mcp-n8n-builder

/** * N8N Workflow Composition Guide * * This guide provides instructions for LLMs on how to effectively compose n8n workflows. * It includes best practices, common patterns, and specific guidance for different workflow types. */ export const WORKFLOW_COMPOSITION_GUIDE = { /** * Core principles for effective workflow composition */ core_principles: ` # Core Principles for N8N Workflow Composition 1. **Start with the right trigger**: Every workflow needs a trigger node that initiates execution. - For automated workflows: Use Schedule, Webhook, or event-based triggers - For manual testing: Use Manual Trigger - For sub-workflows: Use Execute Workflow Trigger - For chat-based workflows: Use Chat Trigger or Manual Chat Trigger 2. **Follow a logical flow**: Arrange nodes in a clear left-to-right sequence that follows data progression. 3. **Transform data incrementally**: Use Set, Function, or Code nodes between major operations to shape data. 4. **Handle errors gracefully**: Include error handling using Error Trigger nodes or IF nodes with error conditions. 5. **Keep workflows focused**: Each workflow should accomplish a specific task. Break complex processes into multiple workflows. 6. **Use descriptive naming**: Give clear names to nodes and workflows to make them self-documenting. 7. **Test incrementally**: Build and test workflows in small increments rather than creating the entire workflow at once. 8. **Consider workflow activation requirements**: Only workflows with automatic trigger nodes (Schedule, Webhook, etc.) can be activated. Workflows with only manual triggers cannot be automatically activated. `, /** * Node categories and their purposes */ node_categories: ` # N8N Node Categories and Their Purposes ## Trigger Nodes Trigger nodes start workflow execution when specific events occur. - **Manual Trigger**: Starts workflow when manually triggered through the UI - **Schedule Trigger**: Starts workflow at specified times using cron expressions - **Webhook Trigger**: Starts workflow when HTTP requests are received - **Error Trigger**: Starts workflow when errors occur in other workflows - **Chat Trigger**: Starts workflow when chat messages are received - **API-specific Triggers**: Start workflows when events occur in specific services (Gmail, Slack, etc.) ## Action Nodes Action nodes perform operations on data or external systems. - **HTTP Request**: Makes HTTP requests to APIs - **Email Send**: Sends emails - **Database Operations**: Interact with databases (Postgres, MySQL, etc.) - **File Operations**: Read, write, or manipulate files - **Service-specific Actions**: Perform actions in specific services (Slack, Google Sheets, etc.) ## Transformation Nodes Transformation nodes modify, filter, or restructure data. - **Set**: Sets or updates values in the data - **Function**: Runs JavaScript code to transform data - **Code**: Runs more complex JavaScript code with more control - **JSON/XML**: Parse or create structured data - **Split/Merge**: Split data into multiple items or merge multiple items - **Filter**: Filter data based on conditions - **Switch**: Route data based on conditions ## AI Nodes AI nodes integrate with AI services and models. - **OpenAI**: Interact with OpenAI models - **LangChain Nodes**: Build complex AI applications with LangChain - **AI Transform**: Transform data using AI - **Agent**: Create AI agents that can use tools and make decisions - **Memory**: Store and retrieve conversation history ## Flow Control Nodes Flow control nodes manage the execution path of workflows. - **IF**: Conditional branching based on conditions - **Switch**: Multi-way branching based on values - **Merge**: Combine data from multiple paths - **Wait**: Pause workflow execution - **Loop**: Process items in a loop `, /** * Common workflow patterns */ common_patterns: ` # Common N8N Workflow Patterns ## Data Fetching and Processing Pattern - **Structure**: Trigger → Fetch Data → Transform Data → Action - **Example**: Schedule Trigger → HTTP Request → Set → Database - **Use Case**: Regularly fetch data from an API, transform it, and store it in a database - **Key Nodes**: Schedule Trigger, HTTP Request, Set, Database nodes ## Webhook Response Pattern - **Structure**: Webhook Trigger → Process Data → Respond to Webhook - **Example**: Webhook Trigger → Function → Respond to Webhook - **Use Case**: Create an API endpoint that processes incoming requests and returns responses - **Key Nodes**: Webhook Trigger, Function/Code, Respond to Webhook ## Conditional Execution Pattern - **Structure**: Trigger → Get Data → IF/Switch → Different Actions Based on Condition - **Example**: Webhook Trigger → HTTP Request → IF → Send Email / Slack Message - **Use Case**: Execute different actions based on data conditions - **Key Nodes**: Any Trigger, IF or Switch, various action nodes ## Batch Processing Pattern - **Structure**: Trigger → Get Batch Data → Split In Batches → Process Each Batch → Merge Results - **Example**: Schedule Trigger → Database → Split In Batches → HTTP Request → Merge - **Use Case**: Process large datasets in smaller chunks - **Key Nodes**: Any Trigger, Split In Batches, Merge ## Error Handling Pattern - **Structure**: Main Workflow with Error Trigger → Error Handling Workflow - **Example**: Main workflow with Error Trigger → Send Error Notification - **Use Case**: Capture and handle errors from other workflows - **Key Nodes**: Error Trigger, notification nodes (Email, Slack, etc.) ## Polling Pattern - **Structure**: Schedule Trigger → Fetch Data → Filter for New Items → Process New Items - **Example**: Schedule Trigger → HTTP Request → Filter → Function → Database - **Use Case**: Regularly check for new data and process only what's new - **Key Nodes**: Schedule Trigger, HTTP Request, Filter, Function ## Data Synchronization Pattern - **Structure**: Trigger → Fetch Source Data → Fetch Target Data → Compare → Update/Create/Delete - **Example**: Schedule Trigger → HTTP Request (Source) → Database (Target) → Function (Compare) → Database (Update) - **Use Case**: Keep data synchronized between different systems - **Key Nodes**: Any Trigger, multiple data source nodes, Function/Code for comparison `, /** * AI workflow patterns */ ai_workflow_patterns: ` # AI Workflow Patterns in N8N ## Basic AI Integration Pattern - **Structure**: Trigger → Prepare Input → AI Model → Process Output → Action - **Example**: Webhook → Set → OpenAI → Function → Send Email - **Use Case**: Enhance workflows with AI capabilities like text generation, summarization, or classification - **Key Nodes**: Any Trigger, OpenAI or other AI nodes, Function/Code for post-processing ## Chained Requests Pattern - **Structure**: Trigger → AI Model 1 → AI Model 2 → ... → AI Model N → Action - **Example**: Webhook → OpenAI (Analysis) → OpenAI (Generation) → Send Email - **Use Case**: Break down complex AI tasks into a series of specialized steps - **Key Nodes**: Any Trigger, multiple AI model nodes, Function/Code for intermediate processing ## Single Agent Pattern - **Structure**: Trigger → Input Preparation → AI Agent with Tools → Output Formatting → Action - **Example**: Chat Trigger → Set → Agent (with HTTP Request Tool) → Set → Respond to Chat - **Use Case**: Create an AI assistant that can use tools to accomplish tasks - **Key Nodes**: Chat Trigger, Agent, Tool nodes, Set for formatting ## Mixture of Experts (MOE) Pattern - **Structure**: Trigger → Input Preparation → Router → Expert Selection → Expert Processing → Merge Results → Output - **Example**: Chat Trigger → Set → Router LLM → Switch → Specialized LLMs → Merge → Respond to Chat - **Use Case**: Route queries to specialized AI models based on the query type - **Key Nodes**: Any Trigger, LLM nodes, Switch, Merge ## Multi-Agent with Gatekeeper Pattern - **Structure**: Trigger → Gatekeeper Agent → Specialist Agents → Gatekeeper for Final Response → Action - **Example**: Chat Trigger → Main Agent → Specialized Agents → Main Agent → Respond to Chat - **Use Case**: Create a system with a main agent that delegates to specialist agents - **Key Nodes**: Chat Trigger, multiple Agent nodes, Function/Code for coordination ## AI with Memory Pattern - **Structure**: Trigger → Retrieve Context → Merge with Input → AI Processing → Store New Context → Action - **Example**: Chat Trigger → Memory Retrieval → Set → LLM → Memory Storage → Respond to Chat - **Use Case**: Create AI systems that remember previous interactions - **Key Nodes**: Chat Trigger, Memory nodes, LLM nodes `, /** * Building blocks for different workflow types */ workflow_building_blocks: ` # Building Blocks for Different N8N Workflow Types ## Data Integration Workflows - **Essential Nodes**: HTTP Request, Database, Set, Function, Filter - **Key Patterns**: Polling, Data Synchronization, Batch Processing - **Best Practices**: - Use the Set node to normalize data between systems - Implement error handling for API failures - Use Filter nodes to process only relevant data - Consider rate limits when making API calls ## Notification Workflows - **Essential Nodes**: Trigger nodes, Filter, Set, Email Send, Slack, Discord, Telegram - **Key Patterns**: Conditional Execution, Error Handling - **Best Practices**: - Format messages appropriately for each platform - Use IF nodes to determine notification urgency - Include relevant context in notifications - Consider time zones for scheduled notifications ## Document Processing Workflows - **Essential Nodes**: Read Binary Files, Extract from File, HTTP Request, Set, Function - **Key Patterns**: Batch Processing, Conditional Execution - **Best Practices**: - Handle different file types appropriately - Implement error handling for file processing failures - Use Split In Batches for large document sets - Consider memory usage for large files ## AI Assistant Workflows - **Essential Nodes**: Chat Trigger, Agent, Memory, LLM nodes, Tool nodes - **Key Patterns**: Single Agent, Multi-Agent with Gatekeeper, AI with Memory - **Best Practices**: - Use Memory nodes to maintain conversation context - Implement clear system prompts for consistent behavior - Use Tool nodes to extend AI capabilities - Consider token usage and costs ## Mixture of Experts (MOE) Workflows - **Essential Nodes**: Chat Trigger, Router LLM, Switch, Expert LLMs, Merge - **Key Patterns**: MOE Pattern, Conditional Execution - **Best Practices**: - Design the router to accurately classify queries - Specialize each expert for its domain - Use consistent output formats across experts - Consider fallback options for unclassified queries ## Automation Workflows - **Essential Nodes**: Schedule Trigger, HTTP Request, Function, Database, IF - **Key Patterns**: Polling, Data Synchronization, Error Handling - **Best Practices**: - Use appropriate scheduling intervals - Implement idempotent operations when possible - Include comprehensive error handling - Log important events for monitoring `, /** * Step-by-step workflow creation process */ workflow_creation_process: ` # Step-by-Step N8N Workflow Creation Process ## 1. Define the Workflow Purpose - Clearly identify what the workflow should accomplish - Determine the trigger events that should start the workflow - Identify the expected output or actions ## 2. Plan the Workflow Structure - Choose the appropriate trigger node - Identify the main processing steps - Determine the necessary data transformations - Plan error handling and edge cases ## 3. Create the Trigger Node - Add the appropriate trigger node (Manual, Schedule, Webhook, etc.) - Configure the trigger settings (schedule, webhook path, etc.) - Test the trigger to ensure it activates correctly ## 4. Build the Core Processing Logic - Add nodes for the main processing steps - Configure each node with the appropriate settings - Use Set or Function nodes to prepare data between steps - Test each step incrementally ## 5. Implement Error Handling - Add IF nodes to check for error conditions - Configure error notification mechanisms - Consider adding Error Trigger nodes for critical workflows ## 6. Test the Complete Workflow - Test the workflow with various inputs - Verify that all paths work as expected - Check error handling by simulating failures ## 7. Refine and Optimize - Look for opportunities to simplify the workflow - Optimize for performance and reliability - Add comments or notes for clarity ## 8. Activate and Monitor - Activate the workflow if it has automatic triggers - Set up monitoring to track workflow execution - Plan for maintenance and updates `, /** * Tips for specific node types */ node_specific_tips: ` # Tips for Specific N8N Node Types ## Trigger Nodes - **Manual Trigger**: Use for testing or workflows that should only run on demand - **Schedule Trigger**: Use cron expressions for precise scheduling - **Webhook Trigger**: Consider authentication for public-facing webhooks - **Error Trigger**: Configure to capture specific error types ## HTTP Request Node - Use authentication when accessing protected APIs - Handle rate limiting with appropriate retry strategies - Parse response data with the JSON parameter - Use expressions for dynamic URLs and parameters ## Function/Code Nodes - Keep code modular and focused on specific tasks - Use comments to explain complex logic - Leverage the built-in n8n libraries and helpers - Return properly structured data for downstream nodes ## Database Nodes - Use parameterized queries to prevent SQL injection - Implement error handling for database operations - Consider connection pooling for high-volume workflows - Use transactions for related operations ## AI Nodes - Provide clear and specific prompts - Use system messages to guide AI behavior - Consider token limits and costs - Implement fallbacks for AI service outages ## Flow Control Nodes - **IF Node**: Use for simple binary conditions - **Switch Node**: Use for multiple possible paths - **Merge Node**: Configure the appropriate merge mode (Append, Merge by Key, etc.) - **Split In Batches Node**: Adjust batch size based on downstream requirements `, /** * Common workflow examples */ workflow_examples: ` # Common N8N Workflow Examples ## Data Synchronization Workflow \`\`\` Schedule Trigger → HTTP Request (Source API) → Function (Transform) → Database (Get Existing) → Function (Compare) → IF → Database (Insert/Update) → Error Handling \`\`\` ## Customer Support Notification Workflow \`\`\` Webhook Trigger → Function (Parse Request) → Switch (Priority) → Set (Format Message) → Slack/Email/SMS → Respond to Webhook \`\`\` ## Document Processing Workflow \`\`\` Webhook Trigger → Read Binary Files → Extract from File → Function (Process Data) → Database (Store Results) → Send Email (Notification) → Respond to Webhook \`\`\` ## AI Content Generation Workflow \`\`\` Schedule Trigger → HTTP Request (Get Topics) → Split In Batches → OpenAI (Generate Content) → Function (Format) → WordPress/CMS (Post Content) → Send Email (Report) \`\`\` ## Mixture of Experts (MOE) Workflow \`\`\` Chat Trigger → Set (Prepare Input) → Router LLM → Switch → Expert LLMs (Code/Math/Creative/General) → Merge → Set (Format Response) → Respond to Chat \`\`\` ## Error Monitoring Workflow \`\`\` Error Trigger → Function (Parse Error) → Switch (Severity) → Set (Format Alert) → Slack/Email/PagerDuty → Database (Log Error) → IF (Critical) → HTTP Request (Restart Service) \`\`\` `, /** * Mixture of Experts (MOE) specific guidance */ moe_specific_guidance: ` # Mixture of Experts (MOE) Workflow Guidance ## What is a Mixture of Experts (MOE) Workflow? A Mixture of Experts workflow routes queries or tasks to specialized "expert" components based on the nature of the input. Each expert is optimized for a specific type of task, allowing the system to provide high-quality responses across a diverse range of queries. ## Key Components of MOE Workflows ### 1. Input Handler - Receives and preprocesses the initial query - Typically a Chat Trigger or Webhook Trigger node - May include Set nodes for input formatting ### 2. Router - Analyzes the input to determine which expert should handle it - Usually an LLM node with a specialized prompt - Should output a clear classification or routing decision ### 3. Switching Mechanism - Directs the query to the appropriate expert based on the router's decision - Typically a Switch node - Should handle all possible routing outcomes ### 4. Expert Components - Specialized nodes or sub-workflows optimized for specific types of queries - Each expert should have a clear domain of expertise - Experts can be LLMs with different prompts, parameters, or even different models ### 5. Result Aggregator - Combines the outputs from the experts - Typically a Merge node - Ensures consistent output format regardless of which expert was used ### 6. Response Formatter - Prepares the final response for delivery - Usually a Set node - Ensures consistent presentation of results ## MOE Implementation Approaches ### 1. Prompt-Based Experts - Use the same LLM with different prompts for each expert - Simpler to implement but may have limitations in specialization - Example: OpenAI node with different system messages for each expert ### 2. Model-Based Experts - Use different AI models for different types of queries - More powerful but potentially more complex and costly - Example: GPT-4 for reasoning, Claude for creative, Mistral for code ### 3. Tool-Based Experts - Combine AI models with specialized tools for different domains - Most flexible but requires more complex workflow design - Example: Code expert with GitHub integration, Math expert with calculator tools ## Best Practices for MOE Workflows 1. **Design an effective router**: The router should accurately classify queries to ensure they reach the appropriate expert. 2. **Balance expert specialization**: Each expert should have a clear domain without too much overlap. 3. **Maintain consistent output formats**: Ensure all experts return results in a compatible format. 4. **Implement a fallback expert**: Include a general-purpose expert for queries that don't clearly match any specialty. 5. **Consider performance and cost**: Different experts may have different resource requirements and costs. 6. **Test with diverse inputs**: Verify that the router correctly classifies a wide range of queries. 7. **Monitor and refine**: Track which experts are used most frequently and refine the system accordingly. `, /** * Error handling best practices */ error_handling: ` # Error Handling Best Practices in N8N Workflows ## Common Error Types - API connection failures - Data validation errors - Authentication issues - Rate limiting - Timeout errors - Resource constraints ## Workflow-Level Error Handling - Use Error Trigger nodes to create dedicated error handling workflows - Configure workflow settings for error behavior - Implement global error notification systems ## Node-Level Error Handling - Use IF nodes to check for error conditions - Implement retry logic for transient errors - Use Try/Catch patterns with Function nodes ## Data Validation - Validate input data early in the workflow - Use Function nodes to check data integrity - Implement type checking and format validation ## Error Notification Strategies - Send alerts for critical errors - Log all errors for analysis - Include context information in error messages - Consider severity levels for different types of errors ## Recovery Strategies - Implement automatic retries with exponential backoff - Create self-healing workflows where possible - Design graceful degradation for partial failures ## Testing for Errors - Simulate error conditions during testing - Verify that error handling works as expected - Test recovery mechanisms `, /** * Performance optimization */ performance_optimization: ` # Performance Optimization for N8N Workflows ## Workflow Design Optimization - Keep workflows focused and modular - Break large workflows into smaller, connected workflows - Use Execute Workflow nodes for reusable components - Minimize unnecessary data processing ## Data Handling Optimization - Filter data early to reduce processing volume - Use pagination for large datasets - Implement batching for bulk operations - Minimize data transformations ## Node-Specific Optimization - Use Set nodes instead of Function nodes for simple transformations - Optimize database queries to retrieve only necessary data - Use appropriate indexing for database operations - Configure HTTP Request nodes with proper caching ## Resource Management - Schedule resource-intensive workflows during off-peak hours - Implement rate limiting for API calls - Monitor memory usage for large data processing - Consider horizontal scaling for high-volume workflows ## Caching Strategies - Implement caching for frequently accessed data - Use If nodes to skip unnecessary processing - Store intermediate results for long-running workflows - Consider external caching solutions for shared data `, /** * Workflow testing and debugging */ testing_and_debugging: ` # Workflow Testing and Debugging in N8N ## Testing Approaches - Test workflows incrementally during development - Use Manual Trigger nodes for isolated testing - Create test datasets for consistent testing - Implement end-to-end testing for critical workflows ## Debugging Techniques - Use the n8n execution history to review workflow runs - Add debug nodes to inspect data at various points - Use console.log() in Function nodes for detailed logging - Implement conditional breakpoints with IF nodes ## Common Issues and Solutions - Data type mismatches: Use Set nodes to ensure correct types - Missing fields: Implement default values or error checking - API rate limiting: Add delays or implement backoff strategies - Authentication failures: Verify credential configuration ## Testing Tools - Use the n8n built-in testing features - Implement monitoring for production workflows - Create dedicated testing workflows - Use version control for workflow configurations `, /** * Security considerations */ security_considerations: ` # Security Considerations for N8N Workflows ## Authentication and Authorization - Use secure authentication methods for API connections - Implement proper authorization checks in webhook endpoints - Rotate API keys and credentials regularly - Use the n8n credentials manager for sensitive information ## Data Security - Minimize storage of sensitive data - Implement data masking for logs and outputs - Consider encryption for sensitive stored data - Implement proper data retention policies ## Access Control - Restrict access to workflow editing - Implement role-based access control - Audit workflow changes regularly - Use separate environments for development and production ## Network Security - Secure webhook endpoints with authentication - Implement rate limiting for public endpoints - Use HTTPS for all external communications - Consider IP restrictions for sensitive endpoints ## Error Handling Security - Avoid exposing sensitive information in error messages - Implement proper logging for security events - Create alerts for suspicious activities - Regularly review error logs for security issues `, };