utilitys.
Home Page: https://risuai.net
License: GNU General Public License v3.0
Make your own story. Software for LLM roleplaying and utilitys.
Home Page: https://risuai.net
License: GNU General Public License v3.0
Risu version: latest(v1.28.2), web / local both
The contents of each preset are as follows:
Check "name2"(Prev: botPresets[1] -> Now: botPresets[0])
You can see that the contents of item "name2" have been overwritten with the contents of item "name1" that I just deleted.
When I delete the preset number 1(in botPresets[0]), the preset number 2(in botPresets[1]) is overwritten with the contents of the preset number 1 that I just deleted.
Let me give you a quick example of this situation:
botPresets = [{name: "name1", data: "data1"}, {name: "name2", data: "data2"}, {name: "name3", data: "data3"}, {name: "name4", data: "data4"} ...];
// After Delete
botPresets = [{name: "name2", data: "data1"}, {name: "name3", data: "data3"}, {name: "name4", data: "data4"} ...];
// Delete Once Again
botPresets = [{name: "name3", data: "data1"}, {name: "name4", data: "data4"} ...];
Originally, the expected results of this operation are as follows:
botPresets = [{name: "name1", data: "data1"}, {name: "name2", data: "data2"}, {name: "name3", data: "data3"}, {name: "name4", data: "data4"} ...];
// After Delete
botPresets = [{name: "name2", data: "data2"}, {name: "name3", data: "data3"}, {name: "name4", data: "data4"} ...];
// Delete Once Again
botPresets = [{name: "name3", data: "data3"}, {name: "name4", data: "data4"} ...];
There is an issue where the {{char}} symbol is displayed as the chat room name, not the character's name, when sending a prompt in a group chat (confirmed by log). Moreover, If I leave the chat room title blank, {{char}} is displayed as a blank in the log, and if I make the chat room title {{char}}, {{char}} is sent as is without conversion to the symbol's name. Below is a partial capture of the log where {{char}} has been changed to the chat room name.
Error
SyntaxError: Unexpected token '<', "<!DOCTYPE "... is not valid JSON
When using ElevenLabs. Im using the installer one it was working well the following days but then this error msg constantly pops up each time AI replies using voice.
I've tried reinstalling it's still the same. What appears to be the problem?
Dear Developers,
I am writing to express my appreciation for the incredible work you have done in developing our favorite AI software. I have been a loyal user for quite some time now and have found it immensely helpful in my daily writing tasks.
I would like to suggest a feature enhancement that I believe would greatly improve the functionality of the software. Currently, we can easily colorize italicized and bolded sentences, which is fantastic. However, I would like to request the ability to colorize dialog sentences in quotes as well.
This feature would be particularly beneficial for writers, as it would allow us to visually differentiate and highlight dialogues in a distinctive manner. By assigning colors to dialog sentences, we can easily identify and emphasize conversations, adding a new level of clarity and readability to our writing.
I understand that implementing new features requires time and effort, but I genuinely believe that this addition would be highly valued by the writing community. It would not only enhance the user experience but also provide a unique and customizable tool for expressing dialogues in a more visually appealing way.
Thank you for considering my suggestion. I appreciate your dedication to continuously improving the software and eagerly look forward to any updates that may include this feature. Please let me know if you require any further information or clarification regarding this request.
Once again, thank you for your hard work, and I wish you all the best in your ongoing endeavors.
This is only a very minor bug xD
The setting button (top of left side) disappear when you scroll it ;)
Since only Global Regex can be easily exported or imported, using Global Regex will be more convenient than Character's Regex Script.
However, sometimes special Characters need different regex scripts, and they may conflict with Global Regex, which can make the modification very troublesome.
So it would be great if the option to disable Global Regex can be added to the Character tab.
Rational:
It's rather tiring to press that three-line icon to surface the options icon(gear), when I am often modifying User/Bot/Advanced Settings.
Request:
Or
ETC:
Currently, the rounded rectangle profile is the default in RisuAI.
It would be nice to have a profile option that could be selected as a fully rounded circle
and the option to add and change fonts
원형 프로필 선택 옵션, 폰트 추가해서 변경할 수 있는 기능 추가 요청
here are some features I found interesting for weak computers.
Thanks for your answer.
what is the appropriate api format to interface with this software
Prefer to not have a installer.
Background: There is no global information about {{User}}, so there is a need to organize information about {{User}}, which is relatively non-intuitively delivered to AI, due to the tendency to be separately entered in the lower book / artist's note / character prompt. .
Purpose: “Global User Lorebook” allows users to set up two or more {{User}} Lorebooks (similar to character cards), so that when you use one of them on a talking AI, it will automatically {{User}} Make settings for {User}} easier to understand
Features I want:
“Global User Lorebook”: A prompt system in the form of a Global Lorebook that is supported similarly to a character card. The purpose is to include the profile picture of {{User}}, so that you can select one of them when starting a conversation.
Additional support within the “Character Card 2.0” form: Certain characters may be set to deal with {{User}} in a specific setting. Expanding the concept of overriding front-end ones when setting some system prompts in “Character Card 2.0”. By providing a more regular and formal area, just setting up a character card supports card creators and users changing and saving the settings of {{User}} they want.
Use the default content of webui url 'http://127.0.0.1:7860/run/textgen' will get Error: error in request:TypeError: Cannot read properties of undefined (reading '0'),
Use webui url 'http://127.0.0.1:5000/api' will get Error: error in request:"SyntaxError: Unexpected token '<', "<!DOCTYPE "... is not valid JSON"
what is correct url for text generation webui?
It seems that not all parameters are sent in the request, in particular max_length, which causes kobold to roll back to the defaults.
RisuAI/src/ts/process/request.ts
Lines 510 to 516 in 89015ac
List of parameters can be viewed here:
https://github.com/LostRuins/koboldcpp/blob/b08b371983932e1f528547b25469a2324d81c835/expose.h#L25-L44
In bot settings, Users can set preset but I think more feature require.
Please add changing the order of presets & duplicating the preset & Setting default preset in the user customized
It would also be great if these presets could also be imported and shared as JSON files.
However, please include SK-keys and put a warning about the risks of sharing them.
봇 설정에서 사용자가 프리셋을 설정할 수 있지만 더 많은 기능이 필요하다고 생각합니다.
봇 프리셋의 순서 변경, 봇 프리셋 복제, 사용자 정의할 수 있는 기본적인 봇 프리셋 서식 기능 등을 추가해 주세요.
또한, 만약 이런 프리셋도 임포트하여 json 파일로 공유할 수 있다면 더 좋을 것 같습니다.
그러나 SK-키를 포함하여 공유할 때는 위험성 경고를 띄워주세요.
Hey there! Thank you for wonderful project and efforts, I appreciate it a lot.
So, I found an issue in the current build (I've even installed an updated 0.7.3 .msi file) where I can't connect Risu to oogabooga WebUI. As I've learned, it happens because current code version relies on the old API.
I've modified the database.ts file to make it run on the new API, but I can't make it work. So I've decided to try to compile it by myself, hence I'm opening this issue.
May you please tell me what dependencies and libraries I need? Or maybe some extensions for Visual Studio? Thank you.
If try to import character .png from TavernAI I get this error: "path not allowed on the configured scope: /home ..." (any folder).
When I try to set image for character - it's not set. This time silent - no errors, just not set.
On previous releases it's been possible.
Manjaro Linux.
Also, with local Oobaboga's text gen UI as backend, generation limited to 80 tokens. Changing option "Max response size" don't change anything.
check --no-stream or add it as a flag to webui.py --> CMD_FLAGS = --no-stream
this is vanilla latest version clean install
Traceback (most recent call last):
File "C:\AI\oobabooga_windows\text-generation-webui\server.py", line 1046, in <module>
create_interface()
File "C:\AI\oobabooga_windows\text-generation-webui\server.py", line 957, in create_interface
shared.gradio['interface'].launch(prevent_thread_lock=True, share=shared.args.share, server_port=shared.args.listen_port, inbrowser=shared.args.auto_launch, auth=auth)
File "C:\AI\oobabooga_windows\installer_files\env\lib\site-packages\gradio\blocks.py", line 1727, in launch
self.validate_queue_settings()
File "C:\AI\oobabooga_windows\installer_files\env\lib\site-packages\gradio\blocks.py", line 1575, in validate_queue_settings
raise ValueError(
ValueError: Queue needs to be enabled! You may get this error by either 1) passing a function that uses the yield keyword into an interface without enabling the queue or 2) defining an event that cancels another event without enabling the queue. Both can be solved by calling .queue() before .launch()
Sorry if this isn't the right place to post this (or the features exist and i haven't found them), loving this project! Thank you for making it
Some Quality of Life ideas:
The title is the content.
In particular, I would appreciate it if you consider the use of this translation API not only supporting various APIs of the translator (Papago, Deepl, 3.5, 4.0, etc.), but also expanding the functions I want to implement additionally.
The extensions I'm currently developing are:
Bots reply, AI language models and more
If you need assistance in this, please join the Official KoboldAI Discord: https://discord.gg/4ZyUnR7D
It would be great if there were a separate feature for an event book, which has the same function as the Lorebook.
For Simulation Bot, using Probability Condition of the Lorebook, it is possible to convert various values into HTML GUI and control events.
To trigger meaningful events in chat, inevitably, the Lorebook should be located to Formationg Order at the bottom.
The problem is that this can overwrite other information such as Author's Note or other chats when loading information using keywords from other items in the Lorebook.
Therefore, it becomes impossible to use the Lorebook for its original purpose.
It would be good to have an Event Book specifically for controlling events.
Lorebook과 완전히 같은 기능을 하는 Event book이 별개로 추가되었으면 좋겠습니다
시뮬레이션 봇의 경우 로어북의 Probability Condition을 활용해 각종 수치를 HTML로 GUI로 변환해 표시하던가, 이벤트를 제어할 수 있는데
채팅에 유의미한 트리거를 일으키려면 필연적으로 Lorebook을 Formationg Order 가장 밑으로 내려야 합니다.
문제는 이 경우 로어북의 다른 항목의 키워드로 정보를 불러왔을 때 Author's Note나 다른 Chat들을 덮어버리는 경우가 생깁니다.
그렇기 때문에 어쩔 수 없이 로어북의 원래 용도로 쓰는 것이 불가능해집니다
이벤트 제어만을 위한 Lorebook, 즉 Event book이 있었으면 좋겠습니다.
Finally it works with Obabooga, thanks for hark work@kwaroran.
But when I talked with Yizu no matter what I sends she only replay "Yuzu, the shy cat girl maid"(seems like a default placeholder reply message). And Rika replays nothing at all. Obabooga backend shows no matter what I send it all take 80 tokens(contents numbers are different).
And I loaded different models in Obaboooga, tried default and notebook mode in setting. In available extensions section I toggled 'api' and 'gallery', in Boolean command-line flags section I toggled 'mode-menu', 'no-stream', 'pin-weight'. I'm using the latest version of Obabooga.
I know that due to a recent issue, the setting for plain fetch has been removed from RISU AI. However, this has the potential issue of disabling some proxies that enforce the use of plain fetch. Please allow plain fetch to be used, at least for local.
The money goes out of the API key, but no spaces related to supermemory appear on the character card - is this the result of supermemory being applied?
This is a issue when using web-Risu with PC browser (Chrome)
Making new group chat on web-Risu needs a refresh to actually see the new group.
today I got home and decided that I want to talk to the ai on risu again. I usually do it on my phone but since it had low battery I did it on my laptop. I logged in and it didn't show my characters even tho I logged in. After that I went on my phone and entered the site and my characters disappeared from there too. I don't know what I should do because I was quite invested in that chat. I tried to import the character but couldn't find it. Please help me
I posted a request for another feature a while back, but I think you may have forgotten about it.
I forgot about it too, but another user mentioned it and I remembered.
A font change feature would be nice!
그때 원형 프로필 요청하면서 폰트 바꾸는 옵션도 같이 요청했었는데
아마 폰트는 까먹으신 거 같아서 다시 글 올립니다. 저도 까먹었었는데 딴 분이 언급하셔서 기억났네요....🙄
RisuAI is the best AI chat frontend project of my life.
Many users have difficulty finding the delete button for bot cards in RisuAI.
My suggestion is to move the Delete Bot Card button to a more direct location, as shown in the image above.
Thank you very much as always.
(RisuAI는 제 인생 최고의 AI 채팅 프론트엔드 프로젝트입니다.
많은 사용자들이 RisuAI에서 봇 카드의 삭제 버튼을 찾는 것을 어려워 합니다.
저는 이에 대해서 위 이미지와 같이, 봇 카드 삭제 버튼을 더 직관적으로 이동하는 것을 제안합니다.
당신의 노고에 항상 감사합니다.)
If I enable that option and click on the first message, it will enter edit mode and will not be able to come out again
When I press Create Character, nothing happens. I see a very small space in the bottom left corner, but it doesn't respond when I press it.
I am currently using Windows 10 and my browser is Chrome. I tried to chat using the official cloud api key, but the following error popped up.
"<!doctype html><meta charset="utf-8"><meta name=viewport content="width=device-width, initial-scale=1"><title>403</title>403 Forbidden<script defer src="https://static.cloudflareinsights.com/beacon.min.js/v52afc6f149f6479b8c77fa569edb01181681764108816\" integrity="sha512-jGCTpDpBAYDGNYR5ztKt4BQPGef1P0giN6ZGVUi835kFF88FOmmn8jBQWNgrNd8g/Yu421NdgWhwQoaOPFflDw==" data-cf-beacon='{"rayId":"7db4ec23aaf2b45e","version":"2023.4.0","r":1,"token":"c622a0c7e1cc41b9b5db4375b44ee190","si":100}' crossorigin="anonymous"></script>\n"
I wondered if it was due to chrome records or extensions, so I tried turning off all extensions in incognito mode, but the same error occurred. It doesn't seem to have anything to do with vpn. Firefox works fine, but there seems to be a compatibility problem with Chrome.
Hello, Developer. Thank you for your hard work.
To begin with, I would like you to add the GPT-3.5-turbo-0301 button to Risu AI's model selection list. As you know, Open AI announced that it will automatically upgrade the GPT-3.5-turbo-0301 model to the GPT-3.5-turbo-0613 model on June 27th.
And they explained that the GPT-3.5-turbo-0301 model will be retired on September 13th. And according to the contents of Open AI's mail and homepage, if a user wants to use the GPT-3.5-turbo-0301 model even after the update, the user specifies the GPT-3.5-turbo-0301 model in the parameter of the API request and can use the model for about 3 months.
After reading these things, I thought it would be nice to add the GPT-3.5-turbo-0301 button to Risu AI's model selection list separately, just like the GPT-3.5-turbo-16k and GPT-3.5-turbo-0613 buttons. This is because there may be users who want to use the GPT-3.5-turbo-0301 model by the expiration date of September 13th.
I'm not familiar with programming stuff like this, so this might be a very silly suggestion. Although this suggestion is foolish, please consider it...
I want to use my CLAUDE API key as the main model and GPT 3.5 proxy as the secondary model, but when I enter my CLAUDE API key in the API key field, my key is leaked to the secondary model proxy. If possible, could you please add a toggle to turn on/off the protection so that the API key is not leaked to the proxy?
Hey there! I have a suggestion that could greatly benefit your UI. How about creating a Discord server where people can provide feedback and share their thoughts? It would be a fantastic way to engage with users who may not have a GitHub account. Discord is widely used, and many individuals prefer to discuss their issues and concerns there. By establishing a Discord server, you would not only address this common problem but also foster a sense of community around your UI. This, in turn, would contribute to increased popularity, as other UIs with Discord servers have successfully achieved. I hope you find this idea helpful!
Hello, I am here to request for OpenAI GPT4 32k model selection which Silly Tavern has but not on Risu.
currently, on firefox risu stop after user name input.
showing
Loading Sample bots...
in window
Uncaught (in promise) DOMException: The quota has been exceeded.
in console
Greetings again,
Sorry for hijacking another persons post! I like your interface a lot! I was trying to get it to work with oobabooga locally, but was having problems. I tried trouble shooting by myself over the past couple days. I wonder if there is a guide to get the whole thing to run locally with oobabooga and a Llamamodel? One is sending HTML and the other is waiting on a JSON that never comes.
I'm pretty new to this stuff and I'm learning a lot!
I found this alternative via a youtube video and was interested in it. However I am not very knowledgeable when it comes to this stuff, so maybe a user-friendly installation instructions would be great! There's little to no information on how to go about this, but I am excited about it.
There is only support for the "Text Generation Webui" which is pretty cool but many people also use KoboldAI. I use this version for example (because it supports 4bit): https://docs.alpindale.dev/local-installation-(gpu)/koboldai4bit/
Your UI looks soooo good and I really appreciate your efforts you did there!
I'm getting this kind of backup error and I've been trying to figure out what's going on, but it seems that the BGM in the asset part of the character card is causing this error. It seems that the BGM in the asset is not backed up properly.
Also, certain emote bot images or bots that have assets added are getting the error and not backing up properly.
Most of the backup loading errors seem to be caused by character cards that contain assets and emote images. In my tests, about 500 character cards without emotes and assets seem to be backed up without errors, and the ones with emotes and assets are the ones that cause backup errors.
Also, images in the emote bot are sometimes not exported character card because they are large.
And when I tested again, I turned off the image compression feature in the advanced settings, and some of the errors in the backup import disappeared. There must be something wrong with the image compression feature.
I think it would be nice to have a way to link files directly to Google Drive in real time like Silly Tavern does with Google Colab, so you don't have to take the time to upload or load backups, just link to Google Drive and you're good to go.
[Below is Korean.]
이런식의 백업 오류가 있는데 제가 직접 뭐가 문제인가 찾아봤는데 캐릭터 카드안에 있는것중에 에셋 부분에 bgm이 이러한 오류를 발생시키는거 같습니다. 에셋에 있는 bgm을 제대로 백업이 안되나 봅니다.
그리고 특정 감정봇 이미지나 에셋이 추가된 봇중에 오류가뜨며 제대로 백업이 안불러와집니다.
대부분의 백업 불러오기 오류는 에셋 및 감정봇 이미지가 들어간 캐릭터 카드들이 일으키는거 같습니다. 테스트를 해봤을때 감정이미지와 에셋이 들어가지않은 500개 가량의 캐릭터카드는 오류없이 백업이 되는거 같으며 오류가 일어나는 캐릭터 카드는 감정이미지와 에셋이 포함된 것들이 그렇습니다.
그리고 감정봇에 있는 이미지가 용량이 커서인지 내보내기가 안되기도 합니다.
그리고 테스트를 또 해봤을때 고급설정에 이미지 압축 기능을 Off 했을때 백업불러오기에서 몇몇 있던 오류가 사라지기도 했습니다. 이미지 압축 기능에 무언가 문제가 있나 봅니다.
차라리 Google Colab으로 돌리는 Silly Tavern처럼 구글드라이브에 직접 실시간으로 파일들을 연동하는 식으로 하면 좋지않을까 싶습니다. 그렇게하면 백업을 시간들여 올리거나 불러올필요 없이 구글드라이브 연동만하면 바로바로 사용이 되는거죠.
AI chats are built by a diverse group of people from around the world. RisuAI's community is built by people all over the world.
AI chat is a community and a hobby that is accessible to people with low vision and blindness using screen readers, especially since it is almost entirely written and red.
Therefore, I would like to make the following suggestions for improving RisuAI's accessibility features.
[ Alternative text is not provided for several features in RisuAI ]
For more details and Korean descriptions, see the links below.
Please help the community of RisuAI get more people, and more various people, involved.
Thank you for your service as always.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.