Include creation parameters inline in launcher instance state replies

VllmInstance.get_status() now spreads VllmConfig fields (options,
gpu_uuids, env_vars) inline into the returned dict via model_dump(),
so callers can inspect the configuration that was used to start each
instance without a separate side-channel.

InstanceStatus in the Go client is updated to match the flat JSON
schema with Options, GpuUUIDs, and EnvVars fields.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Mike Spreitzer <mspreitz@us.ibm.com>

Author: MikeSpreitzer

inference_server/launcher/launcher.py

@@ -177,6 +177,7 @@ def get_status(self) -> dict:
         return {
             "status": "running" if self.process.is_alive() else "stopped",
             "instance_id": self.instance_id,
+            **self.config.model_dump(),
         }
 
     def get_log_bytes(

inference_server/launcher/tests/test_launcher.py

@@ -258,11 +258,14 @@ def test_instance_get_status(
         instance.start()
         status = instance.get_status()
         assert status["status"] == "running"
+        assert status["options"] == vllm_config.options
+        assert status["env_vars"] == vllm_config.env_vars
 
         # Stopped
         mock_process._is_alive = False
         status = instance.get_status()
         assert status["status"] == "stopped"
+        assert status["options"] == vllm_config.options
 
     @patch("launcher.multiprocessing.Process")
     def test_instance_uuid_to_index_translation(
@@ -473,6 +476,8 @@ def test_get_instance_status(self, mock_process_class, manager, vllm_config):
 
         assert status["status"] == "running"
         assert status["instance_id"] == "test-id"
+        assert status["options"] == vllm_config.options
+        assert status["env_vars"] == vllm_config.env_vars
 
     @patch("launcher.multiprocessing.Process")
     def test_get_instance_status_nonexistent(self, mock_process_class, manager):
@@ -494,6 +499,9 @@ def test_get_all_instances_status(self, mock_process_class, manager, vllm_config
         assert status["total_instances"] == 2
         assert status["running_instances"] == 2
         assert len(status["instances"]) == 2
+        for inst in status["instances"]:
+            assert inst["options"] == vllm_config.options
+            assert inst["env_vars"] == vllm_config.env_vars
 
     @patch("launcher.multiprocessing.Process")
     def test_list_instances(self, mock_process_class, manager, vllm_config):
@@ -693,13 +701,18 @@ def test_get_instance_status(self, mock_manager, client):
         mock_manager.get_instance_status.return_value = {
             "status": "running",
             "instance_id": "test-id",
+            "options": "--model test-model",
+            "gpu_uuids": None,
+            "env_vars": {"KEY": "val"},
         }
 
         response = client.get("/v2/vllm/instances/test-id")
 
         assert response.status_code == 200
         data = response.json()
         assert data["status"] == "running"
+        assert data["options"] == "--model test-model"
+        assert data["env_vars"] == {"KEY": "val"}
 
     @patch("launcher.vllm_manager")
     def test_get_nonexistent_instance_status(self, mock_manager, client):

pkg/controller/dual-pods/launcherclient.go

@@ -57,8 +57,11 @@ type VllmConfig struct {
 
 // InstanceStatus returned by status APIs.
 type InstanceStatus struct {
-	InstanceID string `json:"instance_id"`
-	Status     string `json:"status"`
+	InstanceID string            `json:"instance_id"`
+	Status     string            `json:"status"`
+	Options    string            `json:"options"`
+	GpuUUIDs   []string          `json:"gpu_uuids,omitempty"`
+	EnvVars    map[string]string `json:"env_vars,omitempty"`
 }
 
 // AllInstancesStatus response.