more-compute 0.4.3__py3-none-any.whl → 0.5.0__py3-none-any.whl

morecompute/services/providers/runpod_provider.py

@@ -0,0 +1,504 @@
+"""RunPod GPU cloud provider implementation."""
+
+from typing import Any
+from datetime import datetime
+import httpx
+from fastapi import HTTPException
+
+from .base_provider import BaseGPUProvider, NormalizedPod
+from .provider_factory import register_provider
+from ...models.api_models import PodResponse
+
+
+@register_provider
+class RunPodProvider(BaseGPUProvider):
+    """RunPod GPU cloud provider using GraphQL API."""
+
+    PROVIDER_NAME = "runpod"
+    PROVIDER_DISPLAY_NAME = "RunPod"
+    API_KEY_ENV_NAME = "RUNPOD_API_KEY"
+    SUPPORTS_SSH = True
+    DASHBOARD_URL = "https://www.runpod.io/console/user/settings"
+
+    BASE_URL = "https://api.runpod.io/graphql"
+
+    def __init__(self, api_key: str | None = None):
+        super().__init__(api_key)
+
+    def _get_auth_headers(self) -> dict[str, str]:
+        """Get RunPod authentication headers."""
+        return {
+            "Content-Type": "application/json",
+            "Authorization": f"Bearer {self.api_key}" if self.api_key else ""
+        }
+
+    async def _graphql_request(
+        self,
+        query: str,
+        variables: dict[str, Any] | None = None
+    ) -> dict[str, Any]:
+        """Make a GraphQL request to RunPod API.
+
+        Args:
+            query: GraphQL query string
+            variables: Query variables
+
+        Returns:
+            Response data
+
+        Raises:
+            HTTPException: On API errors
+        """
+        async with httpx.AsyncClient() as client:
+            try:
+                response = await client.post(
+                    self.BASE_URL,
+                    headers=self._get_auth_headers(),
+                    json={
+                        "query": query,
+                        "variables": variables or {}
+                    },
+                    timeout=30.0
+                )
+                response.raise_for_status()
+                result = response.json()
+
+                if "errors" in result:
+                    error_msg = result["errors"][0].get("message", "Unknown error")
+                    raise HTTPException(
+                        status_code=400,
+                        detail=f"RunPod API error: {error_msg}"
+                    )
+
+                return result.get("data", {})
+            except httpx.HTTPStatusError as e:
+                raise HTTPException(
+                    status_code=e.response.status_code,
+                    detail=f"RunPod API error: {e.response.text}"
+                )
+            except httpx.RequestError as e:
+                raise HTTPException(
+                    status_code=503,
+                    detail=f"RunPod connection error: {str(e)}"
+                )
+
+    async def get_gpu_availability(
+        self,
+        regions: list[str] | None = None,
+        gpu_count: int | None = None,
+        gpu_type: str | None = None,
+        secure_cloud: bool | None = None,
+        community_cloud: bool | None = None,
+        **kwargs: Any
+    ) -> dict[str, Any]:
+        """Get available GPU types from RunPod.
+
+        Returns dict with available GPUs and their pricing.
+
+        Args:
+            regions: Filter by region (not supported by RunPod GPU types query)
+            gpu_count: Number of GPUs to request pricing for
+            gpu_type: Filter by GPU type name (partial match)
+            secure_cloud: If True, only show GPUs available in Secure Cloud
+            community_cloud: If True, only show GPUs available in Community Cloud
+        """
+        query = """
+        query GpuTypes {
+            gpuTypes {
+                id
+                displayName
+                memoryInGb
+                secureCloud
+                communityCloud
+                lowestPrice(input: {gpuCount: 1}) {
+                    minimumBidPrice
+                    uninterruptablePrice
+                }
+            }
+        }
+        """
+
+        data = await self._graphql_request(query)
+        gpu_types = data.get("gpuTypes", [])
+
+        # Filter by GPU type if specified
+        if gpu_type:
+            gpu_type_lower = gpu_type.lower()
+            gpu_types = [
+                g for g in gpu_types
+                if gpu_type_lower in g.get("displayName", "").lower()
+                or gpu_type_lower in g.get("id", "").lower()
+            ]
+
+        # Filter by cloud type
+        if secure_cloud is True:
+            gpu_types = [g for g in gpu_types if g.get("secureCloud")]
+        if community_cloud is True:
+            gpu_types = [g for g in gpu_types if g.get("communityCloud")]
+
+        # Transform to normalized format
+        gpus = []
+        for gpu in gpu_types:
+            lowest_price = gpu.get("lowestPrice", {})
+            price = lowest_price.get("uninterruptablePrice") or lowest_price.get("minimumBidPrice") or 0
+
+            gpus.append({
+                "gpuType": gpu.get("id"),
+                "gpuName": gpu.get("displayName"),
+                "gpuCount": gpu_count or 1,
+                "priceHr": price,
+                "cloudId": gpu.get("id"),
+                "socket": gpu.get("id"),
+                "memoryGb": gpu.get("memoryInGb"),
+                "secureCloud": gpu.get("secureCloud"),
+                "communityCloud": gpu.get("communityCloud"),
+                "provider": self.PROVIDER_NAME
+            })
+
+        return {
+            "data": gpus,
+            "total_count": len(gpus),
+            "provider": self.PROVIDER_NAME
+        }
+
+    async def create_pod(self, request: Any) -> PodResponse:
+        """Create a new RunPod pod.
+
+        Args:
+            request: CreatePodRequest with pod configuration
+
+        Returns:
+            PodResponse with created pod info
+        """
+        pod_config = request.pod if hasattr(request, 'pod') else request
+
+        mutation = """
+        mutation CreatePod($input: PodFindAndDeployOnDemandInput!) {
+            podFindAndDeployOnDemand(input: $input) {
+                id
+                name
+                desiredStatus
+                imageName
+                gpuCount
+                machineId
+                machine {
+                    gpuDisplayName
+                }
+                runtime {
+                    uptimeInSeconds
+                    ports {
+                        ip
+                        isIpPublic
+                        privatePort
+                        publicPort
+                        type
+                    }
+                }
+            }
+        }
+        """
+
+        variables = {
+            "input": {
+                "name": pod_config.name if hasattr(pod_config, 'name') else pod_config.get("name"),
+                "gpuTypeId": pod_config.gpuType if hasattr(pod_config, 'gpuType') else pod_config.get("gpuType"),
+                "gpuCount": pod_config.gpuCount if hasattr(pod_config, 'gpuCount') else pod_config.get("gpuCount", 1),
+                "volumeInGb": pod_config.diskSize if hasattr(pod_config, 'diskSize') else pod_config.get("diskSize", 20),
+                "containerDiskInGb": 20,
+                "dockerArgs": "",
+                "deployCost": pod_config.maxPrice if hasattr(pod_config, 'maxPrice') else pod_config.get("maxPrice"),
+                "startSsh": True,
+                "imageName": pod_config.image if hasattr(pod_config, 'image') else pod_config.get("image", "runpod/pytorch:2.1.0-py3.10-cuda11.8.0-devel-ubuntu22.04"),
+            }
+        }
+
+        # Remove None values
+        variables["input"] = {k: v for k, v in variables["input"].items() if v is not None}
+
+        data = await self._graphql_request(mutation, variables)
+        pod_data = data.get("podFindAndDeployOnDemand", {})
+
+        # Get SSH connection info
+        ssh_connection = await self._get_ssh_connection(pod_data.get("id"))
+
+        now = datetime.utcnow()
+        return PodResponse(
+            id=pod_data.get("id", ""),
+            userId="",
+            teamId=None,
+            name=pod_data.get("name", ""),
+            status=self._normalize_status(pod_data.get("desiredStatus", "PENDING")),
+            gpuName=pod_data.get("machine", {}).get("gpuDisplayName", ""),
+            gpuCount=pod_data.get("gpuCount", 1),
+            priceHr=0.0,  # Will be fetched separately
+            sshConnection=ssh_connection,
+            ip=None,
+            createdAt=now,
+            updatedAt=now
+        )
+
+    async def _get_ssh_connection(self, pod_id: str) -> str | None:
+        """Get SSH connection string for a pod."""
+        if not pod_id:
+            return None
+
+        query = """
+        query Pod($podId: String!) {
+            pod(input: {podId: $podId}) {
+                id
+                runtime {
+                    ports {
+                        ip
+                        isIpPublic
+                        privatePort
+                        publicPort
+                        type
+                    }
+                }
+            }
+        }
+        """
+
+        try:
+            data = await self._graphql_request(query, {"podId": pod_id})
+            pod = data.get("pod", {})
+            runtime = pod.get("runtime", {})
+            ports = runtime.get("ports", [])
+
+            for port in ports:
+                if port.get("privatePort") == 22 and port.get("isIpPublic"):
+                    ip = port.get("ip")
+                    public_port = port.get("publicPort")
+                    return f"ssh root@{ip} -p {public_port}"
+        except Exception:
+            pass
+
+        return None
+
+    async def get_pods(
+        self,
+        status: str | None = None,
+        limit: int = 100,
+        offset: int = 0
+    ) -> dict[str, Any]:
+        """Get list of all RunPod pods."""
+        query = """
+        query Pods {
+            myself {
+                pods {
+                    id
+                    name
+                    desiredStatus
+                    imageName
+                    gpuCount
+                    costPerHr
+                    machineId
+                    machine {
+                        gpuDisplayName
+                    }
+                    runtime {
+                        uptimeInSeconds
+                        ports {
+                            ip
+                            isIpPublic
+                            privatePort
+                            publicPort
+                            type
+                        }
+                    }
+                }
+            }
+        }
+        """
+
+        data = await self._graphql_request(query)
+        pods_raw = data.get("myself", {}).get("pods", [])
+
+        # Filter by status if specified
+        if status:
+            status_upper = status.upper()
+            pods_raw = [p for p in pods_raw if p.get("desiredStatus", "").upper() == status_upper]
+
+        # Apply pagination
+        pods_raw = pods_raw[offset:offset + limit]
+
+        # Transform to standardized format
+        pods = []
+        for pod in pods_raw:
+            ssh_connection = None
+            runtime = pod.get("runtime", {})
+            if runtime:
+                ports = runtime.get("ports", [])
+                for port in ports:
+                    if port.get("privatePort") == 22 and port.get("isIpPublic"):
+                        ip = port.get("ip")
+                        public_port = port.get("publicPort")
+                        ssh_connection = f"ssh root@{ip} -p {public_port}"
+                        break
+
+            pods.append({
+                "id": pod.get("id"),
+                "name": pod.get("name"),
+                "status": self._normalize_status(pod.get("desiredStatus", "PENDING")),
+                "gpuName": pod.get("machine", {}).get("gpuDisplayName", ""),
+                "gpuCount": pod.get("gpuCount", 1),
+                "priceHr": pod.get("costPerHr", 0),
+                "sshConnection": ssh_connection,
+                "ip": None,
+                "createdAt": datetime.utcnow().isoformat(),
+                "updatedAt": datetime.utcnow().isoformat(),
+                "provider": self.PROVIDER_NAME
+            })
+
+        return {
+            "data": pods,
+            "total_count": len(pods),
+            "offset": offset,
+            "limit": limit,
+            "provider": self.PROVIDER_NAME
+        }
+
+    async def get_pod(self, pod_id: str) -> PodResponse:
+        """Get details for a specific RunPod pod."""
+        query = """
+        query Pod($podId: String!) {
+            pod(input: {podId: $podId}) {
+                id
+                name
+                desiredStatus
+                imageName
+                gpuCount
+                costPerHr
+                machineId
+                machine {
+                    gpuDisplayName
+                }
+                runtime {
+                    uptimeInSeconds
+                    ports {
+                        ip
+                        isIpPublic
+                        privatePort
+                        publicPort
+                        type
+                    }
+                }
+            }
+        }
+        """
+
+        data = await self._graphql_request(query, {"podId": pod_id})
+        pod = data.get("pod")
+
+        if not pod:
+            raise HTTPException(status_code=404, detail=f"Pod {pod_id} not found")
+
+        # Extract SSH connection
+        ssh_connection = None
+        runtime = pod.get("runtime", {})
+        if runtime:
+            ports = runtime.get("ports", [])
+            for port in ports:
+                if port.get("privatePort") == 22 and port.get("isIpPublic"):
+                    ip = port.get("ip")
+                    public_port = port.get("publicPort")
+                    ssh_connection = f"ssh root@{ip} -p {public_port}"
+                    break
+
+        now = datetime.utcnow()
+        return PodResponse(
+            id=pod.get("id", ""),
+            userId="",
+            teamId=None,
+            name=pod.get("name", ""),
+            status=self._normalize_status(pod.get("desiredStatus", "PENDING")),
+            gpuName=pod.get("machine", {}).get("gpuDisplayName", ""),
+            gpuCount=pod.get("gpuCount", 1),
+            priceHr=pod.get("costPerHr", 0),
+            sshConnection=ssh_connection,
+            ip=None,
+            createdAt=now,
+            updatedAt=now
+        )
+
+    async def delete_pod(self, pod_id: str) -> dict[str, Any]:
+        """Delete/terminate a RunPod pod."""
+        mutation = """
+        mutation TerminatePod($podId: String!) {
+            podTerminate(input: {podId: $podId})
+        }
+        """
+
+        await self._graphql_request(mutation, {"podId": pod_id})
+        return {"success": True, "pod_id": pod_id, "provider": self.PROVIDER_NAME}
+
+    async def stop_pod(self, pod_id: str) -> dict[str, Any]:
+        """Stop a RunPod pod (without deleting)."""
+        mutation = """
+        mutation StopPod($podId: String!) {
+            podStop(input: {podId: $podId})
+        }
+        """
+
+        await self._graphql_request(mutation, {"podId": pod_id})
+        return {"success": True, "pod_id": pod_id, "action": "stopped"}
+
+    async def resume_pod(self, pod_id: str) -> dict[str, Any]:
+        """Resume a stopped RunPod pod."""
+        mutation = """
+        mutation ResumePod($podId: String!) {
+            podResume(input: {podId: $podId}) {
+                id
+                desiredStatus
+            }
+        }
+        """
+
+        data = await self._graphql_request(mutation, {"podId": pod_id})
+        return {
+            "success": True,
+            "pod_id": pod_id,
+            "status": data.get("podResume", {}).get("desiredStatus", "RUNNING")
+        }
+
+    def _normalize_status(self, runpod_status: str) -> str:
+        """Convert RunPod status to normalized status."""
+        status_map = {
+            "RUNNING": "ACTIVE",
+            "PENDING": "PENDING",
+            "EXITED": "TERMINATED",
+            "STOPPED": "STOPPED",
+            "STOPPING": "STOPPING",
+            "STARTING": "STARTING",
+            "TERMINATING": "TERMINATING",
+            "TERMINATED": "TERMINATED",
+            "ERROR": "ERROR"
+        }
+        return status_map.get(runpod_status.upper(), runpod_status)
+
+    def normalize_pod(self, pod_data: dict[str, Any]) -> NormalizedPod:
+        """Convert RunPod pod data to normalized format."""
+        ssh_connection = None
+        runtime = pod_data.get("runtime", {})
+        if runtime:
+            ports = runtime.get("ports", [])
+            for port in ports:
+                if port.get("privatePort") == 22 and port.get("isIpPublic"):
+                    ip = port.get("ip")
+                    public_port = port.get("publicPort")
+                    ssh_connection = f"ssh root@{ip} -p {public_port}"
+                    break
+
+        return NormalizedPod(
+            id=pod_data.get("id", ""),
+            name=pod_data.get("name", ""),
+            status=self._normalize_status(pod_data.get("desiredStatus", "PENDING")),
+            gpu_name=pod_data.get("machine", {}).get("gpuDisplayName", ""),
+            gpu_count=pod_data.get("gpuCount", 1),
+            price_hr=pod_data.get("costPerHr", 0.0),
+            ssh_connection=ssh_connection,
+            ip=None,
+            provider=self.PROVIDER_NAME,
+            created_at=datetime.utcnow().isoformat(),
+            updated_at=datetime.utcnow().isoformat()
+        )