Update api/ltx_server_refactored_complete.py

api/ltx_server_refactored_complete.py

@@ -1,7 +1,6 @@
 # FILE: api/ltx_server_refactored_complete.py
 # DESCRIPTION: Final high-level orchestrator for LTX-Video generation.
-# This version delegates all low-level tasks to dedicated utility modules and managers,
-# focusing solely on the business logic of video generation workflows.
+# Incorporates a debug logging decorator for deep inspection of function I/O.
 
 import gc
 import json
@@ -34,6 +33,7 @@ def add_deps_to_path():
     repo_path = str(LTX_VIDEO_REPO_DIR.resolve())
     if repo_path not in sys.path:
         sys.path.insert(0, repo_path)
+        # Usamos logging.info aqui, pois é uma informação importante de inicialização
         logging.info(f"[ltx_server] LTX-Video repository added to sys.path: {repo_path}")
 
 add_deps_to_path()
@@ -52,6 +52,9 @@ try:
         ConditioningItem,
     )
     
+    # Nosso novo decorador de logging para depuração
+    from api.utils.debug_utils import log_function_io
+
 except ImportError as e:
     logging.critical(f"A crucial import from the local API/architecture failed. Error: {e}", exc_info=True)
     sys.exit(1)
@@ -60,9 +63,10 @@ except ImportError as e:
 # --- FUNÇÕES AUXILIARES DO ORQUESTRADOR ---
 # ==============================================================================
 
+@log_function_io
 def calculate_padding(orig_h: int, orig_w: int, target_h: int, target_w: int) -> Tuple[int, int, int, int]:
     """Calculates symmetric padding required to meet target dimensions."""
-    pad_h = target_h - orig_h
+    pad_h = target_h - orig_w
     pad_w = target_w - orig_w
     pad_top = pad_h // 2
     pad_bottom = pad_h - pad_top
@@ -80,8 +84,10 @@ class VideoService:
     tasks to specialized managers and utility modules.
     """
 
+    @log_function_io
     def __init__(self):
         t0 = time.perf_counter()
+        # Logging de alto nível para o usuário
         logging.info("Initializing VideoService Orchestrator...")
         RESULTS_DIR.mkdir(parents=True, exist_ok=True)
 
@@ -100,6 +106,7 @@ class VideoService:
         vae_manager_singleton.attach_pipeline(self.pipeline, device=self.vae_device, autocast_dtype=self.runtime_autocast_dtype)
         logging.info(f"VideoService ready. Startup time: {time.perf_counter()-t0:.2f}s")
 
+    @log_function_io
     def _load_config(self) -> Dict:
         """Loads the YAML configuration file."""
         config_path = LTX_VIDEO_REPO_DIR / "configs" / "ltxv-13b-0.9.8-distilled-fp8.yaml"
@@ -107,6 +114,7 @@ class VideoService:
         with open(config_path, "r") as file:
             return yaml.safe_load(file)
 
+    @log_function_io
     def move_to_device(self, main_device_str: str, vae_device_str: str):
         """Moves pipeline components to their designated target devices."""
         target_main_device = torch.device(main_device_str)
@@ -137,6 +145,7 @@ class VideoService:
     # --- LÓGICA DE NEGÓCIO: ORQUESTRADORES PÚBLICOS ---
     # ==========================================================================
 
+    @log_function_io
     def generate_narrative_low(self, prompt: str, **kwargs) -> Tuple[Optional[str], Optional[str], Optional[int]]:
         """Orchestrates the generation of a video from a multi-line prompt (sequence of scenes)."""
         logging.info("Starting narrative low-res generation...")
@@ -186,6 +195,7 @@ class VideoService:
                 if path.exists(): path.unlink()
             self.finalize()
 
+    @log_function_io
     def generate_single_low(self, **kwargs) -> Tuple[Optional[str], Optional[str], Optional[int]]:
         """Orchestrates the generation of a video from a single prompt in one go."""
         logging.info("Starting single-prompt low-res generation...")
@@ -213,6 +223,7 @@ class VideoService:
     # --- UNIDADES DE TRABALHO E HELPERS INTERNOS ---
     # ==========================================================================
 
+    @log_function_io
     def _generate_single_chunk_low(self, **kwargs) -> Optional[torch.Tensor]:
         """Calls the pipeline to generate a single chunk of latents."""
         height_padded, width_padded = (self._align(d) for d in (kwargs['height'], kwargs['width']))
@@ -237,6 +248,7 @@ class VideoService:
         
         return latents_raw.to(self.main_device)
 
+    @log_function_io
     def _finalize_generation(self, temp_latent_paths: List[Path], base_filename: str, seed: int) -> Tuple[str, str, int]:
         """Consolidates latents, decodes them to video, and saves final artifacts."""
         logging.info("Finalizing generation: decoding latents to video.")
@@ -253,6 +265,7 @@ class VideoService:
         video_path = self._save_and_log_video(pixel_tensor, f"{base_filename}_{seed}")
         return str(video_path), str(final_latents_path), seed
 
+    @log_function_io
     def prepare_condition_items(self, items_list: List, height: int, width: int, num_frames: int) -> List[ConditioningItem]:
         if not items_list: return []
         height_padded, width_padded = self._align(height), self._align(width)
@@ -265,6 +278,7 @@ class VideoService:
             conditioning_items.append(ConditioningItem(tensor, safe_frame, float(weight)))
         return conditioning_items
 
+    @log_function_io
     def _prepare_conditioning_tensor(self, media_path: str, height: int, width: int, padding: Tuple) -> torch.Tensor:
         tensor = load_image_to_tensor_with_resize_and_crop(media_path, height, width)
         tensor = torch.nn.functional.pad(tensor, padding)
@@ -273,7 +287,19 @@ class VideoService:
     def _prepare_guidance_overrides(self, ltx_configs: Dict) -> Dict:
         overrides = {}
         preset = ltx_configs.get("guidance_preset", "Padrão (Recomendado)")
-        # ... (logic for presets remains the same)
+        if preset == "Agressivo":
+            overrides["guidance_scale"] = [1, 2, 8, 12, 8, 2, 1]
+            overrides["stg_scale"] = [0, 0, 5, 6, 5, 3, 2]
+        elif preset == "Suave":
+            overrides["guidance_scale"] = [1, 1, 4, 5, 4, 1, 1]
+            overrides["stg_scale"] = [0, 0, 2, 2, 2, 1, 0]
+        elif preset == "Customizado":
+            try:
+                overrides["guidance_scale"] = json.loads(ltx_configs["guidance_scale_list"])
+                overrides["stg_scale"] = json.loads(ltx_configs["stg_scale_list"])
+            except (json.JSONDecodeError, KeyError) as e:
+                logging.warning(f"Failed to parse custom guidance values: {e}. Falling back to defaults.")
+        if overrides: logging.info(f"Applying '{preset}' guidance preset overrides.")
         return overrides
 
     def _save_and_log_video(self, pixel_tensor: torch.Tensor, base_filename: str) -> Path: