feat: surface /api/status memory+disk pressure on connect/status screens (#903) (#908)

Co-authored-by: M57 <hy4ri@users.noreply.github.com>

app/src/main/java/com/m57/hermescontrol/ui/connect/ConnectScreen.kt

@@ -58,6 +58,7 @@ import androidx.compose.ui.unit.dp
 import androidx.lifecycle.compose.collectAsStateWithLifecycle
 import androidx.lifecycle.viewmodel.compose.viewModel
 import com.m57.hermescontrol.R
+import com.m57.hermescontrol.ui.common.PressureBanner
 
 @Composable
 fun ConnectScreen(
@@ -354,6 +355,12 @@ fun ConnectScreen(
                         }
                     }
                 }
+                // Host-health advisory (issue #903): memory/disk pressure from
+                // the /api/status probe — explains a struggling host before or
+                // after a failed connect attempt.
+                state.status?.let { status ->
+                    PressureBanner(memory = status.memory, disk = status.disk)
+                }
             }
         }
     }

app/src/main/java/com/m57/hermescontrol/ui/connect/ConnectViewModel.kt

@@ -9,6 +9,7 @@ import com.m57.hermescontrol.data.config.ConnectionProfile
 import com.m57.hermescontrol.data.config.resolveBaseUrl
 import com.m57.hermescontrol.data.local.AuthManager
 import com.m57.hermescontrol.data.model.HealthStatus
+import com.m57.hermescontrol.data.model.StatusResponse
 import com.m57.hermescontrol.data.remote.ApiClient
 import com.m57.hermescontrol.data.remote.CleartextPolicy
 import com.m57.hermescontrol.data.remote.NetworkError
@@ -33,6 +34,7 @@ data class ConnectUiState(
     val profiles: List<ConnectionProfile> = emptyList(),
     val selectedProfile: ConnectionProfile? = null,
     val health: HealthStatus? = null,
+    val status: StatusResponse? = null,
 )
 
 class ConnectViewModel(
@@ -61,6 +63,7 @@ class ConnectViewModel(
             )
         }
         loadHealth()
+        loadStatus()
     }
 
     /**
@@ -84,6 +87,28 @@ class ConnectViewModel(
         }
     }
 
+    /**
+     * Best-effort `/api/status` probe (issue #903): surfaces the host
+     * memory/disk pressure banner on the connection screen. Fired on screen
+     * load (saved credentials) and after a failed connect, so an OOM-thrashing
+     * or disk-full host explains itself instead of looking like a dead agent.
+     * Failures are silent (status stays null and the screen shows no banner).
+     */
+    fun loadStatus() {
+        val state = _uiState.value
+        val endpoint = runCatching { ServerEndpoint.parseForBuild(state.baseUrl) }.getOrNull()
+        if (endpoint == null || state.token.isBlank()) return
+        viewModelScope.launch {
+            val result =
+                safeApiCall {
+                    ApiClient.createTempService(endpoint.baseUrl.toString(), state.token).getStatus()
+                }
+            if (result is NetworkResult.Success) {
+                _uiState.update { it.copy(status = result.data) }
+            }
+        }
+    }
+
     fun onProfileNameChange(value: String) {
         _uiState.update { it.copy(profileName = value, errorMessage = null) }
     }
@@ -180,7 +205,12 @@ class ConnectViewModel(
                     }
                     ApiClient.rebuild()
                     _uiState.update {
-                        it.copy(isConnecting = false, connectionSuccess = true, errorMessage = null)
+                        it.copy(
+                            isConnecting = false,
+                            connectionSuccess = true,
+                            errorMessage = null,
+                            status = result.data,
+                        )
                     }
                 }
 
@@ -282,6 +312,8 @@ class ConnectViewModel(
                             }
                         }
                     _uiState.update { it.copy(isConnecting = false, errorMessage = msg) }
+                    // Best-effort: explain a failing host via its pressure rollup.
+                    loadStatus()
                 }
             }
         }