Update app.py

app.py

@@ -47,6 +47,49 @@ prompt = ChatPromptTemplate.from_messages([
 question_answer_chain = create_stuff_documents_chain(chatModel, prompt)
 rag_chain = create_retrieval_chain(retriever, question_answer_chain)
 
+def format_citation(filename):
+    """
+    Convert filename format 'Author_Year - Title' to 'Author (Year) – Title'
+    Handles 'et al.' and '&' in author names properly
+    
+    Examples:
+    'Ng et al._2023 - Problems and Solutions' -> 'Ng et al. (2023) – Problems and Solutions'
+    'Godley & Xia_2016 - Physics Guide' -> 'Godley & Xia (2016) – Physics Guide'
+    'Khan (2003) - Therapy' -> 'Khan (2003) – Therapy'
+    """
+    # Remove file extension if present
+    filename = re.sub(r'\.(pdf|txt)$', '', filename, flags=re.IGNORECASE)
+    
+    # Pattern 1: Handle "Author_YEAR - Title" format
+    match = re.match(r'^(.+?)_(\d{4})\s*-\s*(.+)$', filename)
+    if match:
+        author = match.group(1)
+        year = match.group(2)
+        title = match.group(3)
+        
+        # Replace underscores in author names with spaces
+        author = author.replace('_', ' ')
+        
+        # Format: Author (Year) – Title
+        return f"{author} ({year}) – {title}"
+    
+    # Pattern 2: Already has parentheses "Author (YEAR) - Title"
+    match = re.match(r'^(.+?)\s*\((\d{4})\)\s*-\s*(.+)$', filename)
+    if match:
+        author = match.group(1)
+        year = match.group(2)
+        title = match.group(3)
+        
+        # Replace underscores and hyphens with proper formatting
+        author = author.replace('_', ' ')
+        return f"{author} ({year}) – {title}"
+    
+    # Pattern 3: Just clean up underscores and hyphens in any format
+    filename = filename.replace('_', ' ')
+    filename = re.sub(r'\s*-\s*', ' – ', filename)
+    
+    return filename
+
 def format_latex_for_gradio(text):
     """Convert LaTeX delimiters to Gradio-friendly format while preserving markdown"""
     # Convert display math \[ ... \] to $$ ... $$
@@ -68,10 +111,15 @@ def chat_function(message, history):
     # Extract sources from retrieved documents
     source_documents = response.get("context", [])
     unique_sources = []
+    seen_sources = set()
+    
     for doc in source_documents:
         source = doc.metadata.get('source', None)
-        if source and source not in unique_sources:
-            unique_sources.append(source)
+        if source and source not in seen_sources:
+            # Format the citation properly
+            formatted_source = format_citation(source)
+            unique_sources.append(formatted_source)
+            seen_sources.add(source)
     
     # Remove ANY existing "Sources:" section that the LLM generated
     if "Sources:" in final_answer: