One endpoint for every machine you own.
Roundhouse turns every GPU, NPU and iGPU you own into one OpenAI- and Anthropic-compatible endpoint, with the cloud as one more place a request can run. It decides where each request goes, and tells you why.
● You're on the list.
We'll write to once, on launch day.
$ rh init
✓ found halo Ryzen AI Max, 96 GB unified
✓ found tower 2 × Radeon dGPU
✓ found laptop iGPU + NPU
✓ found studio Apple M4 Max
✓ serving on 127.0.0.1:8080/v1
$ rh why req_7f2a
coder → tower · private, fits, least busy
$ rhfa --ask "can you deploy the decision model
https://huggingface.co/LiquidAI/d1-Omni-2B please?"
· capacity laptop NPU idle, halo busy, tower busy
✓ fit check 2B weights fit laptop NPU, room to spare
✓ canary passed, 0 errors
✓ serving at 127.0.0.1:8080/v1/decisions
wire protocol: OpenAI Decisions