Upload 3 files

output.py

@@ -69,7 +69,10 @@ def generate_conversation_path(cursor, start_convo_id, start_dialogue_id, max_le
             if userscript:
                 placeholder_text += f"Script: {userscript}; "
             placeholder_text = placeholder_text.rstrip("; ") + "]"
-            path.append({"actor": actor_name, "dialogue": placeholder_text})
+            if placeholder_text == "[Action/Check:]":
+                pass
+            else:
+                path.append({"actor": actor_name, "dialogue": placeholder_text})
         else:
             path.append({"actor": actor_name, "dialogue": dialogue_text})
 
@@ -123,8 +126,9 @@ def main():
     num_paths_to_generate = 1000
     paths_generated = 0
     
-    # Initialize Bloom filter
-    bloom = ScalableBloomFilter(initial_capacity=100000, error_rate=0.001)
+    # Track seen conversation patterns
+    seen_windows = set()
+    min_window_size = 5  # Minimum consecutive dialogues to consider for duplication
     
     # Either pick a random starting point:
     starting_dialogues = get_starting_dialogues(cursor)
@@ -138,14 +142,47 @@ def main():
             start_convo_id, start_dialogue_id = random.choice(starting_dialogues)
             path = generate_conversation_path(cursor, start_convo_id, start_dialogue_id)
 
-            # Create unique key for conversation
-            path_text = "".join(entry["dialogue"] for entry in path)
-            path_hash = hashlib.sha256(path_text.encode()).hexdigest()
+            # Skip paths that are too short
+            if len(path) < 5:
+                continue
+
+            # Create semantic key for conversation
+            path_text = " ".join(entry["dialogue"].strip().lower() for entry in path if not entry["dialogue"].startswith("["))
+            
+            def is_duplicate(new_text):
+                """Check if new conversation contains or is contained within existing conversations"""
+                for existing in conversations:
+                    existing_text = " ".join(entry["dialogue"].strip().lower() for entry in existing if not entry["dialogue"].startswith("["))
+                    if len(new_text) < min_window_size or len(existing_text) < min_window_size:
+                        continue
+                    # Check for substring matches in either direction
+                    if new_text in existing_text or existing_text in new_text:
+                        return True
+                return False
             
-            # Check for duplicates using Bloom filter
+            # More efficient window-based check
+            def is_added(current_path):
+                """Check for matching subsequences using sliding window"""
+                path_dialogues = [entry["dialogue"] for entry in current_path if not entry["dialogue"].startswith("[")]
+                window = []
+                for dialogue in path_dialogues:
+                    window.append(dialogue)
+                    if len(window) > min_window_size:
+                        window.pop(0)
+                    if len(window) == min_window_size:
+                        window_str = " ".join(window)
+                        if window_str in seen_windows:
+                            return True
+                # Add windows if not duplicate
+                for i in range(len(path_dialogues) - min_window_size + 1):
+                    seen_windows.add(" ".join(path_dialogues[i:i+min_window_size]))
+                return False
+
+            # Check for duplicates
             try:
-                if path_hash not in bloom and any(entry["dialogue"][0] != "[" for entry in path):
-                    bloom.add(path_hash)
+                if (not is_duplicate(path_text) and 
+                    not is_added(path) and 
+                    any(entry["dialogue"][0] != "[" for entry in path)):
                     conversations.append(path)
                     paths_generated += 1
                     pbar.update(1)