[SDK] Upgrade platform API client to 4.31.0, drop hardware cost

platform 4.31.0 removed cost_per_hr from HardwareInstanceResponse
(#4499), so the pinned 4.28.0 client failed every get_hardware_instances
call with a pydantic ValidationError. Move to the published 4.31.0
client and stop rendering the Cost row, which no longer has a source.

Build the hardware fixture from the real HardwareInstanceResponse
instead of SimpleNamespace; the fake declared cost_per_hr itself, so the
suite stayed green against a client that had dropped it.

Signed-off-by: Honglin Cao <hocao@nvidia.com>
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: V2arK

centml/cli/cluster.py

@@ -237,7 +237,6 @@ def get(type, id):
             ("Status", ready_status),
             ("Created at", deployment.created_at.strftime("%Y-%m-%d %H:%M:%S")),
             ("Hardware", f"{hw.name} ({hw.num_gpu}x {hw.gpu_type})"),
-            ("Cost", f"{hw.cost_per_hr / 100} credits/hr"),
         ]
         if depl_type != DeploymentType.JOB:
             detail_rows.insert(2, ("Endpoint", deployment.endpoint_url))

requirements.txt

@@ -5,6 +5,6 @@ pyjwt>=2.8.0
 cryptography==50.0.0
 websockets>=16.0
 pyte>=0.8.0
-platform-api-python-client==4.28.0
+platform-api-python-client==4.31.0
 click>=8.4.1
 typing-extensions>=4.5

setup.py

@@ -11,7 +11,7 @@
 
 setup(
     name='centml',
-    version='0.7.0',
+    version='0.7.1',
     packages=find_packages(exclude=["tests", "tests.*"]),
     python_requires=">=3.10",
     long_description=open('README.md').read(),

tests/test_cli_cluster.py

@@ -6,7 +6,7 @@
 from click.testing import CliRunner
 
 from centml.cli.cluster import _get_ready_status, _get_service_status, _get_status_error_messages
-from centml.sdk import DeploymentStatus, DeploymentType, RolloutStatus, ServiceStatus
+from centml.sdk import DeploymentStatus, DeploymentType, HardwareInstanceResponse, RolloutStatus, ServiceStatus
 
 
 def test_service_status_uses_legacy_service_status_when_present():
@@ -157,7 +157,9 @@ def test_get_job_routes_to_job_api_and_displays_job_config():
     from centml.cli.cluster import get
 
     deployment = _deployment(status=DeploymentStatus.ACTIVE)
-    hardware = SimpleNamespace(id=2, name="h100", num_gpu=8, gpu_type="H100", cost_per_hr=1200)
+    hardware = HardwareInstanceResponse(
+        id=2, name="h100", gpu_type="H100", num_gpu=8, cpu=96, memory=1024, cluster_id=1
+    )
     runner = CliRunner()
 
     with _patch_cluster_client() as client:
@@ -177,3 +179,5 @@ def test_get_job_routes_to_job_api_and_displays_job_config():
     assert "registry.example.com/job:latest" in result.output
     assert "Completions" in result.output
     assert "Parallelism" in result.output
+    assert "h100 (8x H100)" in result.output
+    assert "credits/hr" not in result.output