DevAtlas AI GPU
The DevAtlas AI service API is served from this same host (/health, /ready, /v1/*). It is called by the Node backend, which authenticates with the X-AI-Service-Key header.
Isolated test for FLUX.2 Klein 4B FP8 loading and inference.
Edit an image with black-forest-labs/FLUX.2-klein-4b-fp8 through the same production provider used by the FastAPI API. One pipeline, one ZeroGPU allocation, and no model is loaded until a request arrives.