AI & LLM Nodes

AI nodes bring the power of large language models and AI capabilities to your workflows. Generate text, analyze images, transcribe audio, create embeddings, and route decisions with AI.

LLM Node

The core AI node for text generation and reasoning.

Providers & Models

Provider	Models
OpenAI	gpt-4.1, gpt-4.1-mini, gpt-4.1-nano, gpt-4o, gpt-4o-mini, o4-mini, o3, o3-mini
Anthropic	claude-opus-4-5, claude-sonnet-4-5, claude-haiku-4-5
Google	gemini-3-pro-preview, gemini-3-flash-preview, gemini-2.5-pro, gemini-2.5-flash
Cohere	command-a-03-2025, command-r-plus, command-r
Hugging Face	Llama-4-Scout, Llama-4-Maverick, Qwen-3-235B, DeepSeek-R1
x.ai	grok-3, grok-3-fast, grok-2-vision

Configuration

{
  provider: "openai",
  model: "gpt-4.1",
  systemPrompt: "You are a helpful assistant that...",
  prompt: "{{trigger.body.question}}",
  temperature: 0.7,
  maxTokens: 2048,
  topP: 1.0,
  outputVariable: "llm_response"
}

Field	Description
provider	AI provider to use
model	Specific model
systemPrompt	System instructions (optional)
prompt	User prompt (supports variables)
temperature	Creativity (0-2, lower = deterministic)
maxTokens	Maximum response length
topP	Nucleus sampling parameter
outputVariable	Variable name for result

Extended Thinking

Some models support extended thinking/reasoning:

{
  provider: "anthropic",
  model: "claude-sonnet-4-5",
  enableThinking: true,
  thinkingBudget: 10000  // Max thinking tokens
}

Supported models:

• OpenAI: o3, o3-mini, gpt-4.1
• Anthropic: claude-opus-4-5, claude-sonnet-4-5
• Google: gemini-3-pro, gemini-3-flash
• Hugging Face: DeepSeek-R1

Output

{
  response: "The answer to your question is...",
  usage: {
    inputTokens: 150,
    outputTokens: 450,
    totalTokens: 600
  }
}

---

Vision Node

Analyze images with AI vision models.

Providers

Provider	Models
OpenAI	gpt-4o, gpt-4o-mini
Anthropic	claude-opus-4-5, claude-sonnet-4-5
Google	gemini-3-pro, gemini-3-flash
x.ai	grok-2-vision

Operations

Operation	Description
analyze	Analyze and describe image content
generate	Generate images (provider-dependent)

Configuration

{
  provider: "openai",
  model: "gpt-4o",
  operation: "analyze",
  imageInput: "{{files_node.output.file}}",  // URL or base64
  prompt: "Describe what you see in this image",
  maxTokens: 1000,
  outputVariable: "vision_result"
}

Use Cases

• Product image analysis
• Document OCR and extraction
• Visual quality inspection
• Chart and graph interpretation
• UI screenshot analysis

---

Audio Input Node (Speech-to-Text)

Transcribe audio to text.

Providers

Provider	Models
OpenAI	whisper-1
Deepgram	nova-2, nova, enhanced, base

Configuration

{
  provider: "openai",
  model: "whisper-1",
  inputName: "audio_file",
  language: "en",        // Optional, auto-detected
  punctuate: true,
  diarize: false,        // Speaker identification
  outputVariable: "transcription",
  label: "Upload Audio",
  description: "Audio file to transcribe"
}

Supported Formats

• MP3, MP4, MPEG, MPGA
• WAV, WEBM
• M4A, OGG, FLAC

Output

{
  text: "Hello, this is the transcribed audio...",
  language: "en",
  duration: 45.2,
  segments: [...]  // With timestamps if requested
}

---

Audio Output Node (Text-to-Speech)

Generate speech from text.

Providers

Provider	Voices
OpenAI	alloy, echo, fable, onyx, nova, shimmer
ElevenLabs	100+ voices, custom voice cloning
Deepgram	aura, aura-asteria, aura-luna

Configuration

{
  provider: "elevenlabs",
  voice: "rachel",
  textInput: "{{llm_node.output.response}}",
  speed: 1.0,
  stability: 0.5,
  similarityBoost: 0.75,
  outputFormat: "mp3",
  returnAsUrl: true,
  outputVariable: "audio_output"
}

Output Options

Option	Description
returnAsUrl: true	Returns URL to audio file
returnAsUrl: false	Returns base64-encoded audio

---

Audio Transcription Node

Advanced transcription with more options.

Configuration

{
  audioSource: "{{input.audio_url}}",
  task: "transcribe",      // or "translate"
  language: "en",
  outputFormat: "json",    // text, json, srt, vtt
  timestamps: true,
  prompt: "Technical discussion about...",
  temperature: 0.2,
  outputVariable: "transcript"
}

Tasks

Task	Description
transcribe	Transcribe in original language
translate	Transcribe and translate to English

---

Embeddings Node

Generate vector embeddings for semantic search.

Providers

Provider	Models
OpenAI	text-embedding-3-small, text-embedding-3-large
Cohere	embed-english-v3.0, embed-multilingual-v3.0
Google	text-embedding-004

Configuration

{
  provider: "openai",
  model: "text-embedding-3-small",
  input: "{{node.output.text}}",
  batchMode: false,
  outputVariable: "embeddings"
}

Batch Mode

For multiple texts:

{
  batchMode: true,
  input: ["Text 1", "Text 2", "Text 3"]
}

Output

{
  embeddings: [0.0123, -0.0456, 0.0789, ...],  // 1536 dimensions
  model: "text-embedding-3-small",
  usage: { totalTokens: 15 }
}

---

Router Node

Use AI to classify inputs and route to different paths.

Configuration

{
  provider: "openai",
  model: "gpt-4o-mini",
  systemPrompt: "Classify the user's intent",
  prompt: "{{trigger.body.message}}",
  routes: [
    { value: "sales", label: "Sales inquiry" },
    { value: "support", label: "Technical support" },
    { value: "billing", label: "Billing question" },
    { value: "other", label: "Other/unknown" }
  ],
  defaultRoute: "other",
  temperature: 0.3,
  outputVariable: "route"
}

How It Works

1. LLM analyzes the input
2. Classifies into one of the defined routes
3. Workflow branches based on classification

Output

{
  route: "support",
  confidence: 0.92,
  reasoning: "The user mentioned a technical error..."
}

Use Cases

• Intent classification for chatbots
• Ticket routing
• Content categorization
• Lead qualification

---

Knowledge Base Query Node

Search your knowledge bases with semantic similarity.

Configuration

{
  knowledgeBaseId: "kb_abc123",
  query: "{{user.question}}",
  topK: 5,
  threshold: 0.7,
  outputVariable: "search_results"
}

Field	Description
knowledgeBaseId	Target knowledge base
query	Search query
topK	Number of results
threshold	Minimum similarity (0-1)

Output

{
  results: [
    {
      content: "Relevant chunk text...",
      score: 0.92,
      metadata: {
        documentId: "doc_123",
        fileName: "guide.pdf",
        pageNumber: 5
      }
    },
    // ... more results
  ],
  totalMatches: 5
}

---

OCR Extraction Node

Extract text from images using OCR.

Configuration

{
  imageSource: "{{input.image_url}}",
  languages: ["en", "es"],
  psm: 3,                  // Page segmentation mode
  outputFormat: "text",    // text, json, hocr
  confidenceThreshold: 0.6,
  preprocessing: {
    deskew: true,
    denoise: true,
    contrast: "auto"
  },
  outputVariable: "ocr_result"
}

Output

{
  text: "Extracted text content...",
  confidence: 0.94,
  blocks: [
    {
      text: "Header text",
      bbox: { x: 10, y: 10, width: 200, height: 30 },
      confidence: 0.98
    }
  ]
}

---

Image Generation Node

Generate images with AI.

Providers

Provider	Models
OpenAI	dall-e-3, dall-e-2
Replicate	flux-pro, flux-schnell
Stability AI	stable-diffusion-3, sdxl
FAL	Various models

Operations

Operation	Description
generate	Create new image from prompt
inpaint	Edit parts of an image
outpaint	Extend image beyond borders
upscale	Increase image resolution
removeBackground	Remove image background
styleTransfer	Apply style to image

Configuration

{
  provider: "openai",
  model: "dall-e-3",
  operation: "generate",
  prompt: "A serene mountain landscape at sunset",
  negativePrompt: "blurry, low quality",
  size: "1024x1024",
  quality: "hd",
  style: "vivid",
  n: 1,
  outputFormat: "url",
  outputVariable: "generated_image"
}

---

Video Generation Node

Generate videos with AI.

Providers

Provider	Models
Google Veo	veo-2
Runway	gen-3
Luma	dream-machine
Replicate	Various

Configuration

{
  provider: "runway",
  model: "gen-3",
  prompt: "A cat playing piano",
  imageInput: "{{input.reference_image}}",  // Optional
  duration: 4,
  aspectRatio: "16:9",
  loop: false,
  outputFormat: "mp4",
  outputVariable: "video_output"
}

---

Best Practices

Model Selection

Task	Recommended
Complex reasoning	claude-opus-4-5, o3, gpt-4.1
Fast responses	gpt-4o-mini, claude-haiku-4-5, gemini-flash
Vision tasks	gpt-4o, claude-sonnet-4-5
Code generation	claude-sonnet-4-5, gpt-4.1
Cost-sensitive	gpt-4o-mini, gemini-flash-lite

Temperature Guidelines

Use Case	Temperature
Factual Q&A	0.0 - 0.3
Code generation	0.2 - 0.4
General assistant	0.5 - 0.7
Creative writing	0.7 - 1.0
Brainstorming	0.9 - 1.2

Error Handling

Always handle potential AI failures:

// Use conditional node after LLM
condition: "{{llm_node.output.error}} == null";

Cost Optimization

• Use smaller models for simple tasks
• Set appropriate maxTokens limits
• Cache responses when possible
• Use embeddings for semantic search vs. LLM comparison