feat(lea): substitute save menu gesture

agent_chat/gesture_catalog.py

@@ -16,6 +16,7 @@ Auteur: Dom — Mars 2026
 
 import logging
 import re
+import unicodedata
 import uuid
 from dataclasses import dataclass, field
 from difflib import SequenceMatcher
@@ -24,6 +25,11 @@ from typing import Dict, List, Optional, Tuple
 logger = logging.getLogger(__name__)
 
 
+SAVE_COMMAND_LABELS = {"enregistrer", "save", "sauvegarder"}
+SAVE_AS_LABELS = {"enregistrer sous", "save as", "sauvegarder sous"}
+FILE_MENU_LABELS = {"fichier", "file", "menu fichier", "file menu"}
+
+
 @dataclass
 class Gesture:
     """Un geste primitif universel."""
@@ -564,6 +570,7 @@ class GestureCatalog:
         Patterns :
         - Clic en haut à droite de la fenêtre (x > 95%, y < 5%) → fermer
         - target_text contenant ✕, ×, X, □, ─, etc.
+        - Commande applicative "Enregistrer" sûre → Ctrl+S
         """
         # Vérifier le target_text
         target_text = (
@@ -583,6 +590,9 @@ class GestureCatalog:
             if target_lower in ("─", "—", "_", "minimize", "réduire"):
                 return self._by_id.get("win_minimize")
 
+        if self._is_save_command_action(action):
+            return self._by_id.get("edit_save")
+
         # Vérifier la position relative (coin haut-droite = fermer)
         x_pct = action.get("x_pct", 0)
         y_pct = action.get("y_pct", 0)
@@ -596,6 +606,128 @@ class GestureCatalog:
 
         return None
 
+    def _normalize_ui_text(self, value: str) -> str:
+        """Normaliser un libellé UI pour comparer accents, casse et raccourcis."""
+        text = str(value or "").strip().lower()
+        text = unicodedata.normalize("NFKD", text)
+        text = "".join(ch for ch in text if not unicodedata.combining(ch))
+        text = text.replace("’", "'")
+        text = re.sub(r"\s+", " ", text)
+        text = re.sub(r"\s*\([^)]*ctrl\s*\+?\s*s[^)]*\)\s*$", "", text)
+        text = re.sub(r"\s+ctrl\s*\+?\s*s\s*$", "", text)
+        return text.strip()
+
+    def _action_text_candidates(self, action: Dict) -> List[str]:
+        """Retourner les libellés utiles d'une action et de son target_spec."""
+        target_spec = action.get("target_spec") or {}
+        candidates = [
+            action.get("target_text", ""),
+            action.get("target_description", ""),
+            action.get("description", ""),
+            target_spec.get("by_text", ""),
+            target_spec.get("target_text", ""),
+            target_spec.get("vlm_description", ""),
+        ]
+        return [str(c) for c in candidates if c]
+
+    def _action_role_text(self, action: Dict) -> str:
+        target_spec = action.get("target_spec") or {}
+        uia = action.get("uia_snapshot") or {}
+        role_parts = [
+            action.get("role", ""),
+            action.get("control_type", ""),
+            target_spec.get("by_role", ""),
+            target_spec.get("role", ""),
+            target_spec.get("control_type", ""),
+            uia.get("control_type", ""),
+            uia.get("class_name", ""),
+        ]
+        return " ".join(self._normalize_ui_text(part) for part in role_parts if part)
+
+    def _action_context_text(self, action: Dict) -> str:
+        target_spec = action.get("target_spec") or {}
+        hints = target_spec.get("context_hints") or {}
+        context_parts = [
+            action.get("window_title", ""),
+            target_spec.get("window_title", ""),
+            target_spec.get("vlm_description", ""),
+            hints.get("window_title", ""),
+            hints.get("interaction", ""),
+            hints.get("source", ""),
+            hints.get("menu_path", ""),
+        ]
+        return " ".join(self._normalize_ui_text(part) for part in context_parts if part)
+
+    def _is_file_menu_action(self, action: Dict) -> bool:
+        labels = {self._normalize_ui_text(text) for text in self._action_text_candidates(action)}
+        return bool(labels & FILE_MENU_LABELS)
+
+    def _is_save_command_label(self, action: Dict) -> bool:
+        for text in self._action_text_candidates(action):
+            label = self._normalize_ui_text(text)
+            if not label:
+                continue
+            if any(save_as in label for save_as in SAVE_AS_LABELS):
+                return False
+            if label in SAVE_COMMAND_LABELS:
+                return True
+        return False
+
+    def _is_save_dialog_action(self, action: Dict) -> bool:
+        context = self._action_context_text(action)
+        if any(save_as in context for save_as in SAVE_AS_LABELS):
+            return True
+        dialog_markers = (
+            "save dialog",
+            "save_dialog",
+            "dialog",
+            "boite de dialogue",
+            "fenetre enregistrer sous",
+            "confirmer l'enregistrement",
+            "save changes",
+        )
+        return any(marker in context for marker in dialog_markers)
+
+    def _is_save_command_action(self, action: Dict) -> bool:
+        if not self._is_save_command_label(action):
+            return False
+        if self._is_save_dialog_action(action):
+            return False
+
+        role = self._action_role_text(action)
+        context = self._action_context_text(action)
+        command_markers = (
+            "menu",
+            "menuitem",
+            "item de menu",
+            "toolbar",
+            "barre d'outils",
+            "tool bar",
+            "ruban",
+            "ribbon",
+            "commande",
+            "command",
+        )
+        return any(marker in role or marker in context for marker in command_markers)
+
+    def _substitute_action(
+        self,
+        action: Dict,
+        gesture: Gesture,
+        *,
+        original_type: str,
+        source_action_ids: Optional[List[str]] = None,
+        reason: str = "",
+    ) -> Dict:
+        new_action = gesture.to_replay_action()
+        new_action["action_id"] = action.get("action_id", new_action["action_id"])
+        new_action["original_type"] = original_type
+        if source_action_ids:
+            new_action["substitution_source_action_ids"] = source_action_ids
+        if reason:
+            new_action["substitution_reason"] = reason
+        return new_action
+
     def optimize_replay_actions(self, actions: List[Dict]) -> List[Dict]:
         """
         Optimiser une liste d'actions de replay en substituant les gestes connus.
@@ -610,13 +742,45 @@ class GestureCatalog:
         substitutions = 0
 
         for action in actions:
+            if (
+                action.get("type") == "click"
+                and optimized
+                and optimized[-1].get("type") == "click"
+                and self._is_file_menu_action(optimized[-1])
+                and self._is_save_command_label(action)
+                and not self._is_save_dialog_action(action)
+            ):
+                gesture = self._by_id.get("edit_save")
+                previous = optimized.pop()
+                source_ids = [
+                    source_id for source_id in (
+                        previous.get("action_id"),
+                        action.get("action_id"),
+                    )
+                    if source_id
+                ]
+                optimized.append(
+                    self._substitute_action(
+                        action,
+                        gesture,
+                        original_type="click_sequence",
+                        source_action_ids=source_ids,
+                        reason="file_menu_save_to_ctrl_s",
+                    )
+                )
+                substitutions += 1
+                logger.debug("Séquence Fichier > Enregistrer substituée par Ctrl+S")
+                continue
+
             gesture = self.match_action(action)
             if gesture and action.get("type") != "key_combo":
                 # Substituer par le raccourci clavier
-                new_action = gesture.to_replay_action()
-                # Conserver l'action_id original pour le tracking
-                new_action["action_id"] = action.get("action_id", new_action["action_id"])
-                new_action["original_type"] = action.get("type")
+                new_action = self._substitute_action(
+                    action,
+                    gesture,
+                    original_type=action.get("type", ""),
+                    reason=f"{gesture.id}_gesture_substitution",
+                )
                 optimized.append(new_action)
                 substitutions += 1
                 logger.debug(