another update

app.py

@@ -5,12 +5,14 @@ import faiss
 from sentence_transformers import SentenceTransformer
 from bs4 import BeautifulSoup
 import gradio as gr
-import os
 
 # Configure Gemini API key
 GOOGLE_API_KEY = 'AIzaSyA0yLvySmj8xjMd0sedSgklg1fj0wBDyyw'  # Replace with your API key
 genai.configure(api_key=GOOGLE_API_KEY)
 
+# Initialize conversation history
+conversation_history = []
+
 # Fetch lecture notes and model architectures
 def fetch_lecture_notes():
     lecture_urls = [
@@ -61,25 +63,22 @@ def initialize_faiss_index(embeddings):
     return index
 
 # Handle natural language queries
-conversation_history = []
-
-# Global variables
-lecture_notes = fetch_lecture_notes()
-model_architectures = fetch_model_architectures()
-all_texts = lecture_notes + [model_architectures]
-embedding_model = SentenceTransformer('paraphrase-MiniLM-L6-v2')
-embeddings = create_embeddings(all_texts, embedding_model)
-faiss_index = initialize_faiss_index(np.array(embeddings))
-
 def handle_query(query, faiss_index, embeddings_texts, model):
+    global conversation_history
+
     query_embedding = model.encode([query]).astype('float32')
-    _, indices = faiss_index.search(query_embedding, 3)
+
+    # Search FAISS index
+    _, indices = faiss_index.search(query_embedding, 3)  # Retrieve top 3 results
     relevant_texts = [embeddings_texts[idx] for idx in indices[0]]
+
+    # Combine relevant texts and truncate if necessary
     combined_text = "\n".join([text for text, _ in relevant_texts])
-    max_length = 500
+    max_length = 500  # Adjust as necessary
     if len(combined_text) > max_length:
         combined_text = combined_text[:max_length] + "..."
 
+    # Generate a response using Gemini
     try:
         response = genai.generate_text(
             model="models/text-bison-001",
@@ -88,21 +87,74 @@ def handle_query(query, faiss_index, embeddings_texts, model):
         )
         generated_text = response.result if response else "No response generated."
     except Exception as e:
-        generated_text = f"An error occurred while generating the response: {str(e)}"
+        print(f"Error generating text: {e}")
+        generated_text = "An error occurred while generating the response."
+
+    # Update conversation history
+    conversation_history.append(f"User: {query}")
+    conversation_history.append(f"System: {generated_text}")
 
+    # Extract sources
     sources = [url for _, url in relevant_texts]
+
     return generated_text, sources
 
+def generate_concise_response(prompt, context):
+    try:
+        response = genai.generate_text(
+            model="models/text-bison-001",
+            prompt=f"{prompt}\n\nContext: {context}\n\nAnswer:",
+            max_output_tokens=200
+        )
+        return response.result if response else "No response generated."
+    except Exception as e:
+        print(f"Error generating concise response: {e}")
+        return "An error occurred while generating the concise response."
+
+# Main function to execute the pipeline
 def chatbot(message, history):
+    lecture_notes = fetch_lecture_notes()
+    model_architectures = fetch_model_architectures()
+
+    all_texts = lecture_notes + [model_architectures]
+
+    # Load the SentenceTransformers model
+    embedding_model = SentenceTransformer('paraphrase-MiniLM-L6-v2')
+
+    embeddings = create_embeddings(all_texts, embedding_model)
+
+    # Initialize FAISS index
+    faiss_index = initialize_faiss_index(np.array(embeddings))
+
     response, sources = handle_query(message, faiss_index, all_texts, embedding_model)
+    print("Query:", message)
+    print("Response:", response)
+
+    # Format the response with conversation history
+    formatted_response = "Conversation History:\n\n"
+    for entry in conversation_history:
+        formatted_response += entry + "\n"
     
-    total_text = response if response else "No response generated."
+    formatted_response += "\nCurrent Response:\n" + response
+
     if sources:
-        relevant_source = "\n".join(sources)
-        total_text += f"\n\nSources:\n{relevant_source}"
+        print("Sources:", sources)
+        formatted_response += "\n\nSources:\n" + "\n".join(sources)
+    else:
+        print("Sources: None of the provided sources were used.")
+
+    # Generate a concise and relevant summary using Gemini
+    prompt = "Summarize the user queries so far"
+    user_queries_summary = " ".join([entry for entry in conversation_history if entry.startswith("User: ")])
+    concise_response = generate_concise_response(prompt, user_queries_summary)
+    print("Concise Response:")
+    print(concise_response)
+    
+    formatted_response += "\n\nConcise Summary:\n" + concise_response
+
+    print("----")
     
-    history.append((message, total_text))
-    return history
+    return formatted_response
 
 iface = gr.ChatInterface(
     chatbot,
@@ -113,12 +165,12 @@ iface = gr.ChatInterface(
         "Explain the transformer architecture.",
         "Tell me about datasets used to train LLMs.",
         "How are LLM training datasets cleaned and preprocessed?",
+        "Summarize the user queries so far"
     ],
     retry_btn="Regenerate",
     undo_btn="Undo",
     clear_btn="Clear",
-    cache_examples=False,  # Disable example caching to avoid file-related errors
 )
 
 if __name__ == "__main__":
-    iface.launch(server_name="0.0.0.0", server_port=7860)
+    iface.launch(server_name="0.0.0.0", server_port=7860)