The response is cutting off

[teal stratus]

I'm new to this but Is there any way to stop the response from cutting off shortly before the first sentence?

[dark lily]

Welcome! If your response is getting cut off (e.g., stopping mid-sentence or just after starting), here are some things to check and try:

1. Model Context Size
   Each model has a token/context limit. If you’re using a free plan, CosmosRP-V2.1 (cosmosrp) allows up to 16,384 tokens per conversation. If your prompt plus expected output is longer, the reply might get clipped.

2. Website Limitation
   Sometimes, the website you’re using (like VenusChat, JanitorAI, etc.) has its own output or token limits that may cause cut-off responses. This is controlled by the site, not the API.

3. Generation Settings
   Try setting the “max tokens” (or similar) setting higher, if your site allows it. This will let the API generate longer results.

4. Temperature Setting
   Make sure you’re using the recommended temperature for the model (for CosmosRP: around 0.85 by default). Too low or too high can sometimes cause odd output lengths.

5. Jailbreak/Custom Prompts
   If you’ve used any special instructions, jailbreaks, or custom prompts, remove them—they can affect generation quality and length.

6. Refresh & New Chat
   If issues persist, start a new chat/session and try again. Websites sometimes cache buggy chat states.

7. Website Bug
   If it always cuts off early (even basic replies), it might be a bug with the website you’re using. Try another supported site or interface.

[dark lily]

If you still experience this issue only on blitzpoint.io, it may be a limitation or bug with that specific website. In that case, please contact blitzpoint.io’s developer for support, as site-specific problems are outside the reach of our API.

For more guidance about using our API, check the Models Channel. If you need help setting up your API elsewhere, let me know which platform you’re using!

Please note: This is an automated response from @dark lily. While it strive to provide accurate assistance, It may occasionally make mistakes. If you find any inaccuracies or need further clarification, feel free to wait for our community helpers to give you further guidance. Thank you.

[molten swift]

Haha, it thinks your name is a website. Anyway, whst are your max new tokens set to? And what website are you using?

[teal stratus]

I'm using janitor and I have my max tokens set to unlimited.

[molten swift]

You have to manually set the tokens, so set them to 1k

Setting it to 0 makes cosmos give itself a manual limit, which causes responses to cut off

[teal stratus]

Okay

Yeah it's still cutting off.

[molten swift]

Refresh and try again, or clear cache and try. I forgot Janitor isnt well built so it doesnt exactly... save settings correctly if at all

[teal stratus]

I did all of that.

[molten swift]

Would that include a new chat as well?

Can you send your api settings and gen settings? Ive been seeing more and more issues with janitor, so it could be the website,

[teal stratus]

It was working fine yesterday

It's working completely fine with a different bot though.

[molten swift]

Then it could just be a website hiccup. I would recommend lowering your context size to 15.5k, if you have it at the max for the model, janitor will give you later on in the roleplay

I am going to delete the images since your full key is in it, but ill repost them here

[teal stratus]

Okay

[south vapor]

Oh! I had this problem too

I keep my messages on 0 token limit

But my context on 16384