Runfra

An API that routes AI image and vision tasks to the appropriate model and cheapest capable backend, classifying requests, dispatching to GPU workers, scoring outputs against a quality threshold, and retrying failures to return the best result. It serves developers and B2B teams building image generation features, and marketplace sellers needing product assets via a no-code studio. It is positioned as a routing layer that uses owned distributed GPUs when capable and external providers only when needed, delivered as a single API with no infrastructure to manage.

Key features

• Classify and route tasks to models
• Cheapest capable backend selection
• Parallel batch generation
• Quality scoring and threshold filtering
• Automatic retry on failure
• Best result selection and return
• Prompt expansion for generation
• 4x Real-ESRGAN upscaling
• 8x gallery print upscaling
• Print-ready DPI export
• Async POST and GET job model
• API key authentication