v1.0 - Version stable: multi-PC, détection UI-DETR-1, 3 modes exécution

- Frontend v4 accessible sur réseau local (192.168.1.40)
- Ports ouverts: 3002 (frontend), 5001 (backend), 5004 (dashboard)
- Ollama GPU fonctionnel
- Self-healing interactif
- Dashboard confiance

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

tests/unit/test_cross_frame_cache.py

@@ -0,0 +1,55 @@
+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 E(eid, role, bbox, label="", etype="ui"):
+    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,
+        embeddings=UIElementEmbeddings(image=None, text=None),
+        visual_features=VisualFeatures(dominant_color="n/a", has_icon=False, shape="rectangle", size_category="medium"),
+        confidence=0.95, tags=[], metadata={}
+    )
+
+def S(elements, state_id="s"):
+    return ScreenState(
+        screen_state_id=state_id,
+        timestamp=datetime.now(),
+        session_id="sess",
+        window=WindowContext(app_name="app", window_title="Login", screen_resolution=[1920,1080]),
+        raw=RawLevel(screenshot_path="x", capture_method="test", file_size_bytes=1),
+        perception=PerceptionLevel(embedding=EmbeddingRef(provider="p", vector_id="v", dimensions=1),
+                                  detected_text=[], text_detection_method="none", confidence_avg=0.0),
+        context=ContextLevel(),
+        ui_elements=elements
+    )
+
+def test_cross_frame_cache_near_bbox_finds_new_id():
+    r = TargetResolver()
+    
+    # Frame 1: input id = inp_1
+    ui1 = [
+        E("lbl", "label", (100,100,120,20), "Username", "label"),
+        E("inp_1", "input", (240,95,260,30), "", "text_input"),
+    ]
+    s1 = S(ui1, "s1")
+    spec = TargetSpec(by_role="input", context_hints={"field_for": "Username"})
+    
+    res1 = r.resolve_target(spec, s1, ResolutionContext(screen_state=s1, previous_target=None))
+    assert res1 is not None
+    assert res1.element.element_id == "inp_1"
+    
+    # Frame 2: même position mais nouvel id = inp_X (simule perception qui renumérote)
+    ui2 = [
+        E("lbl2", "label", (102,98,120,20), "Username", "label"),
+        E("inp_X", "input", (242,96,260,30), "", "text_input"),
+    ]
+    s2 = S(ui2, "s2")
+    
+    res2 = r.resolve_target(spec, s2, ResolutionContext(screen_state=s2, previous_target=None))
+    assert res2 is not None
+    assert res2.element.element_id == "inp_X"
+    assert res2.strategy_used in {"CROSS_FRAME_CACHE", "COMPOSITE"}