"""RunPod serverless backend: submit a job, read its status.""" from __future__ import annotations import httpx from .sim_backend import BackendError class RunPodError(BackendError): pass class RunPodClient: name = "runpod" FUNCTIONS = {"run_simulation", "sweep"} # what worker/Dockerfile's handler can run def __init__( self, api_key: str, endpoint_id: str, transport: httpx.BaseTransport | None = None, timeout: float = 30.0 ): if not api_key or not endpoint_id: raise RunPodError("RUNPOD_API_KEY and RUNPOD_ENDPOINT_ID must be set") self._client = httpx.Client( base_url=f"https://api.runpod.ai/v2/{endpoint_id}", headers={"Authorization": f"Bearer {api_key}"}, timeout=timeout, transport=transport, ) def supports(self, function: str) -> bool: return function in self.FUNCTIONS def submit(self, job_input: dict, execution_timeout_s: int, function: str = "run_simulation") -> dict: """POST /run. Returns {"id": ..., "status": "IN_QUEUE"}.""" if not self.supports(function): raise RunPodError(f"{function} is not available on RunPod (use SIM_BACKEND=modal)") payload = { "input": {**job_input, "function": function}, "policy": {"executionTimeout": int(execution_timeout_s * 1000)}, } try: r = self._client.post("/run", json=payload) except httpx.HTTPError as e: raise RunPodError(f"RunPod unreachable: {e}") from e if r.status_code != 200: raise RunPodError(f"RunPod /run returned HTTP {r.status_code}: {r.text[:300]}") data = r.json() if "id" not in data: raise RunPodError(f"unexpected RunPod response: {str(data)[:300]}") return data def status(self, job_id: str) -> dict: """GET /status/{id}: {"status", "output"?, "error"?, "executionTime"?, "delayTime"?}.""" try: r = self._client.get(f"/status/{job_id}") except httpx.HTTPError as e: raise RunPodError(f"RunPod unreachable: {e}") from e if r.status_code != 200: raise RunPodError(f"RunPod /status returned HTTP {r.status_code}: {r.text[:300]}") return r.json()