Parallelize chunk moderation and restrict connection indicator

Bentlybro · Bentlybro · commit 9e632da8e6e1 · 2025-10-14T15:39:36.000+01:00
Updated ai_moderator.py to process content chunks in parallel using ThreadPoolExecutor for improved performance. Modified base.html to display the real-time connection indicator only for authenticated users.
diff --git a/app/services/ai/ai_moderator.py b/app/services/ai/ai_moderator.py
@@ -1,5 +1,6 @@
 import json
 import time
+from concurrent.futures import ThreadPoolExecutor, as_completed
 
 import openai
 import tiktoken
@@ -310,14 +311,19 @@ def moderate_content(self, content, content_type='text', custom_prompt=None):
                     for i in range(0, content_chars, MAX_CHARS_PER_CHUNK):
                         chunks.append(content[i:i + MAX_CHARS_PER_CHUNK])
 
+                    # Process all chunks IN PARALLEL for maximum speed
                     chunk_results = []
-                    for i, chunk in enumerate(chunks):
-                        result = self._analyze_with_custom_prompt(chunk, custom_prompt)
-                        chunk_results.append(result)
-
-                        # Early exit if chunk is rejected
-                        if result['decision'] == 'rejected':
-                            break
+                    with ThreadPoolExecutor(max_workers=min(len(chunks), 10)) as executor:
+                        # Submit all chunks at once
+                        future_to_chunk = {
+                            executor.submit(self._analyze_with_custom_prompt, chunk, custom_prompt): i
+                            for i, chunk in enumerate(chunks)
+                        }
+
+                        # Collect results as they complete
+                        for future in as_completed(future_to_chunk):
+                            result = future.result()
+                            chunk_results.append(result)
 
                     return self._combine_chunk_results(chunk_results, len(content))
                 else:
@@ -340,14 +346,19 @@ def moderate_content(self, content, content_type='text', custom_prompt=None):
                 # Split content and analyze each chunk
                 chunks = self.split_text_into_chunks(content, max_content_tokens)
 
+                # Process all chunks IN PARALLEL for maximum speed
                 chunk_results = []
-                for i, chunk in enumerate(chunks):
-                    result = self._run_enhanced_default_moderation(chunk)
-                    chunk_results.append(result)
-
-                    # Early exit if chunk is rejected (for efficiency)
-                    if result['decision'] == 'rejected':
-                        break
+                with ThreadPoolExecutor(max_workers=min(len(chunks), 10)) as executor:
+                    # Submit all chunks at once
+                    future_to_chunk = {
+                        executor.submit(self._run_enhanced_default_moderation, chunk): i
+                        for i, chunk in enumerate(chunks)
+                    }
+
+                    # Collect results as they complete
+                    for future in as_completed(future_to_chunk):
+                        result = future.result()
+                        chunk_results.append(result)
 
                 return self._combine_chunk_results(chunk_results, len(content))
 
diff --git a/app/templates/base.html b/app/templates/base.html
@@ -40,12 +40,14 @@
     <link href="{{ url_for('static', filename='css/utilities/responsive.css') }}" rel="stylesheet">
 </head>
 <body>
-    <!-- Real-time connection indicator -->
+    <!-- Real-time connection indicator (only for authenticated users) -->
+    {% if current_user.is_authenticated %}
     <div id="connection-status" class="real-time-indicator">
         <span class="badge bg-secondary" style="font-size: 0.75rem; padding: 0.25rem 0.5rem;">
             <i class="fas fa-circle"></i> <span class="connection-text">Connecting...</span>
         </span>
     </div>
+    {% endif %}
 
     <!-- Toast Notification Container -->
     <div aria-live="polite" aria-atomic="true" class="position-fixed top-0 end-0 p-3" style="z-index: 1080; min-width: 300px;">
