Update app.py

app.py

@@ -162,38 +162,44 @@ def search_memory():
         return jsonify({"error": "No query provided"}), 400
 
     try:
-           # 1. Refined Request
-        search_req = {
-            "top_k": 10,           # Ask for more...
-            "snippet_chars": 150,  # ...but get smaller chunks
-            "no_sketch": False     # Use the sketch track for speed
-        }
+        # 1. Search
+        response = db.find(query, top_k=10, snippet_chars=150)
+        
+        # 2. Parse & Clean
+        clean_results = []
+        hits = response.get('hits', [])
         
-        #response = db.find(query, **search_req)
-        response = db.find(query)
-        print(response)
-        #{'query': 'car', 'hits': [], 'took_ms': 16, 'total_hits': 0, 'engine': 'tantivy', 'context': ''}
+        for hit in hits:
+            score = hit.get('score', 0.0)
+            if score < 0.65: continue
 
+            # --- CLEANING LOGIC ---
+            # 1. Get raw snippet
+            raw_snippet = hit.get('snippet', '')
+            
+            # 2. Split by lines and remove technical metadata headers
+            # (The raw snippet appends metadata at the bottom)
+            lines = raw_snippet.split('\n')
+            content_lines = [
+                line for line in lines 
+                if not line.strip().startswith(('title:', 'tags:', 'labels:', 'extractous_metadata:'))
+            ]
+            clean_text = "\n".join(content_lines).strip()
+            
+            # 3. Use the explicit arrays provided by the SDK for tags/labels
+            # (These are much cleaner than parsing the string)
+            tags = hit.get('tags', [])
+            labels = hit.get('labels', [])
 
-        # 2. Refined Parsing (The Anti-Junk Filter)
-        clean_results = []
-        for hit in response['hits']:
-            print(hit)
-            # Filter low relevance
-            if hit['score'] < 0.65: 
-                continue
-            '''{'frame_id': 0, 'uri': 'mv2://frames/0', 'title': 'Untitled', 'rank': 1, 'score': 5.422323226928711, 'matches': 3, 'snippet': 'the dog is blue\ntitle: Untitled\ntags: blue dog\nlabels: text Blue Dog\nextractous_metadata: {"coverage":1.0,"sections_extracted":1,"sections_total":1,"skim":false}', 'tags': ['blue', 'dog'], 'labels': ['text', 'Blue', 'Dog'], 'track': None, 'created_at': '2026-01-11T03:33:52Z', 'content_dates': []}'''
             clean_results.append({
-                "title": hit['title'] or "Untitled Memory",
-                "snippet": hit['snippet'],   # The highlighted text
-                #"full_text": hit['text'],    # Available on click
-                #"full_text": hit['text'],    
-                "date": hit['created_at'],    # Context!
-                "score": f"{hit['score']:.2f}"
+                "title": hit.get('title') or "Untitled Memory",
+                "text": clean_text,          # The cleaned up content
+                "tags": tags,                # List of strings ['blue', 'dog']
+                "labels": labels,            # List of strings ['text', 'Blue']
+                "date": hit.get('created_at', ''),
+                "score": f"{score:.2f}"
             })
             
-        #return jsonify(clean_results)
-            
         return jsonify({"success": True, "results": clean_results})
     except Exception as e:
         return jsonify({"error": str(e)}), 500