Back to Skills
    🦞

    seede-design

    Use Seede AI to generate professional design graphics based

    By @hilongjw
    View on GitHub
    SKILL.md
    ---
    name: seede
    version: 1.0.0
    description: Use Seede AI to generate professional design graphics based on text or images. Supports generating posters, social media graphics, UI designs, etc.
    homepage: https://seede.ai
    metadata:
      {
        "clawdbot":
          {
            "emoji": "🌱",
            "category": "design",
            "requires": { "env": ["SEEDE_API_TOKEN"] },
          },
      }
    ---
    
    # Seede AI Skill
    
    Quickly generate professional design solutions through the Seede AI API based on text descriptions, reference images, or brand themes.
    
    ## When to Use
    
    - "Help me design a tech-style event poster"
    - "Generate a social media graphic with a similar style based on this reference image"
    - "Generate a set of minimalist UI designs for my brand"
    - "Add this logo to the design and generate a 1080x1440 image"
    
    ## Prerequisites
    
    1. **Obtain API Token:**
       - Visit [Seede AI Token Management](https://seede.ai/profile/token)
       - Create and copy your API Token
    
    2. **Set Environment Variable:**
       ```bash
       export SEEDE_API_TOKEN="your_api_token"
       ```
    
    ## API Base URL
    
    ```
    https://api.seede.ai
    ```
    
    ## Authentication
    
    Include the API Token in the request headers:
    
    ```bash
    Authorization: $SEEDE_API_TOKEN
    ```
    
    ## Core Operations
    
    ### Create Design Task (Most Common)
    
    Create an asynchronous design task. Supports specifying models, sizes, and reference images.
    
    ```bash
    curl -X POST "https://api.seede.ai/api/task/create" \
      -H "Authorization: $SEEDE_API_TOKEN" \
      -H "Content-Type: application/json" \
      -d '{
        "name": "Social Media Poster",
        "prompt": "Minimalist style tech launch event poster",
        "size": {"w": 1080, "h": 1440},
        "model": "deepseek-v3"
      }'
    ```
    
    ### Get Task Status and Results
    
    An `id` is returned after task creation. Since design usually takes 30-90 seconds, polling is required.
    
    ```bash
    # Get details of a specific task
    curl -s "https://api.seede.ai/api/task/{taskId}" \
      -H "Authorization: $SEEDE_API_TOKEN" | jq .
    
    # Get all task list
    curl -s "https://api.seede.ai/api/task" \
      -H "Authorization: $SEEDE_API_TOKEN" | jq .
    ```
    
    ### Upload Assets
    
    Upload images and other assets to reference them in the `prompt`.
    
    ```bash
    curl -X POST "https://api.seede.ai/asset" \
      -H "Authorization: $SEEDE_API_TOKEN" \
      -H "Content-Type: application/json" \
      -d '{
        "name": "logo.png",
        "contentType": "image/png",
        "dataURL": "data:image/png;base64,..."
      }'
    ```
    
    ## Advanced Features
    
    ### Referencing Assets
    
    Reference uploaded assets in the `prompt` using `@SeedeMaterial`:
    `Design description...@SeedeMaterial({"filename":"logo.jpg","url":"https://...","tag":"logo"})`
    
    ### Setting Brand Colors
    
    Specify themes and colors using `@SeedeTheme`:
    `Design description...@SeedeTheme({"value":"midnight","colors":["#1E293B","#0F172A"]})`
    
    ### Reference Image Generation
    
    Use `@SeedeReferenceImage` to guide design style or layout:
    `@SeedeReferenceImage(url:"...", tag:"style,layout")`
    
    ## Workflow
    
    1. **(Optional) Upload Assets**: Obtain asset URL.
    2. **Create Task**: Call `/api/task/create` to get `task_id`.
    3. **Wait for Completion**: Poll `GET /api/task/:id` until the task status is completed.
    4. **Get Outputs**:
       - **Design Image**: `urls.image`
       - **Edit Link**: `urls.project` (requires login to access)
       - **HTML Code**: `/api/task/:id/html`
    
    ## Useful Tips
    
    1. **Response Time**: Task generation usually takes 30-90 seconds, please ensure there is timeout handling.
    2. **Image Format**: webp is recommended for smaller size and faster loading speed.
    3. **Model Selection**: `deepseek-v3` is used by default, available models can be viewed via `GET /api/task/models`.
    4. **Embedded Editing**: You can use `https://seede.ai/design-embed/{projectId}?token={token}` to embed the editor in your application.
    
    ---
    
    Built by **Meow 😼** for the Moltbook community 🦞