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/search_vase_v1.11.py

@@ -0,0 +1,80 @@
+#!/opt/homebrew/bin/python3.11
+"""
+Search for "vase" in the video using OWL-ViT on a subset of frames.
+"""
+
+import os
+import cv2
+import glob
+from PIL import Image
+import torch
+from transformers import OwlViTProcessor, OwlViTForObjectDetection
+
+BASE_DIR = "output/384b0ff44aaaa1f1/full_video_scans"
+RESULTS_DIR = "output/384b0ff44aaaa1f1/vase_search_results"
+os.makedirs(RESULTS_DIR, exist_ok=True)
+
+print("🔍 Searching for vases...")
+
+# Load model
+processor = OwlViTProcessor.from_pretrained("google/owlvit-base-patch32")
+model = OwlViTForObjectDetection.from_pretrained("google/owlvit-base-patch32")
+model.eval()
+
+# Search terms
+SEARCH_TERMS = ["vase", "flower vase", "urn", "pottery", "glass jar"]
+
+frames = sorted(glob.glob(os.path.join(BASE_DIR, "frame_*.jpg")))
+print(f"📸 Scanning {len(frames)} frames...")
+
+found_count = 0
+
+for frame_path in frames:
+    frame_name = os.path.basename(frame_path)
+    sec = frame_name.replace("frame_", "").replace("s.jpg", "")
+
+    image = Image.open(frame_path).convert("RGB")
+    h, w = image.height, image.width
+    target_sizes = torch.Tensor([[h, w]])
+
+    for term in SEARCH_TERMS:
+        inputs = processor(text=[[term]], images=image, return_tensors="pt")
+
+        with torch.no_grad():
+            outputs = model(**inputs)
+
+        results = processor.post_process_object_detection(
+            outputs=outputs, target_sizes=target_sizes, threshold=0.05
+        )
+
+        for score, label, box in zip(
+            results[0]["scores"], results[0]["labels"], results[0]["boxes"]
+        ):
+            s = float(score)
+            if s > 0.08:  # Threshold for visualization
+                x1, y1, x2, y2 = map(int, box.tolist())
+                img = cv2.imread(frame_path)
+                crop = img[y1:y2, x1:x2]
+
+                if crop.size > 0:
+                    crop_name = f"vase_{sec}s_{term.replace(' ', '_')}_{s:.2f}.jpg"
+                    cv2.imwrite(os.path.join(RESULTS_DIR, crop_name), crop)
+
+                    # Annotate
+                    cv2.rectangle(img, (x1, y1), (x2, y2), (0, 0, 255), 3)
+                    cv2.putText(
+                        img,
+                        f"{term} {s:.2f}",
+                        (x1, y1 - 10),
+                        cv2.FONT_HERSHEY_SIMPLEX,
+                        0.7,
+                        (0, 0, 255),
+                        2,
+                    )
+                    ann_name = f"annotated_{sec}s.jpg"
+                    cv2.imwrite(os.path.join(RESULTS_DIR, ann_name), img)
+
+                    print(f"  📍 {sec}s | {term} | {s:.2f}")
+                    found_count += 1
+
+print(f"\n🏁 Done. Found {found_count} candidates.")