feat(testing): stabilize simulation suite and fix gemini caching

tests/test_gui2_parity.py

@@ -22,53 +22,49 @@ def cleanup_callback_file():
     if TEST_CALLBACK_FILE.exists():
         TEST_CALLBACK_FILE.unlink()
 
-def test_gui2_set_value_hook_works(live_gui_2):
+def test_gui2_set_value_hook_works(live_gui):
     """
     Tests that the 'set_value' GUI hook is correctly implemented.
-    This requires a way to read the value back, which we don't have yet.
-    For now, this test just sends the command and assumes it works.
     """
     client = ApiHookClient()
+    assert client.wait_for_server(timeout=10)
     test_value = f"New value set by test: {uuid.uuid4()}"
     gui_data = {'action': 'set_value', 'item': 'ai_input', 'value': test_value}
     
     response = client.post_gui(gui_data)
     assert response == {'status': 'queued'}
     
-    # In a future test, we would add:
-    # time.sleep(0.2)
-    # current_value = client.get_value('ai_input') # This hook doesn't exist yet
-    # assert current_value == test_value
+    # Verify the value was actually set using the new get_value hook
+    time.sleep(0.5)
+    current_value = client.get_value('ai_input')
+    assert current_value == test_value
 
-def test_gui2_click_hook_works(live_gui_2):
+def test_gui2_click_hook_works(live_gui):
     """
     Tests that the 'click' GUI hook for the 'Reset' button is implemented.
-    This will be verified by checking for a side effect (e.g., session is reset,
-    which can be checked via another hook).
     """
     client = ApiHookClient()
+    assert client.wait_for_server(timeout=10)
     
     # First, set some state that 'Reset' would clear.
-    # We use the 'set_value' hook for this.
     test_value = "This text should be cleared by the reset button."
-    client.post_gui({'action': 'set_value', 'item': 'ai_input', 'value': test_value})
-    time.sleep(0.2)
+    client.set_value('ai_input', test_value)
+    time.sleep(0.5)
+    assert client.get_value('ai_input') == test_value
 
     # Now, trigger the click
-    gui_data = {'action': 'click', 'item': 'btn_reset'}
-    response = client.post_gui(gui_data)
-    assert response == {'status': 'queued'}
+    client.click('btn_reset')
+    time.sleep(0.5)
     
-    # We need a way to verify the state was reset.
-    # We can't read the ai_input value back yet.
-    # So this test remains conceptual for now, but demonstrates the intent.
+    # Verify it was reset
+    assert client.get_value('ai_input') == ""
 
-def test_gui2_custom_callback_hook_works(live_gui_2):
+def test_gui2_custom_callback_hook_works(live_gui):
     """
     Tests that the 'custom_callback' GUI hook is correctly implemented.
-    This test will PASS if the hook is correctly processed by gui_2.py.
     """
     client = ApiHookClient()
+    assert client.wait_for_server(timeout=10)
     test_data = f"Callback executed: {uuid.uuid4()}"
 
     gui_data = {