Move platform/ and k8s/ to providers/ directory structure

Reorganize iris cluster infrastructure code into a providers/ directory:
- platform/base.py -> providers/base.py
- platform/gcp*.py, bootstrap, debug, remote_exec, _worker_base -> providers/gcp/
- platform/coreweave.py -> providers/k8s/platform.py
- k8s/* -> providers/k8s/
- platform/manual.py -> providers/manual/platform.py
- platform/factory.py -> providers/factory.py
- controller/vm_lifecycle.py -> providers/gcp/vm_lifecycle.py

All imports updated across src/ and tests/.

Author: rjpower

lib/iris/src/iris/cli/main.py

@@ -72,7 +72,7 @@ def _configure_client_s3(config) -> None:
     R2_ACCESS_KEY_ID/R2_SECRET_ACCESS_KEY), AWS_ENDPOINT_URL, and FSSPEC_S3
     with the correct endpoint.
     """
-    from iris.cluster.platform.coreweave import _needs_virtual_host_addressing
+    from iris.cluster.providers.k8s.platform import _needs_virtual_host_addressing
 
     endpoint = config.platform.coreweave.object_storage_endpoint
     if not endpoint:

lib/iris/src/iris/cluster/config.py

@@ -25,7 +25,7 @@
 from google.protobuf.json_format import MessageToDict, ParseDict
 
 from iris.cluster.constraints import WellKnownAttribute
-from iris.cluster.k8s.provider import KubernetesProvider
+from iris.cluster.providers.k8s.provider import KubernetesProvider
 from iris.cluster.controller.worker_provider import WorkerProvider
 from iris.cluster.types import parse_memory_string
 from iris.managed_thread import ThreadContainer, get_thread_container
@@ -986,7 +986,7 @@ def platform_bundle(self):
             PlatformBundle with both platform and lifecycle
         """
         # Local import: platform.factory imports config.py, creating a circular dependency.
-        from iris.cluster.platform.factory import create_platform_bundle
+        from iris.cluster.providers.factory import create_platform_bundle
 
         return create_platform_bundle(
             platform_config=self._proto.platform,
@@ -1136,7 +1136,7 @@ def make_provider(cluster_config: config_pb2.IrisClusterConfig) -> WorkerProvide
     """
     which = cluster_config.WhichOneof("provider")
     if which == "kubernetes_provider":
-        from iris.cluster.k8s.k8s_service import CloudK8sService
+        from iris.cluster.providers.k8s.service import CloudK8sService
 
         kp = cluster_config.kubernetes_provider
         namespace = kp.namespace or "iris"

lib/iris/src/iris/cluster/controller/autoscaler.py

@@ -28,7 +28,7 @@
 from dataclasses import dataclass
 from enum import Enum
 
-from iris.cluster.platform.base import (
+from iris.cluster.providers.base import (
     CloudSliceState,
     CloudWorkerState,
     CommandResult,

lib/iris/src/iris/cluster/controller/controller.py

@@ -48,7 +48,7 @@
     running_tasks_by_worker,
 )
 from iris.cluster.controller.dashboard import ControllerDashboard
-from iris.cluster.k8s.provider import KubernetesProvider
+from iris.cluster.providers.k8s.provider import KubernetesProvider
 from iris.cluster.controller.provider import TaskProvider
 from iris.cluster.controller.scheduler import (
     JobRequirements,

lib/iris/src/iris/cluster/controller/main.py

@@ -65,9 +65,9 @@ def serve(
     from iris.cluster.controller.autoscaler import Autoscaler
     from iris.cluster.controller.checkpoint import download_checkpoint_to_local
     from iris.cluster.controller.db import ControllerDB
-    from iris.cluster.k8s.provider import KubernetesProvider
+    from iris.cluster.providers.k8s.provider import KubernetesProvider
     from iris.cluster.config import load_config, create_autoscaler, make_provider
-    from iris.cluster.platform.factory import create_platform_bundle
+    from iris.cluster.providers.factory import create_platform_bundle
 
     configure_logging(level=getattr(logging, log_level))
 

lib/iris/src/iris/cluster/controller/scaling_group.py

@@ -18,7 +18,7 @@
 
 from collections.abc import Sequence
 
-from iris.cluster.platform.base import Labels, Platform, SliceHandle
+from iris.cluster.providers.base import Labels, Platform, SliceHandle
 from iris.cluster.constraints import (
     AttributeValue,
     CONSTRAINT_REGISTRY,

lib/iris/src/iris/cluster/k8s/__init__.py

@@ -32,15 +32,15 @@
 )
 from iris.cluster.controller.worker_provider import RpcWorkerStubFactory, WorkerProvider
 from iris.cluster.controller.db import ControllerDB
-from iris.cluster.controller.vm_lifecycle import ControllerStatus
+from iris.cluster.providers.gcp.vm_lifecycle import ControllerStatus
 from iris.cluster.controller.scaling_group import (
     DEFAULT_SCALE_DOWN_RATE_LIMIT,
     DEFAULT_SCALE_UP_RATE_LIMIT,
     ScalingGroup,
 )
-from iris.cluster.platform.base import find_free_port
-from iris.cluster.platform.gcp import GcpPlatform
-from iris.cluster.platform.gcp_fake import InMemoryGcpService
+from iris.cluster.providers.base import find_free_port
+from iris.cluster.providers.gcp.platform import GcpPlatform
+from iris.cluster.providers.gcp.fake_service import InMemoryGcpService
 from iris.cluster.service_mode import ServiceMode
 from iris.cluster.worker.port_allocator import PortAllocator
 from iris.managed_thread import ThreadContainer

lib/iris/src/iris/cluster/local_cluster.py

@@ -0,0 +1,49 @@
+# Copyright The Marin Authors
+# SPDX-License-Identifier: Apache-2.0
+
+"""Infrastructure provider abstraction layer.
+
+Re-exports the core types and protocols. Concrete implementations live in
+subpackages (gcp/, k8s/, local/, manual/).
+"""
+
+from iris.cluster.providers.protocols import ControllerProvider, WorkerInfraProvider
+from iris.cluster.providers.types import (
+    CloudSliceState,
+    CloudWorkerState,
+    CommandResult,
+    Labels,
+    PlatformError,
+    PlatformUnavailableError,
+    QuotaExhaustedError,
+    RemoteWorkerHandle,
+    ResourceNotFoundError,
+    SliceHandle,
+    SliceStatus,
+    StandaloneWorkerHandle,
+    WorkerStatus,
+    default_stop_all,
+    find_free_port,
+    generate_slice_suffix,
+)
+
+__all__ = [
+    "CloudSliceState",
+    "CloudWorkerState",
+    "CommandResult",
+    "ControllerProvider",
+    "Labels",
+    "PlatformError",
+    "PlatformUnavailableError",
+    "QuotaExhaustedError",
+    "RemoteWorkerHandle",
+    "ResourceNotFoundError",
+    "SliceHandle",
+    "SliceStatus",
+    "StandaloneWorkerHandle",
+    "WorkerInfraProvider",
+    "WorkerStatus",
+    "default_stop_all",
+    "find_free_port",
+    "generate_slice_suffix",
+]