improved startup first frame boot

src/app_controller.py

@@ -801,7 +801,7 @@ class AppController:
   Owns the application state and manages background services.
  """
 
- def __init__(self, log_to_stderr: bool = True):
+ def __init__(self, defer_warmup: bool = False, log_to_stderr: Optional[bool] = None):
   """
     [C: src/mcp_client.py:_DDGParser.__init__, src/mcp_client.py:_TextExtractor.__init__]
   """
@@ -823,10 +823,22 @@ class AppController:
 
   # --- Shared background pool + proactive warmup (startup_speedup_20260606) ---
   self._io_pool = make_io_pool()
+  # Warmup progress is a diagnostic; keep stderr quiet unless explicitly asked.
+  # Explicit log_to_stderr arg wins; otherwise default to the SLOP_WARMUP_DEBUG env flag.
+  if log_to_stderr is None:
+   log_to_stderr = bool(os.environ.get("SLOP_WARMUP_DEBUG"))
   self._warmup = WarmupManager(self._io_pool, log_to_stderr=log_to_stderr)
   # Hook warmup completion to stamp warmup_done_ts for startup_timeline().
   self._warmup.on_complete(self._on_warmup_complete_for_timeline)
-  self._warmup.submit(self._compute_warmup_list())
+  self._warmup_started: bool = False
+  self._defer_warmup: bool = defer_warmup
+  self._pending_fetch_provider: Optional[str] = None
+  # The desktop GUI defers warmup until the first frame is painted (see
+  # App._gui_func) so the ~2s of heavy SDK C-extension imports don't hold the
+  # GIL while the window and font atlas are being created — that contention is
+  # what made the window slow to appear. Headless/web/tests warm immediately.
+  if not defer_warmup:
+   self.start_warmup()
 
   # --- Internal State ---
   self._ai_status: str = "idle"
@@ -2132,6 +2144,13 @@ class AppController:
   """
     [C: src/gui_2.py:App.run]
   """
+  # In the desktop GUI, model listing imports the provider SDKs (the same
+  # ~2s C-extension load warmup pays for). Defer it until the first frame is
+  # painted so it doesn't contend for the GIL during window creation; the
+  # deferred fetch is fired from start_warmup().
+  if self._defer_warmup and not self._warmup_started:
+   self._pending_fetch_provider = provider
+   return
   self.ai_status = "fetching models..."
 
   def do_fetch() -> None:
@@ -2213,9 +2232,26 @@ class AppController:
   """
   return self._warmup.is_done()
 
+ def start_warmup(self) -> None:
+  """
+  Submit the heavy-module warmup jobs to the io_pool (idempotent).
+
+  Separated from __init__ so the desktop GUI can call it AFTER the first
+  frame is painted, keeping the ~2s of SDK C-extension imports off the GIL
+  while the window is being created. Safe to call multiple times.
+  """
+  if self._warmup_started:
+   return
+  self._warmup_started = True
+  self._warmup.submit(self._compute_warmup_list())
+  # Run any model fetch that was deferred while the window was being created.
+  if self._pending_fetch_provider is not None:
+   provider, self._pending_fetch_provider = self._pending_fetch_provider, None
+   self._fetch_models(provider)
+
  def wait_for_warmup(self, timeout: Optional[float] = None) -> bool:
   """
-  
+
    Block until warmup completes. Returns True on done, False on timeout.
    [SDM: src/app_controller.py:wait_for_warmup]
   """
@@ -2268,7 +2304,10 @@ class AppController:
 
  def _init_ai_and_hooks(self, app: Any = None) -> None:
   from src import api_hooks
-  ai_client.set_provider(self._current_provider, self._current_model)
+  # validate=False: skip the live model-list lookup (network/subprocess +
+  # provider-SDK import) on the main thread during startup. _fetch_models
+  # corrects the model against the live list after the first frame, off-thread.
+  ai_client.set_provider(self._current_provider, self._current_model, validate=False)
   if self._current_provider == "gemini_cli":
    if not ai_client._gemini_cli_adapter:
     ai_client._gemini_cli_adapter = ai_client.GeminiCliAdapter(binary_path=self.ui_gemini_cli_path)