Skip to content

02-ai-nodes.md

Latest commit

 

History

History
668 lines (578 loc) · 14.6 KB

02-ai-nodes.md

File metadata and controls

668 lines (578 loc) · 14.6 KB
layout default
title n8n AI Tutorial - Chapter 2: AI Nodes
nav_order 2
has_children false
parent n8n AI Tutorial

Chapter 2: AI Nodes and LLM Integration

Welcome to Chapter 2: AI Nodes and LLM Integration. In this part of n8n AI Tutorial: Workflow Automation with AI, you will build an intuitive mental model first, then move into concrete implementation details and practical production tradeoffs.

Configure and use different AI providers, manage credentials, and build multi-model workflows.

AI Node Overview

n8n provides dedicated nodes for various AI providers, each with specific capabilities and configuration options.

OpenAI Nodes

Chat Completion Node

{
  "parameters": {
    "model": "gpt-4o",
    "messages": [
      {
        "role": "system",
        "content": "You are a helpful assistant specialized in {{ $json.domain }}."
      },
      {
        "role": "user",
        "content": "={{ $json.question }}"
      }
    ],
    "temperature": 0.7,
    "maxTokens": 1000,
    "topP": 0.9,
    "frequencyPenalty": 0.0,
    "presencePenalty": 0.0,
    "responseFormat": "text"
  },
  "name": "OpenAI Chat",
  "type": "@n8n/n8n-nodes-langchain.openAi",
  "credentials": {
    "openAiApi": "openai-api"
  }
}

Embeddings Node

{
  "parameters": {
    "model": "text-embedding-ada-002",
    "input": "={{ $json.texts }}",
    "encodingFormat": "float"
  },
  "name": "OpenAI Embeddings",
  "type": "@n8n/n8n-nodes-langchain.openAi",
  "credentials": {
    "openAiApi": "openai-api"
  }
}

Function Calling

{
  "parameters": {
    "model": "gpt-4o",
    "messages": [
      {
        "role": "user",
        "content": "Check the weather in {{ $json.location }}"
      }
    ],
    "functions": [
      {
        "name": "get_weather",
        "description": "Get current weather for a location",
        "parameters": {
          "type": "object",
          "properties": {
            "location": {
              "type": "string",
              "description": "City and country"
            }
          },
          "required": ["location"]
        }
      }
    ]
  },
  "name": "OpenAI Functions",
  "type": "@n8n/n8n-nodes-langchain.openAi"
}

Anthropic Claude Nodes

Claude Chat

{
  "parameters": {
    "model": "claude-3-sonnet-20240229",
    "prompt": "Human: {{ $json.question }}\n\nAssistant:",
    "maxTokensToSample": 1000,
    "temperature": 0.7,
    "topP": 0.9,
    "topK": 250
  },
  "name": "Claude Chat",
  "type": "@n8n/n8n-nodes-langchain.anthropic",
  "credentials": {
    "anthropicApi": "anthropic-api"
  }
}

Local AI with Ollama

Ollama Integration

{
  "parameters": {
    "baseUrl": "http://localhost:11434",
    "model": "llama2:13b",
    "prompt": "{{ $json.prompt }}",
    "options": {
      "temperature": 0.7,
      "top_p": 0.9,
      "num_predict": 500,
      "stop": ["Human:", "\n\n"]
    }
  },
  "name": "Ollama Chat",
  "type": "@n8n/n8n-nodes-langchain.ollama",
  "typeVersion": 1
}

Running Ollama Locally

# Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh

# Pull models
ollama pull llama2
ollama pull codellama
ollama pull mistral

# Start Ollama service
ollama serve

Hugging Face Models

Text Generation

{
  "parameters": {
    "model": "gpt2",
    "inputs": "{{ $json.text }}",
    "parameters": {
      "max_new_tokens": 100,
      "temperature": 0.7,
      "do_sample": true,
      "pad_token_id": 50256
    },
    "options": {
      "use_gpu": false,
      "device": "cpu"
    }
  },
  "name": "HuggingFace Text Gen",
  "type": "@n8n/n8n-nodes-langchain.huggingFaceInference",
  "credentials": {
    "huggingFaceApi": "huggingface-api"
  }
}

AI Agent Node

Basic Agent Configuration

{
  "parameters": {
    "model": "gpt-4o",
    "prompt": "You are a helpful AI assistant. Use the available tools to answer questions.\n\nAvailable tools:\n{{ $json.tools }}\n\nQuestion: {{ $json.question }}",
    "maxIterations": 5,
    "returnIntermediateSteps": false,
    "memory": "buffer",
    "tools": [
      {
        "name": "web_search",
        "description": "Search the web for information",
        "parameters": {
          "query": "string"
        }
      },
      {
        "name": "calculator",
        "description": "Perform mathematical calculations",
        "parameters": {
          "expression": "string"
        }
      }
    ]
  },
  "name": "AI Agent",
  "type": "@n8n/n8n-nodes-langchain.agent",
  "credentials": {
    "openAiApi": "openai-api"
  }
}

Vector Store Integration

Pinecone Setup

{
  "parameters": {
    "operation": "upsert",
    "pineconeIndex": "my-index",
    "items": [
      {
        "id": "={{ $json.id }}",
        "values": "={{ $json.embedding }}",
        "metadata": {
          "text": "={{ $json.text }}",
          "source": "={{ $json.source }}"
        }
      }
    ]
  },
  "name": "Pinecone Upsert",
  "type": "@n8n/n8n-nodes-langchain.pinecone",
  "credentials": {
    "pineconeApi": "pinecone-api"
  }
}

Similarity Search

{
  "parameters": {
    "operation": "getMany",
    "pineconeIndex": "my-index",
    "query": "={{ $json.embedding }}",
    "numberOfResults": 5,
    "includeValues": false,
    "includeMetadata": true
  },
  "name": "Pinecone Search",
  "type": "@n8n/n8n-nodes-langchain.pinecone",
  "credentials": {
    "pineconeApi": "pinecone-api"
  }
}

Credential Management

Setting Up Credentials

  1. Go to Settings → Credentials in n8n UI
  2. Click "Add Credential"
  3. Select credential type (OpenAI, Anthropic, etc.)
  4. Enter API keys and other required information
  5. Test connection
  6. Save credential

Credential Types

// OpenAI Credential
{
  "name": "OpenAI API",
  "type": "openAiApi",
  "data": {
    "apiKey": "sk-..."
  }
}

// Anthropic Credential
{
  "name": "Anthropic API",
  "type": "anthropicApi",
  "data": {
    "apiKey": "sk-ant-..."
  }
}

// Pinecone Credential
{
  "name": "Pinecone API",
  "type": "pineconeApi",
  "data": {
    "apiKey": "...",
    "environment": "us-east-1-aws"
  }
}

Multi-Provider Workflows

Fallback Strategy

{
  "nodes": [
    {
      "parameters": {
        "model": "gpt-4o",
        "messages": [{"role": "user", "content": "={{ $json.question }}"}]
      },
      "name": "Primary AI (GPT-4)",
      "type": "@n8n/n8n-nodes-langchain.openAi",
      "continueOnFail": true
    },
    {
      "parameters": {
        "model": "claude-3-sonnet-20240229",
        "prompt": "Human: {{ $json.question }}\n\nAssistant:"
      },
      "name": "Fallback AI (Claude)",
      "type": "@n8n/n8n-nodes-langchain.anthropic"
    },
    {
      "parameters": {
        "conditions": {
          "string": [
            {
              "value1": "={{ $node['Primary AI (GPT-4)'].error }}",
              "operation": "isEmpty"
            }
          ]
        }
      },
      "name": "Check Primary Success",
      "type": "n8n-nodes-base.if"
    }
  ],
  "connections": {
    "Primary AI (GPT-4)": {
      "main": [
        [
          {
            "node": "Check Primary Success",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Check Primary Success": {
      "main": [
        [
          {
            "node": "Fallback AI (Claude)",
            "type": "main",
            "index": 1
          }
        ]
      ]
    }
  }
}

Model Selection Logic

{
  "nodes": [
    {
      "parameters": {
        "values": {
          "string": [
            {
              "name": "model_choice",
              "value": "={{ $json.complexity === 'high' ? 'gpt-4o' : 'gpt-3.5-turbo' }}"
            }
          ]
        }
      },
      "name": "Model Selector",
      "type": "n8n-nodes-base.set"
    },
    {
      "parameters": {
        "model": "={{ $json.model_choice }}",
        "messages": [{"role": "user", "content": "={{ $json.question }}"}]
      },
      "name": "Dynamic AI",
      "type": "@n8n/n8n-nodes-langchain.openAi"
    }
  ]
}

Custom AI Nodes

JavaScript Code Node

// Custom AI processing
const response = await $node.openAi.default.sendMessage({
  model: 'gpt-4o',
  messages: [
    {
      role: 'system',
      content: 'You are a data analyst. Provide insights in JSON format.'
    },
    {
      role: 'user',
      content: $input.item.json.data
    }
  ]
});

// Parse and structure response
const insights = JSON.parse(response.choices[0].message.content);

return [{
  json: {
    insights: insights,
    timestamp: new Date().toISOString(),
    model: response.model
  }
}];

HTTP Request Node for Custom APIs

{
  "parameters": {
    "method": "POST",
    "url": "https://api.custom-ai.com/generate",
    "sendBody": true,
    "bodyContentType": "json",
    "bodyParameters": {
      "parameters": [
        {
          "name": "prompt",
          "value": "={{ $json.prompt }}"
        },
        {
          "name": "model",
          "value": "custom-model-v1"
        }
      ]
    },
    "options": {}
  },
  "name": "Custom AI API",
  "type": "n8n-nodes-base.httpRequest"
}

Rate Limiting and Cost Control

API Rate Limiting

{
  "parameters": {
    "mode": "queue",
    "batchSize": 1,
    "concurrency": 1,
    "options": {
      "reset": false
    }
  },
  "name": "Rate Limiter",
  "type": "n8n-nodes-base.splitInBatches"
}

Cost Tracking

// Track API costs
const startTime = new Date();
const response = await $node.openAi.default.sendMessage({
  model: 'gpt-4o',
  messages: $input.item.json.messages
});

const endTime = new Date();
const duration = endTime - startTime;

// Estimate cost (rough calculation)
const inputTokens = response.usage.prompt_tokens;
const outputTokens = response.usage.completion_tokens;
const estimatedCost = (inputTokens * 0.00003 + outputTokens * 0.00006);

return [{
  json: {
    response: response.choices[0].message.content,
    usage: response.usage,
    estimated_cost: estimatedCost,
    processing_time: duration,
    model: response.model
  }
}];

Error Handling

Retry Logic

{
  "parameters": {
    "mode": "retry",
    "retryCount": 3,
    "retryInterval": 1000,
    "continueOnFail": true
  },
  "name": "Retry on Error",
  "type": "n8n-nodes-base.errorTrigger"
}

Fallback Workflows

{
  "nodes": [
    {
      "parameters": {
        "errorsToCatch": "all",
        "resume": "withDifferentBranch"
      },
      "name": "Catch Errors",
      "type": "n8n-nodes-base.errorTrigger"
    },
    {
      "parameters": {
        "model": "gpt-3.5-turbo",
        "messages": [
          {
            "role": "user",
            "content": "Simplify this request for a smaller model: {{ $json.original_question }}"
          }
        ]
      },
      "name": "Fallback Model",
      "type": "@n8n/n8n-nodes-langchain.openAi"
    }
  ]
}

Performance Optimization

Batch Processing

{
  "parameters": {
    "batchSize": 10,
    "options": {
      "merge": false
    }
  },
  "name": "Batch AI Requests",
  "type": "n8n-nodes-base.splitInBatches"
}

Caching

{
  "parameters": {
    "dataToSave": {
      "question": "={{ $json.question }}",
      "answer": "={{ $json.answer }}",
      "timestamp": "={{ new Date() }}"
    },
    "keys": {
      "question": "={{ $json.question }}"
    },
    "ttl": 86400
  },
  "name": "Cache Responses",
  "type": "@n8n/n8n-nodes-langchain.memoryBufferWindow"
}

Best Practices

  1. Credential Security: Store API keys securely, never in workflow JSON
  2. Error Handling: Implement comprehensive error handling and fallbacks
  3. Rate Limiting: Respect API limits and implement queuing
  4. Cost Monitoring: Track usage and set budget alerts
  5. Model Selection: Choose appropriate models based on task complexity
  6. Caching: Cache frequent queries to reduce API calls
  7. Testing: Thoroughly test workflows before production deployment
  8. Documentation: Document complex workflows and custom logic

These AI nodes provide powerful capabilities for building intelligent automations. The next chapter will explore document processing with AI.

What Problem Does This Solve?

Most teams struggle here because the hard part is not writing more code, but deciding clear boundaries for name, json, parameters so behavior stays predictable as complexity grows.

In practical terms, this chapter helps you avoid three common failures:

  • coupling core logic too tightly to one implementation path
  • missing the handoff boundaries between setup, execution, and validation
  • shipping changes without clear rollback or observability strategy

After working through this chapter, you should be able to reason about Chapter 2: AI Nodes and LLM Integration as an operating subsystem inside n8n AI Tutorial: Workflow Automation with AI, with explicit contracts for inputs, state transitions, and outputs.

Use the implementation notes around nodes, model, langchain as your checklist when adapting these patterns to your own repository.

How it Works Under the Hood

Under the hood, Chapter 2: AI Nodes and LLM Integration usually follows a repeatable control path:

  1. Context bootstrap: initialize runtime config and prerequisites for name.
  2. Input normalization: shape incoming data so json receives stable contracts.
  3. Core execution: run the main logic branch and propagate intermediate state through parameters.
  4. Policy and safety checks: enforce limits, auth scopes, and failure boundaries.
  5. Output composition: return canonical result payloads for downstream consumers.
  6. Operational telemetry: emit logs/metrics needed for debugging and performance tuning.

When debugging, walk this sequence in order and confirm each stage has explicit success/failure conditions.

Source Walkthrough

Use the following upstream sources to verify implementation details while reading this chapter:

  • View Repo Why it matters: authoritative reference on View Repo (github.com).
  • Awesome Code Docs Why it matters: authoritative reference on Awesome Code Docs (github.com).

Suggested trace strategy:

  • search upstream code for name and json to map concrete implementation paths
  • compare docs claims against actual runtime/config code before reusing patterns in production

Chapter Connections