Personas

Chat AI supports loading preset personas from configurations in the form of JSON files. Each JSON file includes the system prompt, settings, and conversations, allowing you to easily load a persona into Chat AI. While these files can be imported using the import function, Chat AI also supports directly importing public JSON files from the web, by specifying it in the URL.

Example personas

If all you need is a quick link to load a specific persona, this is your chapter. These are some of the interesting and useful personas the AI community came up with:

• Assistant Persona: useful for general-purpose queries
• German to English Translation: translates given German text to English without commentary
• English to German Translation: translates given English text to German without commentary
• Annoying Monster: An annoying monster that speaks in riddles.
• High Society Snob: A high society snob that will only respect you if you are royalty.
• 10-year-old Child: A ten-year-old child.

Using Chat AI Personas

We provide some recommended personas in our Chat AI Personas GitHub repository.

You can create a link to Chat AI with the desired persona from the publicly-available preset personas in Chat AI. To do this, simply add the URL of the JSON file in the import parameter of the URL:

https://chat-ai.academiccloud.de/chat?import=<json_url>

Replace <json_url> with the URL to the JSON file.

Custom Personas

You can also create your own custom personas to load directly in the Chat AI interface. These must be saved as a JSON file.

{
    "model-name": "Meta LLaMA 3.1 8B Instruct", # A short name for the model you wish to use
    "model": "meta-llama-3.1-8b-instruct", # API model name
    "temperature": 0.2, # Custom temperature settings
    "top_p": 0.2, # Custom top_p
    #This is where you can change your system prompt. Role: tells it what the role is in this case "system" other options are "user" and "assistant". Content is where you enter your custom prompt.
    "messages": [
      {
        "role": "system",
        "content": "<Enter your custom prompt here>"
      }
    ]
  }

The latest models are listed here. The API model name is listed here