Update app.py

app.py

@@ -1,26 +1,18 @@
 import os
 import sys
 import logging
+from pathlib import Path
 import json
-import traceback
 from datetime import datetime
-from typing import List, Dict, Any, Optional, Tuple
-
-# Third-party libraries
-import torch
-import numpy as np
-from sentence_transformers import SentenceTransformer
-import chromadb
-from chromadb.utils import embedding_functions
-import gradio as gr
-from openai import OpenAI
-import google.generativeai as genai
+from typing import List, Dict, Any, Optional, Tuple, Union
+import traceback
 
-# ----------------- Logging Configuration -----------------
+# Configure detailed logging with file output
 LOG_DIR = "logs"
 os.makedirs(LOG_DIR, exist_ok=True)
 log_file = os.path.join(LOG_DIR, f"rag_system_{datetime.now().strftime('%Y%m%d_%H%M%S')}.log")
 
+# Set up root logger with both file and console handlers
 logging.basicConfig(
     level=logging.INFO,
     format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
@@ -32,21 +24,58 @@ logging.basicConfig(
 logger = logging.getLogger("rag_system")
 logger.info(f"Starting RAG system. Log file: {log_file}")
 
-# ----------------- Configuration Class -----------------
+# Importing necessary libraries with error handling
+try:
+    import torch
+    import numpy as np
+    from sentence_transformers import SentenceTransformer
+    import chromadb
+    from chromadb.utils import embedding_functions
+    import gradio as gr
+    from openai import OpenAI
+    import google.generativeai as genai
+    logger.info("All required libraries successfully imported")
+except ImportError as e:
+    logger.critical(f"Failed to import required libraries: {e}")
+    print(f"ERROR: Missing required libraries. Please install with: pip install -r requirements.txt")
+    print(f"Specific error: {e}")
+    sys.exit(1)
+
+# Version info for tracking
+VERSION = "1.1.0"
+logger.info(f"RAG System Version: {VERSION}")
+
 class Config:
     """
     Configuration for vector store and RAG system.
+    
+    This class centralizes all configuration parameters for the application,
+    making it easier to modify settings and ensure consistency.
+    
+    Attributes:
+        local_dir (str): Directory for ChromaDB persistence
+        embedding_model (str): Name of the embedding model to use
+        collection_name (str): Name of the ChromaDB collection
+        default_top_k (int): Default number of results to return
+        openai_model (str): Default OpenAI model to use
+        gemini_model (str): Default Gemini model to use
+        temperature (float): Temperature setting for LLM generation
+        max_tokens (int): Maximum tokens for LLM response
+        system_name (str): Name of the system for UI
+        context_limit (int): Maximum characters to include in context
     """
+    
     def __init__(self, 
                  local_dir: str = "./chroma_db",
                  embedding_model: str = "all-MiniLM-L6-v2",
                  collection_name: str = "markdown_docs",
-                 default_top_k: int = 5,
+                 default_top_k: int = 8,  # Increased from 5 to 8 for more context
                  openai_model: str = "gpt-4o-mini",
                  gemini_model: str = "gemini-1.5-flash",
                  temperature: float = 0.3,
-                 max_tokens: int = 1000,
-                 system_name: str = "Document RAG System"):
+                 max_tokens: int = 2000,  # Increased from 1000 to 2000 for more comprehensive responses
+                 system_name: str = "Document RAG System",
+                 context_limit: int = 16000):  # Increased context limit for more comprehensive context
         self.local_dir = local_dir
         self.embedding_model = embedding_model
         self.collection_name = collection_name
@@ -56,15 +85,20 @@ class Config:
         self.temperature = temperature
         self.max_tokens = max_tokens
         self.system_name = system_name
+        self.context_limit = context_limit
         
+        # Create local directory if it doesn't exist
         os.makedirs(local_dir, exist_ok=True)
+        
         logger.info(f"Initialized configuration: {self.__dict__}")
     
     def to_dict(self) -> Dict[str, Any]:
+        """Convert configuration to dictionary for serialization"""
         return self.__dict__
     
     @classmethod
     def from_file(cls, config_path: str) -> 'Config':
+        """Load configuration from JSON file"""
         try:
             with open(config_path, 'r') as f:
                 config_dict = json.load(f)
@@ -76,6 +110,7 @@ class Config:
             return cls()
     
     def save_to_file(self, config_path: str) -> bool:
+        """Save configuration to JSON file"""
         try:
             with open(config_path, 'w') as f:
                 json.dump(self.to_dict(), f, indent=2)
@@ -85,33 +120,59 @@ class Config:
             logger.error(f"Failed to save configuration to {config_path}: {e}")
             return False
 
-# ----------------- Embedding Engine -----------------
 class EmbeddingEngine:
     """
-    Handles text embeddings using a lightweight model.
+    Handle embeddings with a lightweight model.
+    
+    This class manages the embedding model used to convert text to vector 
+    representations for semantic search.
+    
+    Attributes:
+        model (SentenceTransformer): The loaded embedding model
+        model_name (str): Name of the successfully loaded model
+        vector_size (int): Dimension of the embedding vectors
+        device (str): Device used for inference ('cuda' or 'cpu')
     """
+    
     def __init__(self, model_name="all-MiniLM-L6-v2"):
+        """
+        Initialize the embedding engine with the specified model.
+        
+        Args:
+            model_name (str): Name of the embedding model to load
+            
+        Raises:
+            SystemExit: If no embedding model could be loaded
+        """
+        # Use GPU if available
         self.device = "cuda" if torch.cuda.is_available() else "cpu"
         logger.info(f"Using device for embeddings: {self.device}")
         
+        # Try multiple model options in order of preference
         model_options = [
             model_name,
-            "all-MiniLM-L6-v2",
-            "paraphrase-MiniLM-L3-v2",
-            "all-mpnet-base-v2"
+            "all-MiniLM-L6-v2",      # Good balance of speed and quality
+            "paraphrase-MiniLM-L3-v2", # Faster but less accurate
+            "all-mpnet-base-v2"      # Higher quality but larger model
         ]
+        
         self.model = None
         
+        # Try each model in order until one works
         for model_option in model_options:
             try:
                 logger.info(f"Attempting to load embedding model: {model_option}")
                 self.model = SentenceTransformer(model_option)
+                
+                # Move model to device
                 self.model.to(self.device)
+                
                 logger.info(f"Successfully loaded embedding model: {model_option}")
                 self.model_name = model_option
                 self.vector_size = self.model.get_sentence_embedding_dimension()
                 logger.info(f"Embedding vector size: {self.vector_size}")
                 break
+                
             except Exception as e:
                 logger.warning(f"Failed to load embedding model {model_option}: {str(e)}")
         
@@ -121,8 +182,22 @@ class EmbeddingEngine:
             raise SystemExit(error_msg)
     
     def embed(self, texts: List[str]) -> np.ndarray:
+        """
+        Generate embeddings for a list of texts.
+        
+        Args:
+            texts (List[str]): List of texts to embed
+            
+        Returns:
+            np.ndarray: Array of embeddings
+            
+        Raises:
+            ValueError: If the input is invalid
+            RuntimeError: If embedding fails
+        """
         if not texts:
             raise ValueError("Cannot embed empty list of texts")
+        
         try:
             embeddings = self.model.encode(texts, convert_to_numpy=True)
             return embeddings
@@ -130,13 +205,33 @@ class EmbeddingEngine:
             logger.error(f"Error generating embeddings: {e}")
             raise RuntimeError(f"Failed to generate embeddings: {e}")
 
-# ----------------- Vector Store Manager -----------------
 class VectorStoreManager:
     """
-    Manages Chroma vector store operations.
+    Manage Chroma vector store operations - upload, query, etc.
+    
+    This class provides an interface to the ChromaDB vector database,
+    handling document storage, retrieval, and management.
+    
+    Attributes:
+        config (Config): Configuration parameters
+        client (chromadb.PersistentClient): ChromaDB client
+        collection (chromadb.Collection): The active ChromaDB collection
+        embedding_engine (EmbeddingEngine): Engine for generating embeddings
     """
+    
     def __init__(self, config: Config):
+        """
+        Initialize the vector store manager.
+        
+        Args:
+            config (Config): Configuration parameters
+            
+        Raises:
+            SystemExit: If the vector store cannot be initialized
+        """
         self.config = config
+            
+        # Initialize Chroma client (local persistence)
         logger.info(f"Initializing Chroma at {config.local_dir}")
         try:
             self.client = chromadb.PersistentClient(path=config.local_dir)
@@ -146,15 +241,19 @@ class VectorStoreManager:
             logger.critical(error_msg)
             raise SystemExit(error_msg)
         
+        # Get or create collection
         try:
+            # Initialize embedding model
             logger.info("Loading embedding model...")
             self.embedding_engine = EmbeddingEngine(config.embedding_model)
             logger.info(f"Using embedding model: {self.embedding_engine.model_name}")
             
+            # Create embedding function
             sentence_transformer_ef = embedding_functions.SentenceTransformerEmbeddingFunction(
                 model_name=self.embedding_engine.model_name
             )
             
+            # Try to get existing collection or create a new one
             try:
                 self.collection = self.client.get_collection(
                     name=config.collection_name,
@@ -163,15 +262,18 @@ class VectorStoreManager:
                 logger.info(f"Using existing collection: {config.collection_name}")
             except Exception as e:
                 logger.warning(f"Error getting collection: {e}")
+                # Attempt to get a list of available collections
                 collections = self.client.list_collections()
                 if collections:
                     logger.info(f"Available collections: {[c.name for c in collections]}")
+                    # Use the first available collection if any
                     self.collection = self.client.get_collection(
                         name=collections[0].name,
                         embedding_function=sentence_transformer_ef
                     )
                     logger.info(f"Using collection: {collections[0].name}")
                 else:
+                    # Create new collection if none exist
                     self.collection = self.client.create_collection(
                         name=config.collection_name,
                         embedding_function=sentence_transformer_ef,
@@ -185,42 +287,76 @@ class VectorStoreManager:
             raise SystemExit(error_msg)
     
     def query(self, query_text: str, n_results: int = 5) -> List[Dict]:
+        """
+        Query the vector store with a text query.
+        
+        Args:
+            query_text (str): The query text
+            n_results (int): Number of results to return
+            
+        Returns:
+            List[Dict]: List of results with document text, metadata, and similarity score
+        """
         if not query_text.strip():
             logger.warning("Empty query received")
             return []
+        
         try:
             logger.info(f"Querying vector store with: '{query_text[:50]}...' (top {n_results})")
+            
+            # Query the collection
             search_results = self.collection.query(
                 query_texts=[query_text],
                 n_results=n_results,
                 include=["documents", "metadatas", "distances"]
             )
+            
+            # Format results
             results = []
             if search_results["documents"] and len(search_results["documents"][0]) > 0:
                 for i in range(len(search_results["documents"][0])):
                     results.append({
                         'document': search_results["documents"][0][i],
                         'metadata': search_results["metadatas"][0][i] if search_results["metadatas"] else {},
-                        'score': 1.0 - search_results["distances"][0][i],  # convert distance to similarity
+                        'score': 1.0 - search_results["distances"][0][i],  # Convert distance to similarity
                         'distance': search_results["distances"][0][i]
                     })
+                
                 logger.info(f"Found {len(results)} results for query")
             else:
                 logger.info("No results found for query")
+            
             return results
         except Exception as e:
             logger.error(f"Error querying collection: {e}")
             logger.debug(traceback.format_exc())
             return []
 
-    def add_document(self, document: str, doc_id: str, metadata: Dict[str, Any]) -> bool:
+    def add_document(self, 
+                    document: str, 
+                    doc_id: str, 
+                    metadata: Dict[str, Any]) -> bool:
+        """
+        Add a document to the vector store.
+        
+        Args:
+            document (str): The document text
+            doc_id (str): Unique identifier for the document
+            metadata (Dict[str, Any]): Metadata about the document
+            
+        Returns:
+            bool: True if successful, False otherwise
+        """
         try:
             logger.info(f"Adding document '{doc_id}' to vector store")
+            
+            # Add the document to the collection
             self.collection.add(
                 documents=[document],
                 ids=[doc_id],
                 metadatas=[metadata]
             )
+            
             logger.info(f"Successfully added document '{doc_id}'")
             return True
         except Exception as e:
@@ -228,6 +364,15 @@ class VectorStoreManager:
             return False
 
     def delete_document(self, doc_id: str) -> bool:
+        """
+        Delete a document from the vector store.
+        
+        Args:
+            doc_id (str): ID of the document to delete
+            
+        Returns:
+            bool: True if successful, False otherwise
+        """
         try:
             logger.info(f"Deleting document '{doc_id}' from vector store")
             self.collection.delete(ids=[doc_id])
@@ -238,19 +383,31 @@ class VectorStoreManager:
             return False
 
     def get_statistics(self) -> Dict[str, Any]:
+        """
+        Get statistics about the vector store.
+        
+        Returns:
+            Dict[str, Any]: Statistics about the vector store
+        """
         stats = {
             'collection_name': self.config.collection_name,
             'embedding_model': self.embedding_engine.model_name,
             'embedding_dimensions': self.embedding_engine.vector_size,
             'device': self.embedding_engine.device
         }
+        
         try:
+            # Get collection count
             collection_count = self.collection.count()
             stats['total_documents'] = collection_count
+            
+            # Get unique metadata values
             if collection_count > 0:
                 try:
+                    # Get a sample of document metadata
                     sample_results = self.collection.get(limit=min(collection_count, 100))
                     if sample_results and 'metadatas' in sample_results and sample_results['metadatas']:
+                        # Count unique files if filename exists in metadata
                         filenames = set()
                         for metadata in sample_results['metadatas']:
                             if 'filename' in metadata:
@@ -258,28 +415,57 @@ class VectorStoreManager:
                         stats['unique_files'] = len(filenames)
                 except Exception as e:
                     logger.warning(f"Error getting metadata statistics: {e}")
+            
             logger.info(f"Vector store statistics: {stats}")
         except Exception as e:
             logger.error(f"Error getting statistics: {e}")
             stats['error'] = str(e)
+        
         return stats
 
-# ----------------- RAG System -----------------
 class RAGSystem:
     """
-    Handles the Retrieval-Augmented Generation workflow.
+    Retrieval-Augmented Generation with multiple LLM providers.
+    
+    This class handles the RAG workflow: retrieval of relevant documents,
+    formatting context, and generating responses with different LLM providers.
+    
+    Attributes:
+        vector_store (VectorStoreManager): Manager for vector store operations
+        openai_client (Optional[OpenAI]): OpenAI client
+        gemini_configured (bool): Whether Gemini API is configured
+        config (Config): Configuration parameters
     """
+    
     def __init__(self, vector_store: VectorStoreManager, config: Config):
+        """
+        Initialize the RAG system.
+        
+        Args:
+            vector_store (VectorStoreManager): Vector store manager
+            config (Config): Configuration parameters
+        """
         self.vector_store = vector_store
         self.config = config
         self.openai_client = None
         self.gemini_configured = False
+        
         logger.info("Initialized RAG system")
     
     def setup_openai(self, api_key: str) -> bool:
+        """
+        Set up OpenAI client with API key.
+        
+        Args:
+            api_key (str): OpenAI API key
+            
+        Returns:
+            bool: True if successful, False otherwise
+        """
         if not api_key.strip():
             logger.warning("Empty OpenAI API key provided")
             return False
+            
         try:
             logger.info("Setting up OpenAI client")
             self.openai_client = OpenAI(api_key=api_key)
@@ -300,14 +486,27 @@ class RAGSystem:
             return False
     
     def setup_gemini(self, api_key: str) -> bool:
+        """
+        Set up Gemini with API key.
+        
+        Args:
+            api_key (str): Google AI API key
+            
+        Returns:
+            bool: True if successful, False otherwise
+        """
         if not api_key.strip():
             logger.warning("Empty Gemini API key provided")
             return False
+            
         try:
             logger.info("Setting up Gemini client")
             genai.configure(api_key=api_key)
+            
+            # Test the API key with a simple request
             model = genai.GenerativeModel(self.config.gemini_model)
             response = model.generate_content("Test connection")
+            
             self.gemini_configured = True
             logger.info("Gemini client configured successfully")
             return True
@@ -315,44 +514,91 @@ class RAGSystem:
             logger.error(f"Error configuring Gemini: {e}")
             self.gemini_configured = False
             return False
-
+    
     def format_context(self, documents: List[Dict]) -> str:
+        """
+        Format retrieved documents into context for the LLM.
+        
+        Args:
+            documents (List[Dict]): List of retrieved documents
+            
+        Returns:
+            str: Formatted context for the LLM
+        """
         if not documents:
             logger.warning("No documents provided for context formatting")
             return "No relevant documents found."
+        
         logger.info(f"Formatting {len(documents)} documents for context")
         context_parts = []
+        
         for i, doc in enumerate(documents):
             metadata = doc['metadata']
+            # Extract document metadata in a robust way
             title = metadata.get('title', metadata.get('filename', 'Unknown document'))
-            source = metadata.get('source', metadata.get('path', 'Unknown source'))
-            date = metadata.get('date', metadata.get('created_at', 'Unknown date'))
+            
+            # Format header with just essential metadata for cleaner context
             header = f"Document {i+1} - {title}"
-            if source != 'Unknown source':
-                header += f" (Source: {source})"
-            if date != 'Unknown date':
-                header += f" (Date: {date})"
+            
+            # For readability, limit length of context document (using config value)
             doc_text = doc['document']
-            if len(doc_text) > 8000:
-                doc_text = doc_text[:8000] + "... [Document truncated for context]"
+            if len(doc_text) > (self.config.context_limit // len(documents)):  
+                # Divide context limit among the documents
+                max_length = self.config.context_limit // len(documents)
+                doc_text = doc_text[:max_length] + "... [Document truncated for brevity]"
+                
             context_parts.append(f"{header}:\n{doc_text}\n")
+        
         full_context = "\n".join(context_parts)
         logger.info(f"Created context with {len(full_context)} characters")
+        
         return full_context
-
+    
     def generate_response_openai(self, query: str, context: str) -> str:
+        """
+        Generate a response using OpenAI model with context.
+        
+        Args:
+            query (str): User query
+            context (str): Formatted document context
+            
+        Returns:
+            str: Generated response
+        """
         if not self.openai_client:
             logger.warning("OpenAI API key not configured for response generation")
-            return "Error: OpenAI API key not configured. Please enter an API key."
+            return "Please configure an OpenAI API key to use this feature. Enter your API key in the field and click 'Save API Key'."
         
-        system_prompt = (
-            "You are a knowledgeable assistant that answers questions based solely on the provided context. "
-            "Use clear headings and cite the document numbers where the information is found. "
-            "If the context lacks the needed details, say so and suggest what additional details might help."
-        )
+        # Improved system prompt for better, more comprehensive responses
+        system_prompt = """
+        You are an exceptionally helpful, clear, and friendly AI research assistant. Your goal is to provide comprehensive, well-structured, and insightful answers based on the provided document context.
+
+        Guidelines for your response:
+        
+        1. USE ONLY the information contained in the provided context documents to form your answer. If the context doesn't contain enough information to provide a complete answer, acknowledge this limitation clearly.
+        
+        2. Always provide well-structured, detailed responses between 300-500 words that thoroughly address the user's question.
+        
+        3. Format your response with clear headings, bullet points, or numbered lists when appropriate to enhance readability.
+        
+        4. Cite your sources by referring to the document numbers (e.g., "According to Document 1...") to support your claims.
+        
+        5. Use a friendly, conversational, and supportive tone that makes complex information accessible.
+        
+        6. If different documents offer conflicting information, acknowledge these differences and present both perspectives without bias.
+        
+        7. When appropriate, organize information into logical categories or chronological order to improve clarity.
+        
+        8. Use examples from the documents to illustrate key points when available.
+        
+        9. Conclude with a brief summary of the main points if the answer is complex.
+        
+        10. Remember to stay focused on the user's specific question while providing sufficient context for complete understanding.
+        """
         
         try:
-            logger.info(f"Generating response with OpenAI using model {self.config.openai_model}")
+            logger.info(f"Generating response with OpenAI ({self.config.openai_model})")
+            
             start_time = datetime.now()
             response = self.openai_client.chat.completions.create(
                 model=self.config.openai_model,
@@ -363,70 +609,130 @@ class RAGSystem:
                 temperature=self.config.temperature,
                 max_tokens=self.config.max_tokens,
             )
+            
             generation_time = (datetime.now() - start_time).total_seconds()
             response_text = response.choices[0].message.content
+            
             logger.info(f"Generated response with OpenAI in {generation_time:.2f} seconds")
             return response_text
         except Exception as e:
             error_msg = f"Error generating response with OpenAI: {str(e)}"
             logger.error(error_msg)
-            return f"Error: {error_msg}"
+            return f"I encountered an error while generating your response. Please try again or check your API key. Error details: {str(e)}"
     
     def generate_response_gemini(self, query: str, context: str) -> str:
+        """
+        Generate a response using Gemini with context.
+        
+        Args:
+            query (str): User query
+            context (str): Formatted document context
+            
+        Returns:
+            str: Generated response
+        """
         if not self.gemini_configured:
             logger.warning("Gemini API key not configured for response generation")
-            return "Error: Gemini API key not configured. Please enter an API key."
+            return "Please configure a Google AI API key to use this feature. Enter your API key in the field and click 'Save API Key'."
         
-        prompt = (
-            "You are an insightful assistant who provides detailed, well-organized answers based solely on the provided context. "
-            "Answer the question below by clearly citing document numbers where applicable. "
-            "If there is insufficient context, indicate what further details would be needed.\n\n"
-            f"Context:\n{context}\n\nQuestion: {query}\n\nAnswer:"
-        )
+        # Improved Gemini prompt for more comprehensive and user-friendly responses
+        prompt = f"""
+        You are a knowledgeable and friendly research assistant who excels at providing clear, comprehensive, and well-structured responses. Your goal is to help users understand complex information from documents in an accessible way.
+
+        **Guidelines for Your Response:**
+        
+        - Create a detailed, well-organized response of approximately 300-500 words that thoroughly addresses the user's question.
+        - Use ONLY information from the provided context documents.
+        - Structure your answer with clear paragraphs, and use headings, bullet points, or numbered lists when appropriate.
+        - Maintain a friendly, conversational tone that makes information accessible and engaging.
+        - When citing information, reference specific documents by number (e.g., "As mentioned in Document 2...").
+        - If the context doesn't contain enough information for a complete answer, acknowledge this limitation while providing what you can from the available context.
+        - If documents contain conflicting information, present both perspectives fairly.
+        - Conclude with a brief summary if the topic is complex.
         
+        **Context Documents:**
+        {context}
+        
+        **User's Question:** 
+        {query}
+
+        **Your Response:** 
+        """
+
         try:
-            logger.info(f"Generating response with Gemini using model {self.config.gemini_model}")
+            logger.info(f"Generating response with Gemini ({self.config.gemini_model})")
+            
             start_time = datetime.now()
             model = genai.GenerativeModel(self.config.gemini_model)
+            
             generation_config = {
                 "temperature": self.config.temperature,
                 "max_output_tokens": self.config.max_tokens,
                 "top_p": 0.9,
                 "top_k": 40
             }
-            response = model.generate_content(prompt, generation_config=generation_config)
+            
+            response = model.generate_content(
+                prompt,
+                generation_config=generation_config
+            )
+            
             generation_time = (datetime.now() - start_time).total_seconds()
             response_text = response.text
+            
             logger.info(f"Generated response with Gemini in {generation_time:.2f} seconds")
             return response_text
         except Exception as e:
             error_msg = f"Error generating response with Gemini: {str(e)}"
             logger.error(error_msg)
-            return f"Error: {error_msg}"
+            return f"I encountered an error while generating your response. Please try again or check your API key. Error details: {str(e)}"
     
-    def query_and_generate(self, query: str, n_results: int = 5, model: str = "openai") -> Tuple[str, str]:
+    def query_and_generate(self, 
+                          query: str, 
+                          n_results: int = 5, 
+                          model: str = "openai") -> Tuple[str, str]:
+        """
+        Retrieve relevant documents and generate a response using the specified model.
+        
+        Args:
+            query (str): User query
+            n_results (int): Number of documents to retrieve
+            model (str): Model provider to use ('openai' or 'gemini')
+            
+        Returns:
+            Tuple[str, str]: (Generated response, Search results)
+        """
         if not query.strip():
             logger.warning("Empty query received")
             return "Please enter a question to get a response.", "No search performed."
         
-        logger.info(f"Processing query: '{query[:50]}...' using {model} model")
+        logger.info(f"Processing query: '{query[:50]}...' with {model} model")
+        
+        # Query vector store
         documents = self.vector_store.query(query, n_results=n_results)
         
-        # Format retrieval details (hidden by default in the UI)
+        # Format search results (for logs and hidden UI component)
+        # We'll format this in a way that's more useful for reference but not shown in UI
         formatted_results = []
         for i, res in enumerate(documents):
             metadata = res['metadata']
             title = metadata.get('title', metadata.get('filename', 'Unknown'))
-            preview = res['document'][:300] + '...' if len(res['document']) > 300 else res['document']
-            formatted_results.append(f"**Document {i+1}**\nSource: {title}\nPreview:\n{preview}\n")
-        search_output_text = "\n".join(formatted_results) if formatted_results else "No results found."
+            score = res['score']
+            
+            # Only include a very brief preview for reference
+            preview = res['document'][:100] + '...' if len(res['document']) > 100 else res['document']
+            formatted_results.append(f"Document {i+1}: {title} (Relevance: {score:.2f})")
+        
+        search_output_text = "\n".join(formatted_results) if formatted_results else "No relevant documents found."
         
         if not documents:
             logger.warning("No relevant documents found")
-            return "No relevant documents found to answer your question.", search_output_text
+            return "I couldn't find relevant information in the knowledge base to answer your question. Could you try rephrasing your question or ask about a different topic?", search_output_text
         
+        # Format context
         context = self.format_context(documents)
         
+        # Generate response with the appropriate model
         if model == "openai":
             response = self.generate_response_openai(query, context)
         elif model == "gemini":
@@ -438,145 +744,85 @@ class RAGSystem:
         
         return response, search_output_text
 
-# ----------------- Utility Function -----------------
 def get_db_stats(vector_store: VectorStoreManager) -> str:
+    """
+    Function to get vector store statistics.
+    
+    Args:
+        vector_store (VectorStoreManager): Vector store manager
+        
+    Returns:
+        str: Formatted statistics string
+    """
     try:
         stats = vector_store.get_statistics()
         total_docs = stats.get('total_documents', 0)
-        unique_files = stats.get('unique_files', 'Unknown')
-        model = stats.get('embedding_model', 'Unknown')
-        device = stats.get('device', 'Unknown')
-        stats_text = (
-            f"Total documents: {total_docs}\n"
-            f"Unique files: {unique_files}\n"
-            f"Embedding model: {model}\n"
-            f"Device: {device}"
-        )
+        
+        stats_text = f"Documents in knowledge base: {total_docs}"
         return stats_text
     except Exception as e:
         logger.error(f"Error getting statistics: {e}")
         return "Error getting database statistics"
 
-# ----------------- Main Application -----------------
 def main():
-    # Define configuration file path before usage
+    """Main function to run the RAG application"""
+    # Path for configuration file
     CONFIG_FILE_PATH = "rag_config.json"
-    print(f"Starting Document RAG System v1.0.0")
-    print(f"Log file: {log_file}")
     
-    # Load configuration from file or use defaults
+    # Try to load configuration from file, or use defaults
     if os.path.exists(CONFIG_FILE_PATH):
         config = Config.from_file(CONFIG_FILE_PATH)
     else:
-        config = Config(local_dir="./chroma_db", collection_name="markdown_docs")
+        config = Config(
+            local_dir="./chroma_db",  # Store Chroma files in dedicated directory
+            collection_name="markdown_docs"
+        )
+        # Save default configuration
         config.save_to_file(CONFIG_FILE_PATH)
     
+    print(f"Starting Document Knowledge Assistant v{VERSION}")
+    print(f"Log file: {log_file}")
+    
     try:
+        # Initialize vector store manager with existing collection
         vector_store = VectorStoreManager(config)
-        rag_system = RAGSystem(vector_store, config)
-    except Exception as e:
-        print(f"Error initializing system: {e}")
-        sys.exit(1)
-    
-    # ----------------- Gradio Callback Functions -----------------
-    def save_api_key(model_choice: str, api_key: str):
-        if model_choice == "openai":
-            success = rag_system.setup_openai(api_key)
-            return "OpenAI API key saved and configured successfully." if success else "Error configuring OpenAI API key."
-        elif model_choice == "gemini":
-            success = rag_system.setup_gemini(api_key)
-            return "Gemini API key saved and configured successfully." if success else "Error configuring Gemini API key."
-        else:
-            return "Unknown model choice."
-    
-    def process_query(query: str, model_choice: str, n_results: int, temperature: float, max_tokens: int):
-        # Update configuration parameters based on slider values
-        config.temperature = temperature
-        config.max_tokens = max_tokens
-        response_text, search_details = rag_system.query_and_generate(query, n_results=n_results, model=model_choice)
-        return response_text, search_details
-    
-    # ----------------- Gradio Interface -----------------
-    with gr.Blocks(title=config.system_name) as app:
-        gr.Markdown(f"# {config.system_name} v1.0.0")
-        gr.Markdown("Retrieve answers from your documents with AI-powered retrieval and generation.")
-
-        with gr.Row():
-            with gr.Column(scale=1):
-                with gr.Box():
-                    gr.Markdown("### LLM Configuration")
-                    model_choice = gr.Radio(
-                        choices=["openai", "gemini"], 
-                        value="openai", 
-                        label="Select LLM Provider",
-                        info="Choose between OpenAI and Gemini models."
-                    )
-                    api_key_input = gr.Textbox(
-                        label="API Key",
-                        placeholder="Enter your API key here...",
-                        type="password",
-                        info="Your API key is not stored between sessions."
-                    )
-                    save_key_button = gr.Button("Save API Key", variant="primary")
-                    api_status = gr.Markdown("")
-                    
-                with gr.Box():
-                    gr.Markdown("### Search Settings")
-                    n_results_slider = gr.Slider(
-                        minimum=1, 
-                        maximum=20, 
-                        value=config.default_top_k, 
-                        step=1, 
-                        label="Documents to Retrieve",
-                        info="Number of documents for context."
-                    )
-                    temperature_slider = gr.Slider(
-                        minimum=0.0,
-                        maximum=1.0,
-                        value=config.temperature,
-                        step=0.05,
-                        label="Response Temperature",
-                        info="Lower values yield more factual responses."
-                    )
-                    max_tokens_slider = gr.Slider(
-                        minimum=100,
-                        maximum=4000,
-                        value=config.max_tokens,
-                        step=100,
-                        label="Max Output Tokens",
-                        info="Maximum tokens in generated response."
-                    )
-            
-            with gr.Column(scale=2):
-                with gr.Box():
-                    gr.Markdown("### Ask a Question")
-                    query_input = gr.Textbox(
-                        label="Your Question",
-                        placeholder="Enter your question here..."
-                    )
-                    submit_button = gr.Button("Submit")
-                with gr.Box():
-                    answer_output = gr.Markdown(label="Answer")
-                with gr.Accordion("View Document Retrieval Details (hidden)", open=False):
-                    retrieval_output = gr.Markdown(label="Retrieval Details")
-        
-        # Set up callbacks
-        save_key_button.click(
-            save_api_key,
-            inputs=[model_choice, api_key_input],
-            outputs=api_status
-        )
-        
-        submit_button.click(
-            process_query,
-            inputs=[query_input, model_choice, n_results_slider, temperature_slider, max_tokens_slider],
-            outputs=[answer_output, retrieval_output]
-        )
         
-        with gr.Accordion("View Database Statistics", open=False):
-            db_stats = gr.Markdown(get_db_stats(vector_store))
+        # Initialize RAG system without API keys initially
+        rag_system = RAGSystem(vector_store, config)
         
-    app.launch()
-
-if __name__ == "__main__":
-    main()
+        # Custom CSS for better UI
+        custom_css = """
+        .gradio-container {
+            max-width: 1200px;
+            margin: auto;
+        }
+        .gr-prose h1 {
+            font-size: 2.5rem;
+            margin-bottom: 1rem;
+            color: #1a5276;
+        }
+        .gr-prose h3 {
+            font-size: 1.25rem;
+            font-weight: 600;
+            margin-top: 1rem;
+            margin-bottom: 0.5rem;
+            color: #2874a6;
+        }
+        .container {
+            margin: 0 auto;
+            padding: 2rem;
+        }
+        .gr-box {
+            border-radius: 8px;
+            box-shadow: 0 1px 3px rgba(0,0,0,0.12), 0 1px 2px rgba(0,0,0,0.24);
+            padding: 1rem;
+            margin-bottom: 1rem;
+            background-color: #f9f9f9;
+        }
+        .footer {
+            text-align: center;
+            font-size: 0.8rem;
+            color: #666;
+            margin-top: 2rem;
+        }
+        """