W3 Hyperbolic Action

AI inference and GPU compute via Hyperbolic — text, image, audio, vision, and on-demand GPUs.

Quick Start

- name: Ask a question
  id: ai
  uses: w3-io/w3-hyperbolic-action@v0
  with:
    command: chat
    api-key: ${{ secrets.HYPERBOLIC_API_KEY }}
    model: 'deepseek-ai/DeepSeek-V3'
    messages: '[{"role":"user","content":"What is Ethereum?"}]'

Commands

Command	Description
chat	Text generation with 13+ open-source models
generate-image	Image generation with Stable Diffusion, Flux.1, LoRA
generate-audio	Text-to-speech with multi-language support
analyze-image	Image understanding via vision-language models
rent-gpu	Rent on-demand GPU instances
list-gpus	List available GPU types and pricing
terminate-gpu	Terminate a rented GPU instance

Inputs

Input	Required	Default	Description
command	Yes	—	Operation to perform (see Commands)
api-key	Yes	—	Hyperbolic API key
api-url	No	https://api.hyperbolic.xyz	API base URL override
model	No	—	Model ID (e.g. deepseek-ai/DeepSeek-V3)
messages	No	—	Chat messages as JSON array
temperature	No	—	Sampling temperature (0.0-2.0)
top-p	No	—	Nucleus sampling threshold (0.0-1.0)
max-tokens	No	—	Maximum tokens to generate
response-format	No	—	Response format as JSON
seed	No	—	Random seed for reproducibility
stop	No	—	Comma-separated stop sequences
tools	No	—	Tool definitions as JSON array for function calling
prompt	No	—	Text prompt for image generation or vision question
height	No	1024	Image height in pixels
width	No	1024	Image width in pixels
steps	No	25	Number of diffusion steps
lora	No	—	LoRA style adapter name (e.g. Pixel_Art, Logo)
lora-weight	No	0.8	LoRA influence weight (0.0-1.0)
text	No	—	Text to convert to speech
language	No	EN	TTS language: EN, ES, FR, ZH, JP, KR
speaker	No	—	TTS speaker/accent (e.g. EN-US, EN-BR)
speed	No	1.0	TTS speed multiplier (0.1-5.0)
image-url	No	—	URL of image to analyze
image-base64	No	—	Base64-encoded image to analyze
gpu-type	No	—	GPU type for rental (e.g. H100-SXM, A100-SXM)
gpu-count	No	1	Number of GPUs (1, 2, 4, or 8)
instance-id	No	—	Instance ID for terminate-gpu

Outputs

Output	Description
result	JSON result of the operation

Authentication

Get an API key from Hyperbolic and store it as HYPERBOLIC_API_KEY in your repository secrets.