fix(llm-tui): fix card layout overlap, handle HTTP 503 loading state, and auto-switch to status tab

bin/llm-tui

@@ -220,6 +220,15 @@ def check_port_status(port: int) -> tuple[bool, str]:
             data = json.loads(response.read().decode())
             status = data.get("status", "unknown")
             return True, f"Online ({status})"
+    except urllib.error.HTTPError as e:
+        if e.code == 503:
+            try:
+                data = json.loads(e.read().decode())
+                msg = data.get("error", {}).get("message", "Loading model")
+                return True, f"Loading ({msg})"
+            except Exception:
+                return True, "Loading model..."
+        return False, f"HTTP Error {e.code}"
     except Exception:
         return False, "Offline"
 
@@ -246,6 +255,7 @@ class LLMConfigApp(App):
         border: solid #3b4252;
         padding: 1 2;
         margin-bottom: 1;
+        height: auto;
     }
     .title {
         color: #88c0d0;
@@ -302,18 +312,21 @@ class LLMConfigApp(App):
         border-left: thick #88c0d0;
         padding: 1 2;
         margin-bottom: 1;
+        height: auto;
     }
     #vram-meter {
         background: #232834;
         border: solid #4c566a;
         padding: 1;
         margin-top: 1;
+        height: auto;
     }
     #status-card {
         background: #1e222b;
         border: solid #4c566a;
         padding: 1;
         margin-bottom: 1;
+        height: auto;
     }
     Log {
         background: #0f1117;
@@ -339,6 +352,8 @@ class LLMConfigApp(App):
     _sync_lock: bool = False
     _last_selected_model: str = ""
     _last_selected_mmproj: str = ""
+    _last_log_size: int = -1
+    _last_log_name: str = ""
 
     def compose(self) -> ComposeResult:
         yield Header(show_clock=True)
@@ -672,7 +687,12 @@ class LLMConfigApp(App):
         lines = []
         for port, label in [(11434, "Planner / Solo"), (11435, "Worker 1"), (11436, "Worker 2")]:
             online, msg = check_port_status(port)
-            badge = "[green]● ONLINE[/]" if online else "[yellow]○ OFFLINE[/]"
+            if not online:
+                badge = "[yellow]○ OFFLINE[/]"
+            elif "loading" in msg.lower():
+                badge = "[cyan]⏳ LOADING[/]"
+            else:
+                badge = "[green]● ONLINE[/]"
             lines.append(f"Port {port} ({label}): {badge} - {msg}")
 
         card = self.query_one("#status-card", Static)
@@ -684,8 +704,13 @@ class LLMConfigApp(App):
         if logs:
             latest = logs[0]
             try:
+                cur_size = latest.stat().st_size
+                if cur_size == self._last_log_size and latest.name == self._last_log_name:
+                    return
+                self._last_log_size = cur_size
+                self._last_log_name = latest.name
                 with open(latest, "r", errors="ignore") as f:
-                    f.seek(max(0, latest.stat().st_size - 8192))
+                    f.seek(max(0, cur_size - 16384))
                     tail = f.read()
                     log_widget.clear()
                     log_widget.write(f"--- Streaming {latest.name} ---\n" + tail)
@@ -732,7 +757,8 @@ class LLMConfigApp(App):
         )
 
         subprocess.Popen(cmd, env=env, cwd=str(SCRIPT_DIR), start_new_session=True)
-        self.notify(f"Launching {preset} server. Check Status & Live Logs tab.", severity="information")
+        self.notify(f"Launching {preset} server. Switching to Status & Logs...", severity="information")
+        self.query_one(TabbedContent).active = "tab-status"
         self.action_refresh_status()
 
     def action_stop(self) -> None: