feat: Phase 2.6 edges migration to Qdrant (TKG-only architecture)

Phase 2.6.1: co_occurrence_edges migration
- build_co_occurrence_edges_from_qdrant()
- Qdrant embeddings → frame grouping → YOLO objects
- Result: 6679 edges (vs 6701 PostgreSQL)

Phase 2.6.2: face_face_edges migration
- build_face_face_edges_from_qdrant()
- Qdrant embeddings → frame grouping → face pairs
- mutual_gaze detection preserved
- Result: 6 edges (exact match)

Phase 2.6.3: speaker_face_edges migration
- build_speaker_face_edges_from_qdrant()
- Qdrant embeddings → trace_id frame ranges
- SPEAKS_AS edge creation

Architecture:
- All edges use Qdrant payload (no face_detections queries)
- PostgreSQL fallback for empty Qdrant
- Estimated 3.6x performance improvement

Testing:
- Playground (3003): ✓ All Phase 2.6 logs verified
- Edge counts: ✓ Close match with PostgreSQL
- Fallback: ✓ Working

Docs:
- docs_v1.0/DESIGN/TKG_PHASE2_6_EDGES_MIGRATION.md
- docs_v1.0/M4_workspace/2026-06-21_phase2_6_test.md

v1.1/scripts/visualize_stamp_v1.11.py

@@ -0,0 +1,45 @@
+#!/opt/homebrew/bin/python3.11
+"""
+Draw Detection Result on Image
+"""
+
+import cv2
+import os
+
+UUID = "384b0ff44aaaa1f1"
+OUTPUT_DIR = f"output/{UUID}/florence2_results"
+INPUT_IMG = os.path.join(OUTPUT_DIR, "raw_6846.jpg")
+OUTPUT_IMG = os.path.join(OUTPUT_DIR, "raw_6846_detected.jpg")
+
+# Florence-2 Result from previous run: [x_min, y_min, x_max, y_max]
+# Note: Florence-2 usually returns coordinates in normalized 0-1000 scale.
+box_raw = [1721.28, 23.22, 1813.44, 173.34]
+
+# Image dimensions
+img_width, img_height = 1920, 1080
+
+print(f"🎨 Drawing box on {INPUT_IMG}...")
+
+img = cv2.imread(INPUT_IMG)
+if img is None:
+    print("❌ Failed to load image.")
+    exit()
+
+# Convert normalized coordinates to pixel coordinates
+# Florence-2 uses 1000x1000 normalized coordinates
+x1 = int((box_raw[0] / 1000.0) * img_width)
+y1 = int((box_raw[1] / 1000.0) * img_height)
+x2 = int((box_raw[2] / 1000.0) * img_width)
+y2 = int((box_raw[3] / 1000.0) * img_height)
+
+print(f"📍 Pixel Coordinates: ({x1}, {y1}) to ({x2}, {y2})")
+
+# Draw Rectangle (Color: Green, Thickness: 4)
+cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 4)
+
+# Add Label
+cv2.putText(img, "STAMP", (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), 2)
+
+# Save
+cv2.imwrite(OUTPUT_IMG, img)
+print(f"✅ Saved result to {OUTPUT_IMG}")