diff --git a/services/dashboard/VoiceControls.jsx b/services/dashboard/VoiceControls.jsx
index 7365cc7..491789f 100644
--- a/services/dashboard/VoiceControls.jsx
+++ b/services/dashboard/VoiceControls.jsx
@@ -22,7 +22,9 @@ const VoiceControls = ({ onCommand, hotkey = ' ' }) => {
       return;
     }
 
-    // Request microphone permission
+    // Note: We request permission on mount for better UX.
+    // Alternative: Request only on first use by removing this and letting
+    // startListening() handle the permission request
     requestMicrophonePermission();
 
     // Setup hotkey listener
diff --git a/services/hackgpt-api/app/config_validator.py b/services/hackgpt-api/app/config_validator.py
index d2c4514..c4a8304 100644
--- a/services/hackgpt-api/app/config_validator.py
+++ b/services/hackgpt-api/app/config_validator.py
@@ -467,7 +467,10 @@ def _is_valid_target(target: str) -> bool:
     ip_pattern = r'^(\d{1,3}\.){3}\d{1,3}$'
     if re.match(ip_pattern, target):
         parts = target.split('.')
-        return all(0 <= int(part) <= 255 for part in parts)
+        try:
+            return all(0 <= int(part) <= 255 for part in parts)
+        except ValueError:
+            return False
     
     # CIDR notation
     if '/' in target:
@@ -522,7 +525,10 @@ def _is_valid_ip_or_hostname(host: str) -> bool:
     ip_pattern = r'^(\d{1,3}\.){3}\d{1,3}$'
     if re.match(ip_pattern, host):
         parts = host.split('.')
-        return all(0 <= int(part) <= 255 for part in parts)
+        try:
+            return all(0 <= int(part) <= 255 for part in parts)
+        except ValueError:
+            return False
     
     # Hostname
     hostname_pattern = r'^([a-zA-Z0-9]([a-zA-Z0-9\-]{0,61}[a-zA-Z0-9])?\.)*[a-zA-Z0-9]([a-zA-Z0-9\-]{0,61}[a-zA-Z0-9])?$'
diff --git a/services/hackgpt-api/app/llm_help.py b/services/hackgpt-api/app/llm_help.py
index 370e40a..fb2a733 100644
--- a/services/hackgpt-api/app/llm_help.py
+++ b/services/hackgpt-api/app/llm_help.py
@@ -11,7 +11,10 @@ import json
 
 
 # Store conversation history per session
+# Note: In production, use Redis or similar with TTL for scalability
+# This simple in-memory dict will grow unbounded - implement cleanup as needed
 conversation_contexts: Dict[str, List[Dict[str, str]]] = {}
+MAX_SESSIONS = 1000  # Limit number of concurrent sessions
 
 
 async def chat_completion(
@@ -87,6 +90,11 @@ Always emphasize ethical hacking practices and legal considerations."""
                 # Store in conversation history
                 if session_id:
                     if session_id not in conversation_contexts:
+                        # Cleanup old sessions if limit reached
+                        if len(conversation_contexts) >= MAX_SESSIONS:
+                            # Remove oldest session (simple FIFO)
+                            oldest_session = next(iter(conversation_contexts))
+                            del conversation_contexts[oldest_session]
                         conversation_contexts[session_id] = []
                     conversation_contexts[session_id].append({"role": "user", "content": message})
                     conversation_contexts[session_id].append({"role": "assistant", "content": assistant_message})
diff --git a/services/hackgpt-api/app/voice.py b/services/hackgpt-api/app/voice.py
index fe1d97d..3570cf9 100644
--- a/services/hackgpt-api/app/voice.py
+++ b/services/hackgpt-api/app/voice.py
@@ -91,8 +91,8 @@ def _transcribe_with_local_whisper(audio_data: bytes, format: str) -> Dict[str,
             # Clean up temp file
             try:
                 os.unlink(temp_audio_path)
-            except:
-                pass
+            except (OSError, FileNotFoundError) as e:
+                print(f"Warning: Could not delete temp file: {e}")
                 
     except ImportError:
         raise Exception("Whisper not installed. Install with: pip install openai-whisper")