Skip to content

API.md

Latest commit

 

History

History
238 lines (168 loc) · 5.87 KB

API.md

File metadata and controls

238 lines (168 loc) · 5.87 KB
Raw

SweetDesk API Documentation

Backend Go API (Wails Bindings)

The Go backend provides the following methods that can be called from the frontend via Wails bindings:

Image Search

SearchImages(query string, page int, perPage int) ([]ImageResult, error)

Search for images using the Pixabay API.

Parameters:

  • query: Search query string (e.g., "nature", "mountains")
  • page: Page number for pagination (starting from 1)
  • perPage: Number of results per page (max 100)

Returns:

  • Array of ImageResult objects containing image metadata
  • Error if the search fails

Example:

const images = await window.go.main.App.SearchImages("mountains", 1, 12);

Image Download

DownloadImage(imageURL string) (string, error)

Download an image from a URL and return it as base64.

Parameters:

  • imageURL: Full URL to the image to download

Returns:

  • Base64-encoded image data
  • Error if download fails

Example:

const base64Data = await window.go.main.App.DownloadImage(imageURL);

Image Classification

ClassifyImage(base64Data string) (string, error)

Classify an image as "anime", "photo", or "art".

Parameters:

  • base64Data: Base64-encoded image data

Returns:

  • Classification type: "anime", "photo", or "art"
  • Error if classification fails

Note: Falls back to "photo" if Python bridge is not available.

Example:

const imageType = await window.go.main.App.ClassifyImage(base64Data);

Image Upscaling

UpscaleImage(base64Data string, imageType string, scale int) (string, error)

Upscale an image using the appropriate AI model.

Parameters:

  • base64Data: Base64-encoded image data
  • imageType: Type of image ("anime" or "photo")
  • scale: Upscaling factor (2, 4, or 8)

Returns:

  • Base64-encoded upscaled image data
  • Error if upscaling fails

Models Used:

  • Anime: RealCUGAN-pro
  • Photo: Real-ESRGAN (4xLSDIR)

Example:

const upscaled = await window.go.main.App.UpscaleImage(base64Data, "photo", 4);

Full Image Processing

ProcessImage(base64Data string, targetResolution string, useSeamCarving bool) (string, error)

Complete processing pipeline: classify, upscale, and adjust aspect ratio.

Parameters:

  • base64Data: Base64-encoded input image data
  • targetResolution: Target resolution ("4K", "5K", or "8K")
  • useSeamCarving: Use content-aware resize (true) or simple crop (false)

Returns:

  • Base64-encoded processed image data
  • Error if processing fails

Processing Steps:

  1. Classify image type (anime vs photo)
  2. Upscale using appropriate model
  3. Adjust aspect ratio to 16:9
  4. Return final result

Example:

const result = await window.go.main.App.ProcessImage(base64Data, "4K", true);

Data Structures

ImageResult

interface ImageResult {
    id: string;          // Unique image ID
    url: string;         // Image page URL
    downloadURL: string; // Direct download URL
    previewURL: string;  // Thumbnail preview URL
    width: number;       // Original width in pixels
    height: number;      // Original height in pixels
    author: string;      // Author/photographer name
    source: string;      // Source platform (e.g., "Pixabay")
    tags: string[];      // Array of tags
}

Environment Variables

Required

Optional

  • SWEETDESK_DEBUG: Enable debug logging (set to "1")
  • SUPABASE_URL: Supabase project URL (for cloud storage)
  • SUPABASE_KEY: Supabase anonymous key

Error Handling

All methods return errors that should be handled in the frontend:

try {
    const result = await window.go.main.App.ProcessImage(data, "4K", false);
    // Handle success
} catch (error) {
    console.error("Processing failed:", error);
    // Show error to user
}

Performance Notes

Processing Times (Estimated)

Operation Time (M1 Pro) Time (Intel i7)
Classification ~0.5s ~1s
Upscale 2x ~15s ~25s
Upscale 4x ~45s ~60s
Upscale 8x ~120s ~180s
Seam Carving ~10s ~15s

Memory Usage

  • Original image: ~10-20 MB
  • 4K upscaled: ~30-50 MB
  • All processing is in-memory (no disk I/O until save)

Adding New API Providers

To add support for additional image APIs (Unsplash, Pexels, etc.):

  1. Implement the APIProvider interface in Go:
type APIProvider interface {
    Search(query string, options SearchOptions) ([]ImageResult, error)
    Download(imageURL string) ([]byte, error)
    GetName() string
}

  1. Create a new provider struct (e.g., UnsplashProvider)

  2. Register it in app.go startup function

  3. Update frontend to allow provider selection

Python Scripts

classify_image.py

Classifies images using heuristic-based analysis. For production, consider using:

  • DeepGHS/imgutils
  • TensorFlow models
  • PyTorch classifiers

seam_carving.py

Content-aware image resizing. For production, consider using:

  • seam-carving library (faster)
  • OpenCV implementations
  • GPU-accelerated versions

Binary Distribution

AI upscaler binaries should be downloaded from official sources:

Place binaries in:

binaries/
├── darwin/     # macOS
├── linux/      # Linux
└── windows/    # Windows

License Compliance

Component License Commercial Use
Real-ESRGAN-ncnn-vulkan MIT-like ✅ Yes
RealCUGAN-ncnn-vulkan MIT-like ✅ Yes
DeepGHS/imgutils MIT ✅ Yes
seam-carving MIT ✅ Yes
Pixabay API Free ✅ Yes (with attribution)

Note: If using Universal-NCNN-Upscaler instead, be aware of AGPL-3.0 requirements.