spraay_gpu_run
Run AI model inference on GPU. Supports image generation, video, LLMs, audio transcription, and utilities. Use shortcuts or model IDs. Pay $0.05 USDC per request.
Instructions
Run AI model inference on GPU via Replicate. Supports image generation (flux-pro, sdxl, ideogram), video generation (wan-video, minimax-video), LLMs (llama-70b, llama-8b, mixtral), audio (whisper transcription, musicgen), and utilities (esrgan upscaling, rembg background removal). Use shortcuts like 'flux-pro' or full model IDs like 'owner/model'. Returns output directly for fast models, or a poll URL for longer jobs. Costs $0.05 USDC.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| input | Yes | Model-specific input parameters. Image models: { prompt: '...' }. LLMs: { prompt: '...' }. Whisper: { audio: 'https://...' }. ESRGAN: { image: 'https://...' }. | |
| model | Yes | Model shortcut (flux-pro, sdxl, llama-70b, whisper, esrgan, etc.) or full Replicate model ID (owner/model-name). Use spraay_gpu_models to list all shortcuts. | |
| version | No | Specific model version hash (optional — not needed for official models) | |
| webhook | No | Webhook URL for async result delivery (optional) |
Output Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| ok | Yes | True when the gateway call succeeded; false when it returned an error. | |
| data | No | The gateway response payload on success. The exact shape depends on the tool (see the tool description and the JSON in the text content block). | |
| error | No | Human-readable error message, present only when ok is false. |