HomeGuidesSDK ExamplesAnnouncementsCommunity
Guides
Start typing to search…

Creating and Managing Assistants in Skill Studio

This comprehensive guide walks you through creating and managing assistants in Max's Skill Studio. Whether you're building your first assistant or optimizing existing ones, this tutorial covers everything from basic setup to advanced configuration.

Getting Started with Skill Studio

Accessing Skill Studio

To begin creating your assistant:

  1. Open Max AI
  2. Click the Max head icon in the top-left corner
  3. Select Skill Studio from the menu

You'll now see the Skill Studio interface with navigation icons on the left side. Each icon represents different functions, though availability depends on your permissions.

📘

Key Areas for Assistant Creation

The primary areas you'll use include:

  • Assistants - Main configuration area
  • Datasets - Data management
  • Database Connections - Data source setup
  • API Configuration - LLM settings
  • Prompt Library - Prompt management
  • Monitoring - Performance tracking

Creating Your First Assistant

Step 1: Navigate to Assistant Creation

  1. In the Assistants tab, look at the top-left corner
  2. Click the current assistant name to see available assistants
  3. Click the "+" button next to the assistant selector
  4. Select Add Assistant (not Import Assistant)

Step 2: Configure Basic Settings

When creating a new assistant, you'll need:

  1. Assistant Name

    • Choose a descriptive, meaningful name
    • This will be visible to end users

  2. Dataset Selection

    • Select from the dropdown of available datasets
    • If no dataset exists yet, you can create the assistant without one initially
    • You'll need to add a dataset before skills can run properly

  3. Initial Skills

    • Select from available skills on the right side
    • AnswerRocket-managed skills show an "AnswerRocket" tag
    • Recommended starter skills:
      • Trend Analysis - For historical data patterns
      • Data Explorer - For flexible data browsing
💡

Pro Tip

Start with just 2-3 skills to keep things simple. You can always add more skills later as your needs grow.

Step 3: Create the Assistant

Click the Create button to finalize your assistant creation.

Understanding the Assistant Interface

Once created, you'll see several key areas:

Top Navigation Bar

  • Chat - Test skills with a mock chat interface
  • Questions - Manage and run saved questions
  • Tests - Run regression tests on question collections
  • Settings - Configure assistant properties

Main Canvas Area

The layout varies based on skill type:

  • Code Skills - Shows file tree and read-only code
  • Component Skills - Shows visual skill builder
  • Graph Skills - Shows node-based workflow

Skill Management

Next to the Read/Edit toggle, you'll find:

  • Skill Selector - Shows current skill and allows switching
  • Add New Skill button - Access additional skills
⚠️

Edit Mode Warning

Only one user can edit at a time. The Read/Edit toggle shows who's currently editing. Override with caution - communicate with the current editor first!

Configuring Skills

Skill Properties

Access skill settings via the gear icon on the right side:

  1. Basic Properties

    • Active for Chat - Toggle to enable/disable in production
    • Name - Internal skill identifier
    • Description - Brief explanation of skill purpose
    • Dataset - Assign specific dataset or inherit from assistant

  2. LLM Configuration

    • LLM Description - How the LLM understands this skill
    • Capabilities - What the skill can do
    • Limitations - What the skill cannot do
    • Example Questions - Sample queries that trigger this skill
    • Parameter Guidance - Instructions for parameter extraction
🎯

Best Practice

Write clear, specific descriptions in these fields. They directly impact how well the LLM selects and uses your skills.

Working with Parameters

Parameters are the bridge between user questions and skill execution. Access them via the variables icon:

Chat Parameters

Extracted from user queries:

  • Name - Parameter identifier
  • Description - Purpose and usage
  • Limit Values - Restrict to specific data attributes
  • LLM Description - Guide LLM in value assignment
  • Default Value - Fallback when not specified

Code Parameters

Pre-defined by skill creators:

  • Usually read-only except for descriptions
  • Modify descriptions to improve LLM understanding

Prompt Variables

Customize prompts used by skills:

  • Select different prompts from the library
  • Adjust descriptions for better context
🎯

Try This!

Create a chat parameter for "time_period" with:

  • Description: "The date range for analysis"
  • Limit to: Date dimensions from your dataset
  • Default: "Last 30 days"

Global Functions

Add custom functions at two levels:

  1. Skill Functions - Specific to current skill
  2. Assistant Functions - Available to all skills
    • Has separate read/edit lock
    • Green dot indicates existing code

Advanced Configuration

Visual Layouts

Create custom visualization layouts:

  1. Navigate to Visual Layouts in left navigation
  2. Click "+" to create new layout
  3. Add components:
    • Data tables
    • Charts
    • Headers/Paragraphs
    • Containers
    • Markdown blocks
  4. Configure inputs and properties
  5. Save and assign to skills
💡

Layout Tips

  • Start by copying an existing layout's code
  • Use the Generate Layout feature for AI assistance
  • Test with real data before finalizing

Skill Resources

Upload files for skill use:

  1. Click Resources next to Variables
  2. Click Add File and upload
  3. Use the kebab menu (⋮) to:
    • Download files
    • Copy resource links
    • Rename or delete

Common uses:

  • Brand logos for visualizations
  • Configuration files
  • Reference data

Prompt Library

Manage prompts centrally:

  1. Navigate to Prompt Library in left navigation
  2. Create or edit prompts
  3. Configure:
    • Prompt content with variables
    • K-shot examples for better matching
    • Match scores and counts
  4. Test prompts before deployment

Assistant Settings

Access comprehensive settings via the Settings tab:

Appearance Customization

  1. Basic Information

    • Name (visible to users)
    • Description
    • Footer message

  2. Visual Identity

    • Assistant Avatar - Icon in chat messages
    • Assistant Icon - Shown in selectors
    • Banner Image - Full-width header image
🎨

Design Guidelines

  • Avatars: Use simple, recognizable icons (avoid text)
  • Icons: Clear at small sizes
  • Banners: Full-width logos or branded headers

System Prompt Configuration

The system prompt defines core assistant behavior:

  1. Base System Prompt

    • Pre-populated with variables
    • Includes dataset metadata
    • Contains date/time information

  2. Prompt Variables

    • Persona - How the assistant communicates
    • Skill Selection Guidance - Skill choosing logic
    • Parameter Selection Guidance - Parameter extraction rules
    • Answer Response Guidance - Response formatting
📝

System Prompt Best Practice

For assistants created before version 25.07, consider creating a new assistant and copying its system prompt for updated functionality.

Data Configuration

  1. Dataset Assignment

    • Set default dataset for all skills
    • Individual skills can override

  2. Caching Options

    • Enable/disable skill result cache
    • Cache improves response time
    • Disable for always-fresh results

Landing Page

Create custom HTML landing pages:

<div style="padding: 20px;">
  <h1>Welcome to Your Assistant</h1>
  <p>Click questions below to get started:</p>
  <ul>
    <li onclick="window.parent.postMessage({askQuestion: 'Show sales trends'}, '*')">
      Sales Analysis
    </li>
  </ul>
</div>
⚠️

Landing Page Requirements

  • Use inline CSS only
  • No <head> section
  • Keep functionality simple
  • Ensure responsive design

API Configuration

Configure LLMs for different purposes:

  • Chat - Main conversation model
  • Embeddings - Semantic search
  • Narrative - Report generation
  • Evaluations - Quality assessment
  • SQL - Query generation
  • Research - Advanced analysis

Version Management

Creating Versions

Before major changes:

  1. Navigate to Versions in Settings
  2. Click Save Version
  3. Add version name and description
  4. Version captures current state

Comparing Versions

  1. Select a version
  2. Click Compare
  3. Review differences in:
    • Skills
    • Settings
    • Prompts
    • Configuration

Restoring Versions

  1. Select target version
  2. Click Restore
  3. System prompts to save current state
  4. Confirm restoration
🔄

Version Strategy

Save versions:

  • Before major updates
  • After successful deployments
  • When experimenting with new features

Sharing and Permissions

Managing Access

Via the Sharing section:

  1. Add Users

    • Search by name or email
    • Assign as User or Owner

  2. Permission Levels

    • Users - Can use the assistant
    • Owners - Can modify configuration

  3. Group Sharing

    • Share with predefined groups
    • Efficient for team access

Import/Export Operations

Exporting Assistants

  1. Click the three-dot menu (⋮)
  2. Select Export Assistant
  3. Creates transport package with:
    • All skills
    • Resources
    • Configuration
    • Datasets references

Importing Assistants

  1. Click "+" → Import Assistant
  2. Upload transport package (.zip)
  3. Review and confirm import
📦

Transport Package Notes

  • Move between environments easily
  • Version compatibility: Old → New only
    • Check versions via Account menu

Testing and Monitoring

Using the Chat Feature

Test your assistant:

  1. Click Chat in top navigation
  2. Type test questions
  3. Verify skill selection
  4. Check parameter extraction
  5. Review responses

Creating Test Suites

  1. Navigate to Questions tab
  2. Create question collections
  3. Run via Tests tab
  4. Configure:
    • Model overrides
    • Skip skill runs option
    • Evaluation settings

Monitoring Performance

Regular monitoring ensures quality:

  1. Check skill execution logs
  2. Review user feedback
  3. Analyze parameter extraction accuracy
  4. Track response quality
🎯

Testing Challenge

Create a test suite with:

  1. 5 questions covering different skills
  2. Edge cases for parameter extraction
  3. Follow-up questions requiring context

Run the suite and review results for improvement opportunities.

Troubleshooting Common Issues

Skill Not Triggering

  1. Check skill is Active for Chat
  2. Review Example Questions
  3. Verify LLM Description clarity
  4. Test with exact example questions

Parameter Extraction Failures

  1. Review parameter descriptions
  2. Check limit values configuration
  3. Add more specific LLM descriptions
  4. Set appropriate default values

Performance Issues

  1. Enable result caching
  2. Optimize dataset queries
  3. Review skill complexity
  4. Check API configuration

Best Practices Summary

  1. Start Simple

    • Begin with 2-3 essential skills
    • Add complexity gradually
    • Test thoroughly at each stage

  2. Document Everything

    • Clear descriptions for all components
    • Meaningful example questions
    • Comprehensive parameter guidance

  3. Version Regularly

    • Before major changes
    • After stable releases
    • During experimentation

  4. Monitor Continuously

    • Review logs regularly
    • Act on user feedback
    • Track performance metrics

  5. Collaborate Effectively

    • Communicate about edits
    • Share knowledge via resources
    • Document custom functions
🏆

You're Ready!

You now have the knowledge to create and manage sophisticated assistants in Skill Studio. Start with a simple assistant, test thoroughly, and gradually add complexity as you master each feature.

Remember: Great assistants evolve through iteration. Don't aim for perfection on the first try - build, test, learn, and improve!

Updated 6 days ago