Update services/intent_classifier.py

services/intent_classifier.py

@@ -86,7 +86,7 @@ class IntentClassifier:
                     "suggested_response_type": "friendly"
                 }
         
-        # Question patterns (informational)
+        # Question patterns (informational) - CHECK THIS FIRST!
         is_question = False
         for pattern in IntentClassifier.QUESTION_PATTERNS:
             if re.search(pattern, message, re.IGNORECASE):
@@ -99,7 +99,37 @@ class IntentClassifier:
             if keyword in message
         )
         
-        # If has pipeline keywords, it's likely a pipeline request
+        # IMPORTANT: Questions take priority over pipeline keywords!
+        # "tell me about X" is a question, not a pipeline request
+        if is_question:
+            # If it's clearly asking about the system itself, it's informational
+            if any(term in message for term in ["masterllm", "you", "this", "system", "bot", "chatbot", "assistant"]):
+                return {
+                    "intent": "question",
+                    "confidence": 0.95,
+                    "requires_pipeline": False,
+                    "suggested_response_type": "informational"
+                }
+            # If it's a question but with strong pipeline keywords, could be pipeline request
+            # But only if the keywords are very explicit (e.g., "extract text from page 5")
+            elif pipeline_keyword_count >= 2 and not message.startswith("tell me"):
+                return {
+                    "intent": "pipeline_request",
+                    "confidence": 0.7,
+                    "requires_pipeline": True,
+                    "suggested_response_type": "pipeline_generation",
+                    "keyword_matches": pipeline_keyword_count
+                }
+            # Otherwise, it's just a question
+            else:
+                return {
+                    "intent": "question",
+                    "confidence": 0.85,
+                    "requires_pipeline": False,
+                    "suggested_response_type": "informational"
+                }
+        
+        # If has pipeline keywords but is NOT a question, then pipeline request
         if pipeline_keyword_count > 0:
             return {
                 "intent": "pipeline_request",
@@ -109,15 +139,6 @@ class IntentClassifier:
                 "keyword_matches": pipeline_keyword_count
             }
         
-        # Questions without pipeline keywords
-        if is_question:
-            return {
-                "intent": "question",
-                "confidence": 0.8,
-                "requires_pipeline": False,
-                "suggested_response_type": "informational"
-            }
-        
         # Default: treat as casual if short, otherwise might be pipeline request
         if len(message.split()) < 3:
             return {
@@ -183,9 +204,41 @@ class IntentClassifier:
 3. I'll create a pipeline for you to approve
 4. Watch the magic happen! ✨"""
         
-        # Who are you
-        if "who are you" in message_lower or "what are you" in message_lower:
-            return "I'm MasterLLM 🤖, an AI-powered document processing orchestrator. I use advanced AI models (Bedrock Claude & Google Gemini) to understand your requests and automatically create processing pipelines for your documents!"
+        # Who are you / What are you / Tell me about MasterLLM
+        if any(phrase in message_lower for phrase in ["who are you", "what are you", "about masterllm", "about you", "tell me about"]):
+            return """**About MasterLLM** 🤖
+
+I'm MasterLLM, an intelligent AI-powered document processing orchestrator that helps you automate complex document workflows.
+
+**What I Do:**
+- 🔍 **Analyze** your requests and understand what you need
+- 🛠️ **Create** custom processing pipelines automatically
+- ⚡ **Execute** multi-step document workflows
+- 📊 **Deliver** structured results
+
+**My Capabilities:**
+- Extract text from PDFs and images (OCR)
+- Summarize long documents
+- Translate content to different languages
+- Classify documents
+- Extract tables and structured data
+- Detect signatures and stamps
+- Describe images
+- Extract named entities (people, places, organizations)
+
+**How I'm Different:**
+I'm not just a single tool - I'm a **smart orchestrator** that:
+1. Understands your natural language requests
+2. Determines the best sequence of operations
+3. Uses powerful AI models (AWS Bedrock Claude & Google Gemini)
+4. Combines multiple tools into efficient pipelines
+5. Executes everything seamlessly
+
+**My Role:**
+Think of me as your AI assistant for document processing. I have a natural conversation with you, understand what you need, propose a plan, and execute it - all while keeping you informed every step of the way!
+
+Want to see me in action? Just upload a document and tell me what you'd like to do with it! 🚀"""
+
         
         # Help
         if message_lower in ["help", "?"] or "help me" in message_lower: