Hi! If the API is giving you short responses and seems limited, here are the most common reasons and fixes:
-
Jailbreaks/Custom Prompts: Make sure you are NOT using any jailbreaks, custom instructions, or "special prompts" in your settings. Our models are unfiltered and work best with a blank prompt (just enter a single space in that field). Custom instructions can severely lower response quality and length.
-
Model/Website:
- Some sites, characters, or bots may restrict output or use short prompt templates. Try a different character or website to see if it improves.
- For quality and full-length responses, always use the recommended OpenAI default generation settings, and set the temperature around
0.85. You can also check the Models Channel for specific model settings.
-
Conversation History: Long, repeated, or short messages in your chat history can influence model output length. Start a new chat and try again.
-
Which Site Are You Using?:
- JanitorAI: Be sure youβre using the FULL endpoint URL, e.g.,
https://api.pawan.krd/cosmosrp-2.5/v1/chat/completions, and not just the base. - VenusChatAI/ChubAI/SillyTavern: Use the base URL only, e.g.,
https://api.pawan.krd/cosmosrp-2.5/v1 - For setup guides, see Setup Guides Channel.
- JanitorAI: Be sure youβre using the FULL endpoint URL, e.g.,