AI Settings - Configure Local AI Models

Tabbed Configuration System

• Connection Tab: API URL, model selection, and connection testing
• Default Prompt Tab: System prompt customization and behavior configuration
• Advanced Tab: Response length controls and performance optimization

Real-Time Validation

• Connection Testing: Built-in connectivity verification
• Model Discovery: Automatic detection of available local models
• Status Feedback: Real-time status updates and error reporting
• Settings Persistence: Automatic saving to workbook and global configuration

API Connection Settings

• API URL: Ollama server endpoint (default: http://localhost:11434)
• Model Selection: Choose from available local AI models
• Model Refresh: Dynamic model list updates with 🔄 refresh button
• Connection Testing: "Test Connection" button for connectivity verification

Model Management

• Available Models: Dropdown list of installed Ollama models
• Model Types: Support for llama2, codellama, mistral, phi, and other Ollama-compatible models
• Refresh Capability: Update model list without restarting Excel
• Selection Persistence: Remember selected model across sessions

Connection Testing

• Test Connection Button: Verify connectivity to Ollama service
• Status Display: Real-time connection status and error messages
• Help Integration: Built-in help system with installation guidance
• Troubleshooting: Detailed error messages for connection issues

System Prompt Editor

• Multi-line Editor: Rich text editor for system prompt customization
• Default Behavior: Pre-configured prompt optimized for Excel assistance
• Custom Personalities: Create specialized AI assistants for different use cases
• Reset Functionality: Quick restoration to default settings with Reset button

Prompt Engineering

• Behavior Specification: Define how AI should respond to different question types
• Context Integration: Specify how AI should handle Excel data and context
• Response Style: Control formality, detail level, and explanation depth
• Domain Expertise: Configure AI for specific domains (finance, marketing, technical)

Default Prompt Template

The system includes a professional default prompt:

You are an Excel AI assistant. Help with formulas, data analysis, and spreadsheet tasks. Be concise and practical.

Response Length Controls

• Chat Response Max Tokens: Slider control (500-8000 tokens)
• Function Response Max Tokens: Slider control (100-2000 tokens)
• Real-time Display: Token count display updates as you adjust sliders
• Performance Guidance: Built-in tips for optimal token settings

Token Limit Guidelines

• 500-1000 tokens: Quick responses (~375-750 words)
• 1000-2000 tokens: Standard responses (~750-1500 words)
• 2000-4000 tokens: Comprehensive analysis (~1500-3000 words)
• 4000-8000 tokens: Extensive responses (~3000-6000 words)

Performance Optimization

• Response Speed: Lower token limits for faster responses
• Quality Balance: Higher limits for more detailed, comprehensive answers
• Memory Management: Efficient handling of conversation context
• Resource Considerations: Guidance on system resource usage

Multi-Level Configuration

• Global Settings: Stored in user AppData for system-wide defaults
• Workbook Settings: Stored in Excel workbook properties for project-specific configuration
• Priority System: Workbook settings override global settings when available

Configuration Storage

• JSON Format: Settings stored in structured JSON format
• Location: %AppData%\UnleashedFlow\ai-settings.json
• Backup: Automatic backup of configuration settings
• Portability: Settings can be shared across systems

Configuration Management

• Document Settings: Keep records of configuration changes and purposes
• Test Thoroughly: Always test configuration changes with representative tasks
• Backup Configurations: Keep backups of working configurations
• Version Control: Track configuration versions for different use cases

Model Selection Strategy

• Start Simple: Begin with general-purpose models like llama2
• Specialize Gradually: Move to specialized models for specific needs
• Consider Resources: Balance model capability with system resources
• Update Regularly: Keep models updated for improved performance

Performance Optimization

• Token Management: Adjust token limits based on response requirements
• Connection Stability: Ensure stable Ollama connection
• Resource Monitoring: Monitor system resources during AI operations
• Error Handling: Configure appropriate error handling and timeouts

Connection Issues

• "Cannot connect to Ollama": Verify Ollama is installed and running
• "No models available": Install models using ollama pull [model-name]
• "Connection timeout": Check firewall settings and Ollama service status
• "Invalid API URL": Verify URL format and port number

Model Issues

• "Model not found": Refresh model list or reinstall model
• "Model loading slowly": Large models take time to load initially
• "Out of memory": Use smaller models or increase system RAM
• "Inconsistent responses": Clear conversation history or restart service

Settings Issues

• "Settings not saving": Check file permissions in AppData directory
• "Configuration reset": Verify workbook properties are accessible
• "Prompt not working": Test prompt changes with simple questions first

Workflow Integration

AI Settings provides the foundation for all AI tools:

• AI Companion: Uses connection and prompt settings
• AI Data Tools: Leverage model selection and performance settings
• UDF Functions: Use function token limits and model configuration

Consistent Configuration

• Unified Settings: All AI tools use the same configuration
• Seamless Experience: Settings apply across all AI features
• Centralized Management: Single location for all AI configuration