Is your feature request related to a problem? Please describe.
Scale is still a functional API and is useful in many contexts. Would be nice to have official support like in other forks, especially if they walk their content moderation back. I can cobble it in myself but would prefer not to brute-force it.

Is your feature request related to a problem? Please describe.
When using the sd_api_picture in Oobabooga, the chatbot can generate images but SillyTavern show the html code instead.
When receiving a <img src=...>, it put it inside double quote, preventing the image from showing up.
Removing the quotes manually allow to see the image correctly, meaning the issue is on the formatting itself.

Describe the solution you'd like
In case you receive a <img src=...> from your connected API, it should not be put between quotes.

Describe alternatives you've considered
//

Additional context

Is your feature request related to a problem? Please describe.
3.5 turbo works like trash right now. It sends the error below. Which requires repeated clicks on the send button to get through

Describe the solution you'd like
Checkbox that makes client send request in a loop if it gets the error about an overloaded server

Describe alternatives you've considered
Manually clicking send button

Additional context
error: {
message: 'That model is currently overloaded with other requests. You can retry your request, or contact us through our help center at help.openai.com if the error persists. (Please include the request ID %ID% in your message.)',
type: 'server_error',
param: null,
code: null
}

Is your feature request related to a problem? Please describe.
Sometimes there are interesting moments in the chats that I would like to play out in different ways, i.e bot responds interestingly in the message history or I get a good swipe and I want to save it for the reuse. For this I copy the chat file and then use a copy. But this is tedious, because you have to go into the tavern folder, click publc, chats, {{char}} folder, search for the latest file, then copy it and rename the copy.

Describe the solution you'd like
Chat cloning functionality in the options menu.

Describe alternatives you've considered
Manually copying chat files.

Some users, including myself, were able to obtain access to the Anthropic Claude API
I would then like to request that you add a way to use it with tavern, since some users prefer using it rather than agnai.

*Bug Description
I have noticed that certain web extensions, such as anti-banner and ad blockers, are causing the API connections icon to become disabled.

*Reproduce the Bug

1. Enable the anti-banner feature of the Kaspersky Chrome extension.
2. Visit the 127.0.0.1:8000 or localhost:8000

*image

I solved this easily by exception handling localhost from the anti-banner feature.

I suspect that this issue may be caused by the hyperlinks inside the drawer.

Thanks as always. 👍

MathJax support in tavern would be an amazing addition. I think it's really important to have a reliable way of rendering math equations and symbols on the page.

https://github.com/mathjax/MathJax-node
For reference.

AI Horde is a crowdsourced cluster of workers for text and image generation that can be used anonymously and/or free. It provides a fully documented REST API you can easily integrate to your front-end. It can also generate images for further immersion in the chat. It can also handle the captioning and other image interrogation

I am the creator, so feel free to hit me up on questions on discord.

Is your feature request related to a problem? Please describe.

• I realized [Author's Note / Character Bias] are sometimes saved and sometimes not... not sure this is intended

Describe the solution you'd like

• It would be nice if we can have a save button for [Author's Note / Character Bias] to let us save it when needed

Describe alternatives you've considered

• Or... maybe such a function can be a normal function like the world info main prompt.

I don't know if I'm missing something or it is intended, but currently placeholders({{char}}/{{user}}) are not working in the Author's Note.

This also happens with the NSFW/jailbreak prompt, when 'Jailbreak as system message' checked.

Also, if someone use the A/N, it seems like they often place it after the user's last chat, reason below. Please consider changing the default setting to that(In-chat, every 1 message, depth 0). Availability to create two A/Ns to use After Scenario and In-chat simultaneously would be fine addition, too.

Personally, I'd like to suggest feature somewhat similar to A/N but globally applies to all chats. This can be accomplished to some extent by checking Constant in World Info, but it's not really the same because I can't enhance its impact by making it placed after the user's last chat.

Thank you for your consideration!

Describe the bug
SillyTavern correctly boots up, but the characters refuse to respond.

To Reproduce
Steps to reproduce the behavior:

1. Install oobabooga 1-click-install text generation webui (and nodejs and sillytavern)
2. edit "start-webui.bat" to "call python server.py --load-in-8bit --model pygmalion-6b --no-stream --extensions api"
3. Launch oobabooga, if you test it by itself (against the default notebook bot), it works normally
4. Try using SillyTavern, no response

Expected behavior
It should reply

Desktop (please complete the following information):

• OS/Device: Windows 10
• Environment: local
• Browser: Brave (Chromium-based)
• Generation API: Oobabooga
• Branch: main
• Model: Pygmalion 6b (default release, not dev nor first build)

Additional context
Error log on oobabooga side:

127.0.0.1 - - [11/Apr/2023 22:14:00] "GET /api/v1/model HTTP/1.1" 200 - 127.0.0.1 - - [11/Apr/2023 22:14:00] code 404, message Not Found 127.0.0.1 - - [11/Apr/2023 22:14:00] "GET /api/v1/config/soft_prompts_list HTTP/1.1" 404 - Traceback (most recent call last): File "C:\Program Files\oobabooga-windows\installer_files\env\lib\site-packages\gradio\routes.py", line 393, in run_predict output = await app.get_blocks().process_api( File "C:\Program Files\oobabooga-windows\installer_files\env\lib\site-packages\gradio\blocks.py", line 1108, in process_api result = await self.call_function( File "C:\Program Files\oobabooga-windows\installer_files\env\lib\site-packages\gradio\blocks.py", line 929, in call_function prediction = await anyio.to_thread.run_sync( File "C:\Program Files\oobabooga-windows\installer_files\env\lib\site-packages\anyio\to_thread.py", line 31, in run_sync return await get_asynclib().run_sync_in_worker_thread( File "C:\Program Files\oobabooga-windows\installer_files\env\lib\site-packages\anyio_backends_asyncio.py", line 937, in run_sync_in_worker_thread return await future File "C:\Program Files\oobabooga-windows\installer_files\env\lib\site-packages\anyio_backends_asyncio.py", line 867, in run result = context.run(func, *args) File "C:\Program Files\oobabooga-windows\installer_files\env\lib\site-packages\gradio\utils.py", line 490, in async_iteration return next(iterator) File "C:\Program Files\oobabooga-windows\text-generation-webui\modules\api.py", line 28, in generate_reply_wrapper for i in generate_reply(params[0], generate_params): File "C:\Program Files\oobabooga-windows\text-generation-webui\modules\text_generation.py", line 167, in generate_reply input_ids = encode(question, generate_state['max_new_tokens'], add_bos_token=generate_state['add_bos_token']) KeyError: 'add_bos_token'

Error log on SillyTavern side:

{ error: null }

Another notable thing is that previously I ran oobabooga using the same settings on WSL2, and SillyTavern worked fine, but now I tried switching to Windows 10 running directly and it broke.

Is your feature request related to a problem? Please describe.
Presets have a strong effect on a character. And strongly influence role play. I ask to add functionality which remembers prompts and adds them to list, so I won't copypaste text manually everytime I want a different roleplay narration. So, for example, I could create preset for vivid, verbose NSFW, preset where character initiates NSFW, preset where I just want SFW slice of life, etc. Also I would like if you add a checkbox (or what you think would work best), which links selected preset to a opened character.

Describe the solution you'd like
Ability to link a character to a preset. Ability to open, edit, add and delete presets. Presets contain nsfw prompt and main prompt.

Describe alternatives you've considered
Manually copypasting prompts

Additional context

Used this service: https://tiktokenizer.vercel.app.
인생은 아름답고 멋지다
Жизнь прекрасна и удивительна

Jailbreak prompt Redoing this since SillyLossy said OAI it's in the dev branch.
Someone from /aids/ made this jailbreak prompt.

Need testing to see its efficiency but with the right tweaking it can work great. If not, a toggle for this would be great too.

Describe the bug
Invalid/Expired Token warning when using Poe's cookies when when there are custom bots

To Reproduce
Steps to reproduce the behavior:

1. Log in to Poe
2. Make a custom bot
3. Boot up SillyTavern
4. Get Poe's p-b cookie value
5. Insert Poe's p-b cookie value in SillyTavern
6. Error Pop-up

Expected behavior
It should work regardless if the user has a custom bot or not.

Desktop (please complete the following information):

• OS/Device: Windows 11
• Environment: tested locally, untested on cloud
• Browser: Firefox
• Generation API: Poe
• Branch: main
• Model: none

Additional context
Works normally on Poe accounts without custom bots. SillyTavern seems to throw this error when it tried to download the custom bot after ChatGPT.

Is your feature request related to a problem? Please describe.
There is no simple way inside the GUI of identifying when a past chat message is truncated out of the prompt because of the user/model context token length restriction. This would be useful to visualize chat length that the bot is aware of, and to help manually manage bot knowledge, memories, story context, etc.

Describe the solution you'd like
I would like a checkbox inside the power users section of the user tab with the option to grey out a box around messages outside of context. One way it could be done is storing the data locally in a file, with char and a number of messages that fit in the last prompt. Ideally it would update when user changes the context size slider but that may trigger a lot of recalculations at once.

Describe alternatives you've considered
You can scroll up in the terminal and see what the last message sent before permanent block is but it can be a lot of scrolling, also you have to send a message to see, there is no way that I know of to do it for a character after restarting sillytavern without sending a message.

Additional context

If I let's say have tokens set at 200, why does it fill the entire 200 tokens instead of cutting the reply when it finishes writing an answer? There is an option for single line mode in both koboldcpp and sillytavern, but it doesn't stop generation like it does in ooba webui, it only stops them from appearing in the ui, so the wait times are long even if the returned reply is like 10 words. And if I set the token count too low, then the replies will have no potential to be longer

Is there a way to stop it from having a conversation with itself to fill the entire token limit?

Also, what is the processing prompt BLAS thing that takes 2-3 minutes to go through frequently? And is there a way to enable text streaming?

Thanks!

Describe the bug

Not sure if it is a bug but definitely it is something. I do run KoboldAI locally and API from KoboldAI works fine on my local TavernAI but it has issues on SillyTavern

To Reproduce
Run KoboldAI locally and try to use API from it on SillyTavern.

Expected behavior
As TavernAI have no problem using local API so SillyTavern should not either?

Screenshots

Desktop (please complete the following information):

• OS/Device: Windows 10
• Environment: local
• Browser Opera
• Generation API: KoboldAI
• Branch: main
• Model: Pygmalion 6b

Anyways, still. Thank you for amazing job. For bringing to live TavernAI :)

Cannot scroll on Android Chrome
(Google Pixel 6, Android 13)
Using Termux (Open AI API), everything works fine except scrolling up and down.
Thanks.

Describe the bug

When Always add character's name to prompt is disabled, the character's name is still prepended in the prompt.

To Reproduce

Steps to reproduce the behavior:

1. Disable Always add character's name to prompt
2. Generate
3. View context in terminal

Expected behavior

The character's name should not be prepended in the prompt.

Screenshots

Character's name is DEBUG: and Always add character's name to prompt was disabled:

Desktop (please complete the following information):

• OS/Device: Windows 10
• Environment: Local
• Browser: Chrome
• Generation API: KoboldAI
• Branch: Dev
• Model: Pygmalion 6B

Describe the bug
Using the Colab, no characters I upload will appear in the character list. I've tried importing it from both card.png and .json files, as well as placing the files in the MyDrive/TavernAI/characters folder outlined as the character directory by your code. Adding the character by manually filling it out also doesn't work, resulting in no saving occuring. My Drive is mounted, and the default three characters show up no matter what.

To Reproduce
Steps to reproduce the behavior:

1. Go to the Colab linked in the Readme on this repo.
2. Make copy in your Google Drive (I did this in order to save a custom model entry so I wouldn't have to redo it every time.)
3. Load a model. I loaded a custom model from huggingface, not any of the default ones, if that matters.
4. Wait for cell to run, get link.
5. Connect to OAI API using key, success. GPT 3.5 AND 4 result in the issue, version does not matter.
6. Go to character tab
7. Attempt to upload character from computer
8. Tavern pauses, then acts as if nothing has happened
   9.Check for characters, only Megumin, Aqua and Darkness are there, still.
9. Open the plus tab for new characters
10. Fill out all sections
11. Hit save, only for it to not.
12. Disconnect and delete runtime.
13. Put .png of character card in MyDrive/TavernAI/Characters
14. Relaunch, wait for it to run, click the link
15. Check characters, only to still only have the default three.

Expected behavior
I expected my character to upload in at least one of those instances, especially since it's in the folder with the rest of them, and in the same format.

Screenshots
If needed, I can, but it's less of a screenshotable issue, since it doesn't provide error codes or obviously break. It's more of a "nothing happens at all" error.

Desktop (please complete the following information):

• OS/Device: Linux Mint/HP Pavillion 15a
• Environment: Cloud, Colab
• Browser: Firefox
• Generation: API OpenAI
• Branch: Unsure. Probably main. I didn't specify on my Vicuna model, could that be a source of the problem?
• Model: vicuna-13b-4bit-128g, tacked it on at the end in the code since Pygmalion threatens banning and there's no way to NOT load a model implemented.

Additional context
Let me know if you'd like the stack.

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

It's difficult to make the desired character speak in the next conversation. Currently, we proceed with the conversation by excluding all other characters in the group, so the desired character can speak. However, as the number of characters increases, this becomes difficult.

Describe the solution you'd like
A clear and concise description of what you want to happen.

I would appreciate it if you could add a feature to the group chat that allows us to choose the character to speak next directly.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

I believe a feature in "Character Management" is necessary to set the order of the conversation. (I understand that in group chats, the character at the top of the list speaks first. It would also be helpful if we could change the character's order by dragging, or set the order using numbers. since there are cases where we want only one specific person to speak, it would be great if we could set the number of people speaking next.)

Additional context

"This translation has been done by Chat GPT-4."

Edit: So I just realized that Firefox is showing 100% GPU usage the second I submit my prompt. Tabbing out or minimizing the browser immediately drops the usage back down. Disabling hardware acceleration completely fixes the issue, but I have blur and chat bubbles disabled, so it's still odd it would be using this many resources.

Describe the bug
I have a bit of a strange issue going on. Output generation is extremely slow when I have the chat window open. I ran some tests by re-generating the same reply on the same seed. The difference in generation time between the window being visible or not ranges from ~5.5 to 6.5 seconds for a new chat and ~13 to 60+ seconds for a long-running one. This was tested with Firefox. I tried Edge as well. Strangely, the difference there is less significant but still noticeable.

Some generation times in seconds tracked with WSplit:

Fresh chat, first reply
Window Open
6.50 6.49 6.59 6.51
Tabbing Out
5.40 5.66 5:63 5:45

Long chat, 46th reply
Window Open
120.00+ (gave up) 120.00+ (gave up)
Tabbing Out
13.01 13.20

To Reproduce
Steps to reproduce the behavior:

1. Launch SillyTavern, open with Firefox, and connect to Oogabooga
2. Pick a bot and generate a response
3. Set to a fixed seed. Regenerate the response. Track generation time.
4. Click regenerate again and immediately minimize the browser
5. Check generation time difference

Expected behavior
Generation time with the window open is the same as minimized.

Desktop (please complete the following information):

• OS/Device: [e.g. Windows 11]
• Environment: [cloud, local]
• Browser [e.g. chrome, safari]
• Generation API [e.g. KoboldAI, OpenAI]
• Branch [main, dev]
• Model [e.g. Pygmalion 6b, LLaMa 13b]

Additional context

• Just being out of focus isn't enough. I must switch to a different tab or minimize the browser to fix the generation speed
• I am using this connected locally to Oogabooga launching with the following parameters: --wbits 4 --groupsize 128 --auto-devices --notebook --auto-devices --no-stream --gpu-memory 9536MiB
• Using an 11gb 1080TI on Windows 10
• I noticed the issue when replies would take 4+ minutes to generate. Eventually I realized that I'd always hear the notification sound 10-15 seconds after changing tabs.
• Oogabooga's "Output generated in x seconds" log is not accurate and a bad metric for time tracking.
• Using gpt-x-alpaca-13b-native-4bit-128g-cuda.pt

Describe the bug
Crash on startup. The webpage loads, but the server crashes immediately.

To Reproduce
Steps to reproduce the behavior:

1. Run Start.bat
2. Server crashes immediately after opening the webpage.

Expected behavior
Start the server without any issues.

Log

E:\ProgramsE\TavernAI-SillyLossy Mod>call npm install

up to date, audited 194 packages in 659ms

22 packages are looking for funding
  run `npm fund` for details

7 high severity vulnerabilities

To address all issues possible (including breaking changes), run:
  npm audit fix --force

Some issues need review, and may require choosing
a different dependency.

Run `npm audit` for details.
Launching...
TavernAI started: http://127.0.0.1:8000
E:\ProgramsE\TavernAI-SillyLossy Mod\server.js:680
    var base64DecodedData = Buffer.from(textChunks[0].text, 'base64').toString('utf8');
                                                      ^

TypeError: Cannot read properties of undefined (reading 'text')
    at charaRead (E:\ProgramsE\TavernAI-SillyLossy Mod\server.js:680:55)
    at E:\ProgramsE\TavernAI-SillyLossy Mod\server.js:700:28
    at Array.forEach (<anonymous>)
    at E:\ProgramsE\TavernAI-SillyLossy Mod\server.js:698:18
    at FSReqCallback.oncomplete (node:fs:198:23)

Node.js v18.15.0
Press any key to continue . . .

Desktop (please complete the following information):

• OS/Device: Windows 10
• Environment: local
• Browser: Firefox
• Generation API: KoboldAI
• Branch: main
• Model: KoboldAI_OPT-13B-Nerybus-Mix

Additional context
It has been starting normally last week. It only started crashing when I started it today, I guess because it updated itself to the latest version. It said that there was an update to NPM and it updated itself.

Is your feature request related to a problem? Please describe.
I liked the streaming implemented in TAI-Turbo since I could see the reply coming in which felt natural and intuitive.

Describe the solution you'd like
Snooping around the code reveals you were probably trying to get it working in your fork at some point. The solution would be to have that happen. (i wanted to look into it to maybe make a PR but I know jack about JS)

Describe alternatives you've considered
I tend to still use TAI-Turbo just out of liking that feature....

Additional context
(I was curious about the OAI model being used but seems you can change it just fine now.)

[BUG] Poe.com is broken in main branch

Switch to dev branch if you use that. Thanks.

I know you can drag and drop your characters, chat logs and settings by copying the contents of the public folder from your old version to your new one but it would be nice to not have to remote into the server to do that process and instead just do an import/export of all the data right from the website.

This might also tie in with just the ability to update it as well from the website but that may be a bit more involved since it's essentially keeping a shell that stays alive long enough to overwrite data files.

Describe the bug
When setting the OpenAI connection to GPT-4, then typing a chat, the console throws an error from OpenAI saying GPT-4 is not a valid model. The other model selections do not do this. I have an OpenAI subscription, and can manually choose GPT-4 on their website.

{
error: {
message: 'The model: gpt-4 does not exist',
type: 'invalid_request_error',
param: null,
code: 'model_not_found'
}
}

To Reproduce
Steps to reproduce the behavior:

1. Go to 'Settings'
2. Click on 'Model' and choose GPT-4
3. Close settings and start a chat
4. When there is no reply, look at the console window and see the error

Expected behavior
A response.

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS/Device: Windows 11
• Environment: Local
• Browser Edge or Chrome
• Generation API OpenAI
• Branch main
• Model GPT-4

Additional context

Implement colored tokenizer for making bots. I tried using this one (https://tiktokenizer.vercel.app) but it ignores some characters, line breaks at least.

Heard from someone that they're doing this, so I set everything up to try it out, but when I go on to connect the api the running tavern dies immediately. Any idea why that would be or if there are any logs anywhere that might give a hint

Updated server.js from the 13 hours ago edit and now I get this

An error occurred: Error parsing response. response: [Error: HTTP Server is running, but this endpoint does not exist. Please check the URL.], error: [SyntaxError: Unexpected token E in JSON at position 0]

koboldcpp is spamming this

127.0.0.1 - - [09/Apr/2023 00:45:10] "GET / HTTP/1.1" 200 -
127.0.0.1 - - [09/Apr/2023 00:45:10] "GET /api/latest/model HTTP/1.1" 200 -
127.0.0.1 - - [09/Apr/2023 00:45:11] "GET /sw.js HTTP/1.1" 404 -
127.0.0.1 - - [09/Apr/2023 00:45:11] "GET /manifest.json HTTP/1.1" 404 -
127.0.0.1 - - [09/Apr/2023 00:45:28] "GET /api/v1/model HTTP/1.1" 200 -
127.0.0.1 - - [09/Apr/2023 00:45:28] "GET /api/v1/config/soft_prompts_list HTTP/1.1" 404 -
127.0.0.1 - - [09/Apr/2023 00:49:02] "GET /api/v1/model HTTP/1.1" 200 -
127.0.0.1 - - [09/Apr/2023 00:49:02] "GET /api/v1/config/soft_prompts_list HTTP/1.1" 404 -
127.0.0.1 - - [09/Apr/2023 00:52:04] "GET /api/v1/model HTTP/1.1" 200 -
127.0.0.1 - - [09/Apr/2023 00:52:04] "GET /api/v1/config/soft_prompts_list HTTP/1.1" 404 -
127.0.0.1 - - [09/Apr/2023 00:52:07] "GET /api/v1/model HTTP/1.1" 200 -
127.0.0.1 - - [09/Apr/2023 00:52:10] "GET /api/v1/model HTTP/1.1" 200 -
127.0.0.1 - - [09/Apr/2023 00:52:13] "GET /api/v1/model HTTP/1.1" 200 -
127.0.0.1 - - [09/Apr/2023 00:52:16] "GET /api/v1/model HTTP/1.1" 200 -
127.0.0.1 - - [09/Apr/2023 00:52:19] "GET /api/v1/model HTTP/1.1" 200 -
127.0.0.1 - - [09/Apr/2023 00:52:22] "GET /api/v1/model HTTP/1.1" 200 -

Nvm it's running even with the error message.

Is your feature request related to a problem? Please describe.

Entries added to the top of context have less weight than entries towards the bottom of the context. If you have context size set to 2048, then the character card has little weight after awhile and this also means world info entries also have little weight, as well.

Describe the solution you'd like

Adding insertion depth to the various context boxes/character card/world info entires would provide greater control. This would also allow the characters to stay 'in character' even when using a context size of 2048, as it would allow the character card to sit lower in the context. This also improves short-term memory, as the conversation would be allowed to "flow" over the character card. Right now, you would limit context size on purpose to increase the strength of your character card, which is unideal as you lose short-term memory.

Describe alternatives you've considered

• Insertion Depth could be just one setting that's applied to an entire character card itself (as like a big block). So there'd be only 1 insertion depth box for it. Then WI with "After Chara" or "Before Chara" would automatically have stronger effect depending.
• Insertion Depth could be just added to WI, then people could use WI as a means to create their characters to have the same effect.

There could be other solutions, but this is the first one that comes to mind! Thank you!

Currently Description, Impersonation, Jailbreak, NSFW, Main, First message textarea's are too small for comfortable use (in my experience). Especially Description one.
Describe the solution you'd like
If its possible I'd like to have them resizable. If not, increase the default size for the description field at least, since it doesn't really affect anything negatively.

You said to create a request here so it doesn't get lost in the reddit backlog

The Poe integration seems to be broken. I get this error in the console:
"OPTIONS /api/poe/status HTTP/1.1" 404 -

Is your feature request related to a problem? Please describe.
Prior to commit 0c8f068, the character menu was allowed to overlap the message area like so:

Commit 0c8f068 changed it so that this menu will not overlap the text area. Furthermore, the presets menu us now similarly locked to the margin left of the text area.

The problem I have is that now these two menus get squished and are very annoying to use because they are not allowed to overlap the message area anymore. See below for my experience (locks were activated just for taking the screenshot).

Describe the solution you'd like
Please allow these menus to overlap like before prior to 0c8f068

Describe alternatives you've considered
I can zoom-out my browser to increase the size of the menus, but this reduces the size of the font for the whole UI and makes things hard to read. My eyes are not very good, and I typically need 120% or more zoom.

The jailbreak as system message feature is great but having to use it together with the NSFW prompt makes the AI too sexually charged no matter what. Would it be possible to introduce a toggle so that only the UJB prompt is sent as the last system message? Thanks in advance.

Hi! Yesterday I was running sillytavern on my android normaly, but today I couldn't use it anymore. I'm using Termux as listed on the guide. Can you help me fix this issue?
~/TavernAI $ node server.js
node:events:490
throw er; // Unhandled 'error' event
^

Error: listen EADDRINUSE: address already in use 0.0.0.0:8000
at Server.setupListenHandle [as _listen2] (node:net:1740:16)
at listenInCluster (node:net:1788:12)
at doListen (node:net:1937:7)
at process.processTicksAndRejections (node:internal/process/task_queues:83:21)
Emitted 'error' event on Server instance at:
at emitErrorNT (node:net:1767:8)
at process.processTicksAndRejections (node:internal/process/task_queues:82:21) {
code: 'EADDRINUSE',
errno: -98,
syscall: 'listen',
address: '0.0.0.0',
port: 8000
}

Node.js v19.6.1

Describe the bug
when I try to use with oobabooga, the first time I type something
I get a #### { error: 'This app has no endpoint /api/textgen/.' } in the Silly Tavern console window.
the ui just stays on a spinning circle, but never generates anything

No debug output in the text-generation-webui window

To Reproduce
Steps to reproduce the behavior:

1. Attempt to connect to Text generation web ui
2. Attempt to chat with a bot

Expected behavior
To respond

Screenshots
I'm pretty sure before updating oobabooga to the latest version where it says None here, it had some type of information about what it connected to, but it was still having the same exact issue.

Desktop (please complete the following information):

• OS/Device: Windows 10
• Environment: local
• Browser: Brave (chromium)
• Generation API: text-generation-webui
• Branch main and dev (just updated dev to latest before writing this bug report
• Model: Pygmalion 6b

Additional context
I made sure to run with --no-stream, and to choose "notebook" under Interface Mode
though there is another mode that lets you choose cai-chat, chat, or instruct in the actual chat tab of the ui, and I just chose cai-chat
I tried updating text-generation-webui

Is your feature request related to a problem? Please describe.
During congestion times it can take up to 2 minutes for a response to come from OAI. And sometimes when it's a gateway timeout, it takes like 5 minutes. For people using local models, a generation can also be very slow. During this time, the chat symbol is spinning, waiting for the response. I usually alt-tab to a different window and check back occasionally.

It would be nice if Tavern played a sound to notify the user that there's new activity.

Describe the solution you'd like
Allow the user to play a file name notification.wav (or whatever format) and play it when a response arrives or fails. In Automatic's SD webui, if the user places a notification.mp3 in the root folder, it will play automatically when a generation completes.

Info on AI Horde: https://horde.koboldai.net

Hello, I appreciate all what you've done so far it's the best TavernAI fork I've found, I just wish I could use AI Horde with it.
I'm requesting it if possible for all those without suitable GPUs or find CPU mode too slow they can utilize the faster AI Horde.

A fork from 3 weeks ago has done so but I much prefer to use yours for its enhancements, features and extensions, much better.
https://github.com/Aspartame-e951/TavernAI/releases/tag/1.0

Thanks for any consideration.

Are the dice bugged atm? clicking the icon does nothing for me.

Amazing UI btw, Very impressive how easy it was to connect to OpenAI or a local oobabooga run model.

Describe the bug
I click on the <div id="options_button"> and I get error below. The menu for creating a new chat does not open
bookmarks.js:69

   Uncaught TypeError: characters[this_chid].chat.includes is not a function
at showBookmarksButtons (bookmarks.js:69:42)
at HTMLDivElement.<anonymous> (script.js:3811:13)
at HTMLDivElement.dispatch (jquery-3.5.1.min.js:2:43090)
at v.handle (jquery-3.5.1.min.js:2:41074)

To Reproduce
Steps to reproduce the behavior:

1. Go to the TavernAI
2. Load character
3. Click options button in the left bottom
4. See nothing

Expected behavior
Options menu opens, I can create new chat, regenerate, etc.

Screenshots
Nothing to show.

Desktop (please complete the following information):

• OS/Device: Windows 10
• Environment: local
• Browser: Edge, Google Chrome, Firefox
• Generation API: OpenAI
• Branch: tried latest main and dev
• Model: Turbo 0301

Additional context
I have ublock, hola vpn, dark reader installed on chrome.
This error happens with or without TavernAI-extensions.
I have many character cards. Possibly some of them are breaking your code. If I clone a clean repo, and add a few cards, I can't replicate this error.
I can't create new chats so it's hard to enjoy your Tavern 😢.

Describe the bug
When trying to connect to koboldcpp using the KoboldAI API, SillyTavern crashes/exits.

To Reproduce
Steps to reproduce the behavior:

1. Go to 'API Connections'
2. Enter API url: 'http://localhost:5001/api'
3. Click on 'Connect'
4. The SillyTavern console window of Node.js disappears unexpectedly, i. e. SillyTavern crashed/exited

Expected behavior
Connection is established and SillyTavern doesn't crash or exit.

Desktop (please complete the following information):

• OS/Device: Windows 11
• Environment: local
• Browser: Firefox
• Generation API: koboldcpp (KoboldAI)
• Branch: main (v1.2.3)
• Model: alpaca-7b-native-enhanced, ggml-vicuna-7b-4bit

Additional context
Regular TavernAI works with koboldcpp. No crashes/exits there.
The koboldcpp console log shows two entries:
"GET /api/v1/config/soft_prompts_list HTTP/1.1" 404 -
"GET /api/v1/model HTTP/1.1" 200 -
So SillyTavern is definitely able to access the endpoint.

Got this when running Tavern now.

Any attempt at dealing with this is either ineffective or makes the problem worse.
Any idea what it means? Could my IP and API keys be compromised?

Describe the bug
On iPad everything works, but on iPhone XS Max I am unable to scroll through chats in portrait mode (with or without full screen). In landscape on iPhone, scrolling works.

To Reproduce
Steps to reproduce the behavior:

1. Go to local IP on iPhone
2. Can browse menus and chats, but there is no scroll bar or ability to scroll through chat conversations while in portrait mode.

Expected behavior
Expect to be able to scroll normally in portrait on iPhone.

Screenshots
N/A

Desktop (please complete the following information):

• OS/Device: [iPhone XS Max, iOS 14]
• Environment: [local]
• Browser [Safari]
• Generation API [Text generation web UI]
• Branch [main]
• Model [Pygmalion 6b]

Additional context
Tried adding <meta name="viewport" content="width=device-width, initial-scale=1"> to index.html, but did not help.

Describe the bug
Disabling Author's Note, after it's previously been activated, doesn't properly disable it.

To Reproduce
Steps to reproduce the behavior:

1. Enter text into "Append the following text:" box.
2. "Every N messages you send" set to 1.
3. Generate
4. "Every N messages you send" set to 0.
5. Generate or Regenerate

Expected behavior
Setting "Every N messages you send" to 0 should disable it and the Author's Note should not be sent to the context.

Screenshots

Desktop (please complete the following information):

• OS/Device: Windows 10
• Environment: Local
• Browser: Chrome, Firefox, OperaGX (tested on all)
• Generation API: KoboldAI
• Branch: Dev
• Model: pygmalion-6b