Update app.py

app.py

@@ -388,93 +388,77 @@ class RAGIndex:
         return answer
 
     def answer(self, question: str) -> str:
-        """Answer a question using RAG - simplified extractive approach."""
+        """Answer a question using RAG with a generative seq2seq model (Flan-T5, BART, etc.)."""
         if not self.initialized:
             return "❌ Assistant not properly initialized. Please check the logs."
-    
+
         if not question or not question.strip():
             return "Please ask a question."
-    
+
         if self.index is None or not self.chunks:
             return (
                 f"📚 Knowledge base is empty.\n\n"
                 f"Please add documents to: `{KB_DIR}`\n"
                 f"Supported formats: .txt, .md, .pdf, .docx"
             )
-    
-        # Retrieve relevant contexts
+
+        # 1) Retrieve relevant contexts
         contexts = self.retrieve(question, top_k=3)
-    
+
         if not contexts:
             return (
                 f"{NO_ANSWER_MSG}\n\n"
                 f"💡 Try rephrasing your question or check if relevant documents exist in the knowledge base."
             )
-    
+
         used_sources = set()
-        best_context = None
-        best_score = 0
-        
-        # Find the best matching context
+        context_texts = []
+
         for ctx, source, score in contexts:
             used_sources.add(source)
-            if score > best_score:
-                best_score = score
-                best_context = ctx
-        
-        if not best_context:
-            return f"{NO_ANSWER_MSG}"
-        
-        # AGGRESSIVE cleaning of the context
-        def deep_clean(text):
-            """Remove ALL markdown, bullets, numbers, emojis, and formatting."""
-            # Remove emojis and special characters
-            text = re.sub(r'[📘📄🟢🟡🟠✓✗❌⚠️💡📚]', '', text)
-            # Remove markdown headers (# ## ###)
-            text = re.sub(r'^#{1,6}\s+', '', text, flags=re.MULTILINE)
-            # Remove numbered lists (1. 2. 3.)
-            text = re.sub(r'^\s*\d+\.\s+', '', text, flags=re.MULTILINE)
-            # Remove bullet points (- * •)
-            text = re.sub(r'^\s*[-*•]\s+', '', text, flags=re.MULTILINE)
-            # Remove bold/italic (**text** *text*)
-            text = re.sub(r'\*\*([^*]+)\*\*', r'\1', text)
-            text = re.sub(r'\*([^*]+)\*', r'\1', text)
-            # Remove extra colons from labels
-            text = re.sub(r':\s*$', '', text, flags=re.MULTILINE)
-            # Clean multiple spaces
-            text = re.sub(r'\s+', ' ', text)
-            # Remove "Good:" "Bad:" type prefixes
-            text = re.sub(r'^(Good|Bad|Example|Note):\s*', '', text, flags=re.MULTILINE)
-            return text.strip()
-        
-        cleaned = deep_clean(best_context)
-        
-        # Extract just the most relevant sentences (3-4 sentences max)
-        sentences = [s.strip() + '.' for s in cleaned.split('.') if len(s.strip()) > 20]
-        answer_text = ' '.join(sentences[:4])  # First 4 good sentences
-        
-        # If we got good text, try to generate a natural answer
-        if len(answer_text) > 50:
-            # Simple prompt for FLAN-T5
-            prompt = f"Question: {question}\n\nInformation: {answer_text[:800]}\n\nWrite a clear answer in 2-3 sentences:"
-            
-            try:
-                generated = self._generate_from_context(prompt, max_new_tokens=150)
-                generated = generated.strip()
-                
-                # Only use generated answer if it looks good
-                if (len(generated) > 30 and 
-                    not generated.startswith(("Do NOT", "You are", "##", "**")) and
-                    generated.count(':') < 3):
-                    answer_text = generated
-                # Otherwise, keep the cleaned extractive answer
-                    
-            except Exception as e:
-                print(f"Generation error (using extractive fallback): {e}")
-                # Keep the cleaned extractive answer
-        
+            cleaned_ctx = clean_context_text(ctx)
+            if cleaned_ctx:
+                context_texts.append(cleaned_ctx)
+
+        if not context_texts:
+            return (
+                f"{NO_ANSWER_MSG}\n\n"
+                f"💡 Try adding more detailed documents to the knowledge base."
+            )
+
+        # 2) Combine contexts into a single evidence block
+        combined_context = "\n\n".join(context_texts)
+
+        # Keep context at a reasonable size for the model
+        max_context_chars = 3000
+        if len(combined_context) > max_context_chars:
+            combined_context = combined_context[:max_context_chars]
+
+        # 3) Build a prompt that works for both BART (summarization-style)
+        #    and instruction-tuned models like Flan-T5.
+        prompt = (
+            "You are an assistant that answers questions about a knowledge base.\n"
+            "Using only the information in the passages below, answer the question in 2–4 sentences.\n"
+            "Explain in clear, natural language. Do NOT copy section numbers, markdown headings, or bullet symbols.\n\n"
+            f"Passages:\n{combined_context}\n\n"
+            f"Question: {question}\n\n"
+            "Answer:"
+        )
+
+        try:
+            answer_text = self._generate_from_context(prompt, max_new_tokens=180).strip()
+        except Exception as e:
+            print(f"Generation error: {e}")
+            return (
+                "There was an error while generating the answer. "
+                "Please try again with a shorter question or different wording."
+            )
+
+        if not answer_text:
+            answer_text = NO_ANSWER_MSG
+
         sources_str = ", ".join(sorted(used_sources)) if used_sources else "N/A"
-        
+
         return (
             f"**Answer:** {answer_text}\n\n"
             f"**Sources:** {sources_str}"