n8n-nodes-siliconflow

n8n-nodes-siliconflow

An n8n community node to integrate with SiliconFlow AI models, providing access to chat completions, vision analysis, text embeddings, and document reranking capabilities.

Features

• 🤖 Chat Completions: Access to 20+ AI models including Qwen, GLM, DeepSeek, Hunyuan, and MiniMax
• 👁️ Vision Analysis: Analyze images with powerful vision-language models (Qwen2.5-VL, QVQ, DeepSeek-VL2)
• 🧮 Text Embeddings: Support for 9 embedding models including BGE, Qwen3-Embedding, and YoudAo
• 📊 Document Reranking: Rerank documents for better relevance with 6 reranking models
• 🎯 AI Agent Integration: Dedicated node for seamless integration with n8n AI Agent nodes
• 🔧 Flexible Output: Choose between simple (message-only) and detailed (with metadata) output modes
• ⚡ Advanced Parameters: Full support for reasoning models, streaming, and custom formats
• 🔐 Secure Authentication: API key-based authentication with configurable base URL

Supported Models

Chat Completion Models

• Qwen Series: QwQ-32B (reasoning), Qwen3-235B-A22B, Qwen3-32B, Qwen3-14B, Qwen3-8B, Qwen2.5 series
• GLM Series: GLM-Z1-32B, GLM-4-32B, GLM-Z1-Rumination-32B (reasoning), GLM-4-9B series
• DeepSeek: DeepSeek-V2.5, DeepSeek-R1 (reasoning)
• Others: Hunyuan-A13B-Instruct, MiniMax-M1-80k, QwenLong-L1-32B

Embedding Models

• BGE Series: bge-large-zh-v1.5, bge-large-en-v1.5, bge-m3 (multilingual)
• Qwen3-Embedding: 8B, 4B, 0.6B (up to 32K tokens)
• Others: YoudAo BCE, sentence-transformers models

Reranking Models

• Qwen3-Reranker: 8B, 4B, 0.6B
• BGE-Reranker: v2-m3, Pro/v2-m3
• YoudAo: bce-reranker-base_v1

Installation

Community nodes are not available in the npm registry. To install this node, you have the following options:

Option 1: User Installation via n8n Community Nodes

• Go to Settings > Community Nodes.
• Select Install.
• Enter n8n-nodes-siliconflow in Enter npm package name.
• Agree to the risks of using community nodes: select I understand the risks of installing unverified code from a public source.
• Select Install.

After installing the node, you can use it like any other node. n8n displays the node in search results in the Nodes panel.

Option 2: Manual installation

To get started install the package in your n8n root directory:

npm install n8n-nodes-siliconflow

For Docker-based deployments, add the following line before the font installation command in your n8n Dockerfile:

RUN cd /usr/local/lib/node_modules/n8n && npm install n8n-nodes-siliconflow

Configuration

You will need a SiliconFlow API key to use this node. You can get one from SiliconFlow.

• Create a SiliconFlow account at https://cloud.siliconflow.cn
• Generate an API key from your account settings
• In n8n, create a new SiliconFlow API credential with:
  • API Key: Your SiliconFlow API key
  • Base URL: https://api.siliconflow.cn/v1 (default)

Operations

The SiliconFlow node provides two types of nodes:

SiliconFlow Node

The main node supports three operations:

1. Chat Completion

Generate text completions using SiliconFlow's language models. Features include:

• Output Modes:
  • Simple: Returns only the message content as a string (ideal for chat workflows)
  • Detailed: Returns structured data with message, usage, metadata, and reasoning content
• Support for reasoning models (QwQ-32B, GLM-Z1-Rumination, DeepSeek-R1)
• Advanced parameters: temperature, top_p, top_k, min_p, frequency_penalty
• Thinking mode for chain-of-thought reasoning
• Streaming and custom response formats
• System, user, and assistant message roles

2. Text Embedding

Generate vector embeddings for text using various embedding models:

• Support for Chinese, English, and multilingual models
• Token limits from 512 to 32,768 depending on model
• Float or Base64 encoding formats
• Batch processing support

3. Document Reranking

Rerank documents based on relevance to a query:

• Support for multiple reranking algorithms
• Configurable top-N results
• Document chunking for long texts
• Relevance scoring

SiliconFlow Chat Model Node

A specialized node designed for AI Agent integration:

• AI Agent Compatibility: Can be used as a language model provider in n8n AI Agent nodes
• Seamless Integration: Works with official AI Agent, AI Tool, and AI Memory nodes
• Multi-turn Conversations: Supports complex dialogue scenarios
• Reasoning Support: Full support for reasoning models with thinking capabilities

Usage Examples

Basic Chat with Simple Output

{
  "resource": "chat",
  "operation": "complete",
  "model": "Qwen/QwQ-32B",
  "prompt": "Hello, how are you?",
  "outputMode": "simple"
}

Output: "Hello! I'm doing well, thank you for asking. How can I help you today?"

Chat with Detailed Output

{
  "resource": "chat", 
  "operation": "complete",
  "model": "Qwen/QwQ-32B",
  "prompt": "Explain quantum computing",
  "outputMode": "detailed"
}

Output:

{
  "message": "Quantum computing is a type of computation that harnesses quantum mechanical phenomena...",
  "model": "Qwen/QwQ-32B",
  "finishReason": "stop",
  "usage": {
    "prompt_tokens": 15,
    "completion_tokens": 150,
    "total_tokens": 165
  },
  "reasoning": "The user is asking about quantum computing..."
}

AI Agent Integration

• Add an AI Agent node to your workflow
• Add a SiliconFlow Chat Model node
• In the AI Agent configuration, select the SiliconFlow Chat Model as your language model
• Configure other AI Agent settings (tools, memory, etc.)
• Your AI Agent now uses SiliconFlow's powerful models!

Advanced Features

• Reasoning Models: Enable thinking mode for step-by-step problem solving
• Long Context: Support for models with up to 80K context length
• Multilingual: Chinese, English, and multilingual model options
• Flexible Input: Support for simple prompts or structured message arrays

Compatibility

Tested with n8n version 1.0+.

Resources

• n8n community nodes documentation
• SiliconFlow API documentation

License

MIT