n8n-nodes-universal-llm-vision

Add vision capabilities to your n8n workflows - Analyze images with AI using any LLM provider, with flexible integration options for every use case.

Installation

Install via n8n's community node interface:

1. Open n8n in your browser
2. Go to Settings > Community Nodes
3. Search for n8n-nodes-universal-llm-vision and click Install

Both nodes will be available in your node palette under the "AI" category.

Choose Your Approach

This package provides two nodes with different integration approaches:

🎯 Universal LLM Vision

Custom credentials node - Connect to any OpenAI-compatible vision API with your own credentials.

• ✅ Any custom provider: Configure your own API endpoints and credentials
• ✅ Vision model discovery: Auto-fetch vision-capable models from models.dev with pricing
• ✅ Full API control: Custom headers, parameters, JSON response format
• ✅ Rich metadata: Token usage, costs, and model info in output
• ✅ AI Agent ready: Use as a tool in AI Agent workflows

Best for: Production workflows, custom APIs, full parameter control

🔗 Vision Chain

Langchain integration node - Reuse your existing n8n chat model connections (OpenAI, Anthropic, OpenRouter, etc.).

• ✅ Reuse chat models: Connect any n8n chat model node you already have configured
• ✅ Simpler setup: No need to duplicate credentials
• ✅ Quick switching: Change models by swapping the connected chat model node

⚠️ Note: Vision Chain cannot be used as a tool in AI Agents. For AI Agent tool integration, use Universal LLM Vision instead.

Important Note

Both nodes only work with vision-capable models. Regular text-only models are not supported. Most modern multimodal models from OpenAI (GPT-4o), Anthropic (Claude Sonnet), Google (Gemini), and OpenRouter support vision.

Common features:

• Binary data, URL, and base64 image sources
• Customizable prompts with intelligent defaults
• Auto/Low/High detail control for cost optimization
• Production-ready with 241 tests

📖 Read the detailed comparison in the Vision Chain documentation

Quick Start

Vision Chain (Fastest Setup)

Perfect for getting started with LLM Vision:

1. Add any Chat Model node (e.g., OpenAI Chat Model)
2. Add Vision Chain node
3. Connect: Chat Model → Vision Chain (Chat Model input)
4. Configure image source and prompt
5. Done! ✨

[Your Data] → [Vision Chain] → [Next Node]
                    ↑
              [Chat Model]

📖 Complete workflow example

Universal LLM Vision (Full Control)

For production workflows with specific requirements:

1. Add Universal LLM Vision node
2. Configure credentials (provider + API key)
3. Select from available vision models
4. Configure image source and prompt
5. Customize parameters, headers, system prompt as needed
6. Done! ✨

[Your Data] → [Universal LLM Vision] → [Next Node]

📖 Complete workflow example

Use Cases

Universal LLM Vision - Best for:

• 🏭 Production pipelines: Batch image processing with metadata tracking
• 📊 Custom APIs: Integration with proprietary vision models
• 🔍 Structured extraction: OCR with JSON mode for invoices, receipts, forms
• 🎯 Full control workflows: Custom headers, parameters, response formats

Vision Chain - Best for:

• 🤖 AI Agents: Customer support bots, visual Q&A assistants
• ⚡ Rapid prototyping: Quick model testing and switching
• 🔄 Dynamic workflows: Model selection based on conditions
• 🔗 Multi-step analysis: Chaining different models for specialized tasks

Both nodes - Common uses:

• Product catalog descriptions and quality inspection
• Document processing (text extraction, handwriting recognition)
• Specialized analysis (medical, architectural, fashion)
• Scene understanding and object detection

Detailed Configuration

Universal LLM Vision Node

Providers & Models

Supported Providers:

• OpenAI (GPT-4o, GPT-4 Turbo with Vision)
• Google Gemini (Flash, Pro Vision)
• Anthropic (Claude Sonnet, Opus with Vision)
• OpenRouter (vision models from multiple providers)
• Groq (Llama Vision, Mixtral Vision)
• Grok/X.AI (Grok Vision)
• Custom (any OpenAI-compatible vision API)

Model Selection:
The node auto-fetches all vision-capable models from models.dev, displaying:

• Model name
• Pricing (input/output per 1M tokens)
• Model ID in parentheses (e.g., $2.5 / $10 per 1M tokens (gpt-4o))

Tested Models:

• OpenAI: GPT 5, GPT 4.1, GPT 4o
• Google: Gemini 2.5 Flash Lite, Gemini 3.0 Flash
• OpenRouter: Gemma 3 27B, GLM 4.6V, Ministral 3, Nemotron VL, Qwen3 VL
• Grok/X.AI: Grok 4.1 Fast

Credentials

1. Select your provider
2. Enter API key
3. (Optional) Custom base URL for custom providers

For custom OpenAI-compatible APIs:

• Select "Custom Provider"
• Provide Base URL (e.g., https://your-api.com/v1 or http://localhost:11434/v1 for Ollama)
• API Key is optional for local providers like Ollama - leave empty if not needed
• The node will attempt to auto-fetch available models
• Use Manual Model ID if auto-fetch fails

Parameters

Required:

• Model: Select from dropdown or enter manually
• Image Source: Binary Data / URL / Base64
• Prompt: Your analysis instruction

Optional (Model Parameters):

• Temperature (0-2): Creativity level
• Max Tokens: Response length limit
• Top P: Nucleus sampling parameter

Advanced Options:

• System Prompt: Guide model behavior (intelligent default provided)
• Response Format: Text or JSON
• Custom Headers: Add custom HTTP headers
• Additional Parameters: Provider-specific parameters
• Manual Model ID: Override model selection
• Output Property Name: Where to store result (default: analysis)
• Include Metadata: Add usage stats and token counts

Vision Chain Node

Setup

1. Add Chat Model: Any n8n chat model (OpenAI, Anthropic, etc.)
2. Connect to Vision Chain: Use the "Chat Model" input
3. Configure Image Source: Binary / URL / Base64
4. Write Prompt: Your analysis instruction

Parameters

Required:

• Image Source: Binary Data / URL / Base64
• Prompt: Analysis instruction

Options:

• Image Detail: Auto / Low / High (affects cost and quality)
• System Prompt: Comprehensive default for image understanding (customizable)
• Output Property Name: Configure result property (default: analysis)

Note: Temperature, max tokens, and model selection are configured at the chat model node level.

Examples & Workflows

📥 Example Files

• example-workflow.json - Complete Universal LLM Vision workflow
• example-workflow-chain.json - Vision Chain with AI Agent

Universal LLM Vision Examples

Image Analysis from URL

Webhook → Set (image URL) → Universal LLM Vision → Respond

Perfect for: Product catalog automation, web scraping with analysis

Batch Image Processing

Read Binary Files → Universal LLM Vision → IF → Split → [Process Results]

Perfect for: Quality control, document classification, content moderation

OCR with Structured Output

HTTP Request (get image) → Universal LLM Vision (JSON mode) → Set → Database

Configure:

• Response Format: JSON
• Prompt: "Extract text with structure: {title, date, amount, items: []}"

Vision Chain Examples

Image Analysis with Chat Models

[Data] → Vision Chain → [Process Results]
             ↑
        [Chat Model]

Perfect for: Image classification, visual Q&A, content moderation

Dynamic Model Switching

[Data] → Vision Chain → [Output]
             ↑
        [Different Chat Models based on conditions]

Perfect for: Cost optimization, fallback strategies, A/B testing

Image Analysis Pipeline

Download File → Vision Chain (describe) → Vision Chain (extract text) → Process
                     ↑                         ↑
                [Model A]                  [Model B]

Perfect for: Multi-step analysis, combining different model strengths

Development & Contributing

This package is built using the n8n-community-node-starter boilerplate, providing:

• Robust programmatic node architecture
• Comprehensive testing framework (Jest)
• CI/CD pipelines
• AI-assisted development tools

Contributions are welcome! Feel free to:

• 🐛 Report bugs by opening an issue
• 💡 Suggest features or improvements
• 🔧 Submit pull requests with fixes or enhancements

License

MIT License - see LICENSE file for details.

Links

• n8n Documentation
• Community Nodes Guide
• n8n-community-node-starter - The boilerplate this node is based on