Skip to main content

What are Assistants?

Assistants are pre-configured AI agents with custom instructions, parameters, and settings. They allow you to:
  • Define specialized behavior and tone
  • Set default model parameters
  • Create reusable personas for different tasks
  • Associate assistants with projects

Creating an Assistant

Basic Setup

  1. Go to Settings > Assistants
  2. Click Add Assistant
  3. Fill in the basic information:

Name

A descriptive name for your assistant (e.g., “Code Reviewer”, “Creative Writer”)

Avatar

Choose an emoji or custom image to identify your assistant visually.

Description (Optional)

A brief explanation of what the assistant does.

Instructions

Instructions define your assistant’s behavior, knowledge, and response style.
Dynamic Date Insertion: Use {{current_date}} in your instructions to automatically insert the current date. This helps the model provide time-aware responses.Example: Today's date is {{current_date}}. Consider current events when responding.

Parameters

Configure model behavior with custom parameters:

Predefined Parameters

Jan provides quick-access buttons for common parameters:
  • Temperature - Controls randomness (0.0-2.0)
  • Top P - Nucleus sampling threshold (0.0-1.0)
  • Top K - Limits token selection (1-100)
  • Max Tokens - Maximum response length
  • Presence Penalty - Reduces repetition (-2.0 to 2.0)
  • Frequency Penalty - Discourages repeated phrases (-2.0 to 2.0)
Click a predefined parameter to add it with default values.

Custom Parameters

Add your own parameters:
  1. Click Add Parameter (+)
  2. Enter a Key name
  3. Select a Type:
    • String - Text value
    • Number - Numeric value
    • Boolean - True/false
    • JSON - Complex object
  4. Enter the Value

Using Assistants

In New Chats

When starting a new chat:
  1. Click the Assistant dropdown in the chat input area
  2. Select an assistant from the list
  3. The assistant’s instructions and parameters will be applied

In Projects

Assign a default assistant to a project:
  1. Create or edit a project
  2. Select an assistant from the dropdown
  3. All new threads in that project will use this assistant
Project assistants are automatically applied to new conversations, but you can override them by manually selecting a different assistant.

Managing Assistants

Editing Assistants

  1. Go to Settings > Assistants
  2. Click the assistant you want to edit
  3. Modify any field (name, avatar, instructions, parameters)
  4. Click Save
Changes to assistants only affect new conversations. Existing threads continue using the original instructions.

Deleting Assistants

  1. Go to Settings > Assistants
  2. Click the delete icon next to the assistant
  3. Confirm deletion
Deleting an assistant removes it from all projects and new chat options. Existing threads are not affected.

Assistant Best Practices

Clear Instructions

  • Be specific about the assistant’s role and expertise
  • Define the expected behavior and tone
  • Include examples when helpful
  • Keep instructions focused and concise

Parameter Tuning

  • 0.0 - 0.3 - Focused, deterministic (code, facts, analysis)
  • 0.4 - 0.7 - Balanced (general chat, explanations)
  • 0.8 - 1.2 - Creative (writing, brainstorming)
  • 1.3 - 2.0 - Highly creative (experimental, artistic)
  • Presence Penalty - Increases with unique topics (0.6 for diverse responses)
  • Frequency Penalty - Reduces word repetition (0.5 for varied phrasing)
  • Start with low values and adjust based on results

Testing Assistants

Before using an assistant in production:
  1. Create test conversations with typical queries
  2. Verify the tone and style match expectations
  3. Adjust parameters based on response quality
  4. Iterate on instructions for edge cases

Example Assistants

Instructions:
Parameters:
  • temperature: 0.3
  • max_tokens: 2000
  • presence_penalty: 0.1
Instructions:
Parameters:
  • temperature: 0.4
  • top_p: 0.85
  • max_tokens: 1500
Instructions:
Parameters:
  • temperature: 1.1
  • presence_penalty: 0.6
  • frequency_penalty: 0.4
  • max_tokens: 3000