SiliconFlow

Overview

SiliconFlow provides high-quality GenAI services based on excellent open-source foundation models. You can use SiliconFlow in Dify through consolidated image generation, image editing, and video generation tools backed by current SiliconFlow models.

Configuration

1. Apply for SiliconCloud API Key

Create a new API Key on the SiliconCloud API management page and ensure that you have sufficient balance. SiliconFlow offers two separate endpoints:

• China endpoint: cloud.siliconflow.cn
• International endpoint: cloud.siliconflow.com

Note: These are separate account systems. Make sure to select the matching endpoint option when configuring the plugin in Dify.

2. Get SiliconFlow tools from Plugin Marketplace

The SiliconFlow tools could be found at the Plugin Marketplace, please install it first.

3. Fill in the configuration in Dify

On the Dify navigation page, click and fill in the API Key.

• API Key: Your SiliconCloud API Key
• Use International Endpoint (optional): Choose whether to use the international endpoint () or China endpoint (, default). Select "True" if you created your account on the international site, "False" for the China site.

4. Using the tools

You can use the SiliconFlow tool in the following application types:

Chatflow / Workflow applications

Both Chatflow and Workflow applications support the SiliconFlow tool node. After adding it, fill in "Input Variables → Prompt" with variables that reference the user's prompt or the content generated by the previous node. Then reference the structured output from the SiliconFlow tool in the "End" node.

Agent applications

Add the SiliconFlow tool in the Agent application, then send an image-generation, image-editing, or video-generation prompt in the dialog box to call the corresponding unified tool.

5. Features

Image Generation

Image Generate

The tool uses the current API and supports both text-to-image and image-to-image generation.

• Parameters:

  • : Choose between and ; this choice determines which model-specific parameters are effective
  • (required): Text prompt used to generate the image
  • : Optional input image URL or base64 data for image-to-image generation; leave empty for text-to-image
  • : Optional content to avoid in the image, supported by both models
  • : Model-dependent recommended sizes; pick a size that matches the selected model
  • : 1-4 images per generation, only effective for Kolors
  • : 0-20 (default: 7.5), only effective for Kolors
  • : 0.1-20 (default: 4), only effective for Qwen Image
  • : 1-100 (default: 20), supported by both models
  • : Optional parameter for reproducible results, supported by both models

• Available Models:

Image Edit

The tool uses the current API for image editing.

• Parameters:

  • : Choose between and
  • (required): Text prompt describing the target edit result
  • : 1-100 (default: 20)
  • : 0.1-20 (default: 4)
  • : Optional parameter for reproducible results
  • (required): Required first input image URL or base64 data
  • / : Optional additional input images for

• Available Models:

Video Generation

Video Generate

The tool uses SiliconFlow's current and workflow, supporting both text-to-video and image-to-video depending on the selected model.

• Parameters:

  • (required): Detailed text description for video generation
  • : Choose between and
  • : Required when model is , ignored for text-to-video
  • : Supported sizes , , and
  • : Optional parameter for reproducible results
  • : Control what you don't want in the video

• Available Models:

(May take 4 minutes to generate 1 video)

About SiliconFlow

SiliconFlow is committed to building a scalable, standardized, and high-performance AI Infra platform. It offers SiliconCloud (the model cloud service platform), SiliconLLM (the LLM inference engine), and OneDiff (the high-performance text-to-image/video acceleration library). These solutions help enterprises and individual users deploy AI models efficiently and cost-effectively.

Website | SiliconCloud | Discord | X