Ziddny LogoZiddny Docs
Assistants

Adjust Behavior

Configure your assistant’s behavior, tone, and knowledge source

The Adjust Behavior page allows you to define how your assistant interacts with users.
Here, you can set its greeting, tone, rules, and the AI model behind it. Prompt Interface


Start Greeting

Enable the Start Greeting option if you want your assistant to begin the conversation automatically.
You can then write the Greeting Text, which is what the assistant will say at the start of each new session.
Example:

Example

“Hello, I would be happy to talk to you.”


System Prompt

The System Prompt is the foundation of your assistant’s personality and behavior.
This text acts as instructions to the large language model, guiding how the assistant should:

  • Interact with users
  • Follow specific rules and safety guardrails
  • Ask certain questions or collect defined information
  • Provide answers based on selected knowledge or context
  • Maintain a particular tone, such as formal, friendly, or neutral

Example:

Example

“You are Alex, the interviewer for ABC Company. Greet users politely and ask structured interview questions.”

Your prompt defines exactly what the assistant should do and how it should behave during every interaction.


Model Selection

Choose the Large Language Model (LLM) that powers your assistant.
Ziddny supports multiple advanced models, including:

  • ziddny-gemini 2.0 Flash
  • ziddny-gemini 2.5 Flash
  • ziddny-gpt-4 Omni
  • ziddny-gpt-4.1 Mini

Each model varies in speed and reasoning ability.
You can experiment with different ones to match your assistant’s purpose — for example, faster models for casual Q&A, and more capable models for deep reasoning or interviews.


Response Length Limit

This defines how long the assistant’s replies can be.
It is measured in tokens, where each token roughly equals 0.75 words.
For example, a limit of 300 tokens corresponds to about 225 words.


Creativity Level

Adjust the Creativity Level slider to control how flexible the assistant’s responses are:

  • 0.0 (Precise): Follows instructions exactly and avoids creative variations
  • 1.0 (Creative): Produces more expressive, open-ended, or varied responses
  • You can set it anywhere in between for a balanced tone

This setting helps you tailor the assistant’s behavior — for example, use a low setting for structured tasks (like interviews) and a higher setting for storytelling or conversational uses.


Knowledge Source

You can connect your assistant to an existing Knowledge Source from the dropdown menu.
Once linked, the assistant will reference that knowledge to provide context-based answers.

Learn more

Learn more about how to create and manage knowledge sources in Knowledge Setup.


Next Steps

After adjusting the behavior settings:

  1. Click Save to store your configuration.
  2. Use the Test button to try your assistant immediately.
  3. When you’re satisfied, click Update Live to apply changes to the published version.