rpa_vision_v3/demo_fiche12_form_rows.py

#!/usr/bin/env python3
"""
Démonstration Fiche #12 - Form Rows/Columns

Auteur : Dom, Alice Kiro
Date : 19 décembre 2024
"""

from datetime import datetime
from core.execution.target_resolver import TargetResolver, ResolutionContext
from core.models.workflow_graph import TargetSpec
from core.models.screen_state import ScreenState, RawLevel, PerceptionLevel, ContextLevel, WindowContext, EmbeddingRef
from core.models.ui_element import UIElement, UIElementEmbeddings, VisualFeatures


def create_element(eid, role, bbox, label="", etype="ui", conf=0.95):
    """Helper pour créer un UIElement"""
    return UIElement(
        element_id=eid, type=etype, role=role, bbox=bbox,
        center=(bbox[0]+bbox[2]//2, bbox[1]+bbox[3]//2),
        label=label, label_confidence=1.0 if label else 0.0,
        embeddings=UIElementEmbeddings(image=None, text=None),
        visual_features=VisualFeatures(
            dominant_color="#ffffff", has_icon=False, shape="rectangle", size_category="medium"
        ),
        confidence=conf, tags=[], metadata={}
    )


def create_screen(elements):
    """Helper pour créer un ScreenState"""
    return ScreenState(
        screen_state_id="demo_screen", timestamp=datetime.now(), session_id="demo_session",
        window=WindowContext(app_name="Demo App", window_title="Form Demo", screen_resolution=[1920,1080]),
        raw=RawLevel(screenshot_path="demo.png", capture_method="test", file_size_bytes=1024),
        perception=PerceptionLevel(
            embedding=EmbeddingRef(provider="demo", vector_id="demo_vector", dimensions=512),
            detected_text=[], text_detection_method="ocr", confidence_avg=0.9
        ),
        context=ContextLevel(), ui_elements=elements
    )


def demo_form_login():
    """Démonstration : Formulaire de connexion classique"""
    print("🔐 Démonstration : Formulaire de connexion")
    print("=" * 50)
    
    # Créer un formulaire de connexion réaliste
    elements = [
        # Panel de connexion
        create_element("login_panel", "panel", (50, 50, 400, 300), "Login"),
        
        # Ligne 1 : Username
        create_element("lbl_username", "label", (80, 100, 100, 20), "Username"),
        create_element("inp_username", "input", (200, 95, 200, 30), "", etype="text_input"),
        
        # Ligne 2 : Password  
        create_element("lbl_password", "label", (80, 150, 100, 20), "Password"),
        create_element("inp_password", "input", (200, 145, 200, 30), "", etype="text_input"),
        
        # Ligne 3 : Boutons
        create_element("btn_login", "button", (200, 200, 80, 35), "Login"),
        create_element("btn_cancel", "button", (300, 200, 80, 35), "Cancel"),
        
        # Éléments distracteurs
        create_element("inp_search", "input", (500, 100, 150, 30), "", etype="text_input"),
        create_element("lbl_search", "label", (500, 80, 100, 20), "Search"),
    ]
    
    screen = create_screen(elements)
    resolver = TargetResolver()
    context = ResolutionContext(screen_state=screen)
    
    # Test 1 : Trouver le champ Username
    print("\n📝 Test 1 : Chercher le champ pour 'Username'")
    spec1 = TargetSpec(by_role="input", context_hints={"field_for": "Username"})
    result1 = resolver.resolve_target(spec1, screen, context)
    
    if result1:
        print(f"✅ Trouvé : {result1.element.element_id} (confiance: {result1.confidence:.2f})")
        print(f"   Stratégie : {result1.strategy_used}")
        print(f"   Ancre : {result1.resolution_details.get('anchor_id', 'N/A')}")
    else:
        print("❌ Aucun résultat")
    
    # Test 2 : Trouver le champ Password
    print("\n🔒 Test 2 : Chercher le champ pour 'Password'")
    spec2 = TargetSpec(by_role="input", context_hints={"field_for": "Password"})
    result2 = resolver.resolve_target(spec2, screen, context)
    
    if result2:
        print(f"✅ Trouvé : {result2.element.element_id} (confiance: {result2.confidence:.2f})")
        print(f"   Stratégie : {result2.strategy_used}")
        print(f"   Ancre : {result2.resolution_details.get('anchor_id', 'N/A')}")
    else:
        print("❌ Aucun résultat")
    
    # Test 3 : Bonus same_row_as_text
    print("\n🎯 Test 3 : Bouton sur même ligne que 'Cancel'")
    spec3 = TargetSpec(by_role="button", by_text="Login", context_hints={"same_row_as_text": "Cancel"})
    result3 = resolver.resolve_target(spec3, screen, context)
    
    if result3:
        print(f"✅ Trouvé : {result3.element.element_id} (confiance: {result3.confidence:.2f})")
        print(f"   Stratégie : {result3.strategy_used}")
    else:
        print("❌ Aucun résultat")


def demo_form_complex():
    """Démonstration : Formulaire complexe avec fallback"""
    print("\n\n🏢 Démonstration : Formulaire complexe")
    print("=" * 50)
    
    # Formulaire avec layout vertical (fallback)
    elements = [
        # Section 1 : Informations personnelles
        create_element("section1", "panel", (50, 50, 350, 200), "Personal Info"),
        
        # Nom (layout vertical - fallback)
        create_element("lbl_name", "label", (80, 80, 80, 20), "Full Name"),
        create_element("inp_name", "input", (80, 110, 250, 30), "", etype="text_input"),
        
        # Email (layout horizontal - priorité)
        create_element("lbl_email", "label", (80, 160, 60, 20), "Email"),
        create_element("inp_email", "input", (160, 155, 200, 30), "", etype="text_input"),
        
        # Section 2 : Adresse
        create_element("section2", "panel", (50, 280, 350, 150), "Address"),
        
        # Ville (avec contrainte de conteneur)
        create_element("lbl_city", "label", (80, 320, 60, 20), "City"),
        create_element("inp_city", "input", (160, 315, 150, 30), "", etype="text_input"),
        
        # Input distracteur hors section
        create_element("inp_other", "input", (500, 320, 150, 30), "", etype="text_input"),
    ]
    
    screen = create_screen(elements)
    resolver = TargetResolver()
    context = ResolutionContext(screen_state=screen)
    
    # Test 1 : Fallback vertical pour "Full Name"
    print("\n📝 Test 1 : Fallback vertical pour 'Full Name'")
    spec1 = TargetSpec(by_role="input", context_hints={"field_for": "Full Name"})
    result1 = resolver.resolve_target(spec1, screen, context)
    
    if result1:
        print(f"✅ Trouvé : {result1.element.element_id} (confiance: {result1.confidence:.2f})")
        print(f"   Mode : {'Même ligne' if result1.confidence >= 0.95 else 'Fallback vertical'}")
    else:
        print("❌ Aucun résultat")
    
    # Test 2 : Priorité horizontale pour "Email"
    print("\n📧 Test 2 : Priorité horizontale pour 'Email'")
    spec2 = TargetSpec(by_role="input", context_hints={"field_for": "Email"})
    result2 = resolver.resolve_target(spec2, screen, context)
    
    if result2:
        print(f"✅ Trouvé : {result2.element.element_id} (confiance: {result2.confidence:.2f})")
        print(f"   Mode : {'Même ligne' if result2.confidence >= 0.95 else 'Fallback vertical'}")
    else:
        print("❌ Aucun résultat")
    
    # Test 3 : Contrainte de conteneur
    print("\n🏠 Test 3 : Champ 'City' avec contrainte de conteneur")
    spec3 = TargetSpec(
        by_role="input", 
        context_hints={"field_for": "City"},
        hard_constraints={"within_container_text": "Address"}
    )
    result3 = resolver.resolve_target(spec3, screen, context)
    
    if result3:
        print(f"✅ Trouvé : {result3.element.element_id} (confiance: {result3.confidence:.2f})")
        print(f"   Respecte contrainte conteneur : ✅")
    else:
        print("❌ Aucun résultat")


def demo_multi_anchor():
    """Démonstration : Support multi-anchor"""
    print("\n\n🌍 Démonstration : Support multi-anchor (multilingue)")
    print("=" * 60)
    
    # Interface multilingue
    elements = [
        # Labels en français
        create_element("lbl_identifiant", "label", (80, 100, 120, 20), "Identifiant"),
        create_element("inp_user", "input", (220, 95, 200, 30), "", etype="text_input"),
        
        # Labels en anglais (autre section)
        create_element("lbl_password_en", "label", (80, 150, 100, 20), "Password"),
        create_element("inp_pass", "input", (200, 145, 200, 30), "", etype="text_input"),
    ]
    
    screen = create_screen(elements)
    resolver = TargetResolver()
    context = ResolutionContext(screen_state=screen)
    
    # Test multi-anchor : chercher Username OU Identifiant
    print("\n🔍 Test : Chercher champ pour ['Username', 'Identifiant']")
    spec = TargetSpec(
        by_role="input", 
        context_hints={"field_for": ["Username", "Identifiant", "User"]}
    )
    result = resolver.resolve_target(spec, screen, context)
    
    if result:
        print(f"✅ Trouvé : {result.element.element_id} (confiance: {result.confidence:.2f})")
        print(f"   Ancre utilisée : {result.resolution_details.get('anchor_id', 'N/A')}")
        print(f"   Labels cherchés : {result.resolution_details.get('criteria_used', {}).get('field_for', [])}")
    else:
        print("❌ Aucun résultat")


if __name__ == "__main__":
    print("🚀 Démonstration Fiche #12 - Form Rows/Columns")
    print("Auteur : Dom, Alice Kiro - 19 décembre 2024")
    print("=" * 60)
    
    try:
        demo_form_login()
        demo_form_complex()
        demo_multi_anchor()
        
        print("\n\n🎉 Démonstration terminée avec succès !")
        print("La Fiche #12 fonctionne parfaitement pour l'association label→champ.")
        
    except Exception as e:
        print(f"\n❌ Erreur durant la démonstration : {e}")
        import traceback
        traceback.print_exc()