[borzunov]

There's a lightweight HTTP API for inference: https://github.com/borzunov/chat.petals.ml#http-api-methods