Update app.py

app.py

@@ -1,7 +1,6 @@
-# Updated app.py with torch import and error handling
 import gradio as gr
 import os
-import torch  # Missing import added here
+import torch
 from langchain_community.document_loaders import PyPDFLoader
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain_community.embeddings import HuggingFaceEmbeddings
@@ -14,78 +13,75 @@ EMBEDDING_MODEL = "sentence-transformers/all-MiniLM-L6-v2"
 MODEL_NAME = "microsoft/phi-2"
 
 def initialize_system():
-    try:
-        # Verify documents
-        if not os.path.exists(DOCS_DIR):
-            raise FileNotFoundError(f"Missing {DOCS_DIR} folder")
-            
-        pdf_files = [os.path.join(DOCS_DIR, f) 
-                    for f in os.listdir(DOCS_DIR) 
-                    if f.endswith(".pdf")]
-        if not pdf_files:
-            raise ValueError(f"No PDFs found in {DOCS_DIR}")
-        
-        # Process documents
-        text_splitter = RecursiveCharacterTextSplitter(
-            chunk_size=800,
-            chunk_overlap=100
-        )
-        
-        texts = []
-        for pdf in pdf_files:
-            loader = PyPDFLoader(pdf)
-            pages = loader.load_and_split(text_splitter)
-            texts.extend(pages)
-        
-        # Create embeddings
-        embeddings = HuggingFaceEmbeddings(
-            model_name=EMBEDDING_MODEL,
-            model_kwargs={'device': 'cpu'},
-            encode_kwargs={'normalize_embeddings': False}
-        )
-        
-        # Create vector store
-        vector_store = FAISS.from_documents(texts, embeddings)
-        
-        # Load Phi-2 model
-        tokenizer = AutoTokenizer.from_pretrained(
-            MODEL_NAME,
-            trust_remote_code=True,
-            padding_side="left"
-        )
-        
-        model = AutoModelForCausalLM.from_pretrained(
-            MODEL_NAME,
-            trust_remote_code=True,
-            device_map="auto",
-            load_in_4bit=True,
-            torch_dtype=torch.float16
-        )
-        
-        return vector_store, model, tokenizer
-        
-    except Exception as e:
-        raise RuntimeError(f"Initialization failed: {str(e)}")
+    # Document verification
+    if not os.path.exists(DOCS_DIR):
+        raise FileNotFoundError(f"Missing {DOCS_DIR} folder")
+    
+    pdf_files = [os.path.join(DOCS_DIR, f) 
+                for f in os.listdir(DOCS_DIR) 
+                if f.endswith(".pdf")]
+    
+    if not pdf_files:
+        raise ValueError(f"No PDFs found in {DOCS_DIR}")
+
+    # Document processing
+    text_splitter = RecursiveCharacterTextSplitter(
+        chunk_size=800,
+        chunk_overlap=100
+    )
+    
+    texts = []
+    for pdf in pdf_files:
+        loader = PyPDFLoader(pdf)
+        pages = loader.load_and_split(text_splitter)
+        texts.extend(pages)
+
+    # Create embeddings
+    embeddings = HuggingFaceEmbeddings(
+        model_name=EMBEDDING_MODEL,
+        model_kwargs={'device': 'cpu'}
+    )
+    
+    # Vector store
+    vector_store = FAISS.from_documents(texts, embeddings)
+
+    # Model loading
+    tokenizer = AutoTokenizer.from_pretrained(
+        MODEL_NAME,
+        trust_remote_code=True,
+        padding_side="left"
+    )
+    
+    model = AutoModelForCausalLM.from_pretrained(
+        MODEL_NAME,
+        trust_remote_code=True,
+        device_map="auto",
+        load_in_4bit=True,
+        torch_dtype=torch.float16
+    )
+
+    return vector_store, model, tokenizer
 
 try:
     vector_store, model, tokenizer = initialize_system()
-    print("✅ System initialized successfully")
+    print("System initialized successfully ✅")
 except Exception as e:
-    print(f"❌ Initialization error: {str(e)}")
+    print(f"Initialization failed ❌: {str(e)}")
     raise
 
 def generate_response(query):
     try:
-        # Retrieve context
+        # Context retrieval
         docs = vector_store.similarity_search(query, k=2)
         context = "\n".join([d.page_content for d in docs])
         
         # Phi-2 optimized prompt
         prompt = f"""<|system|>
-        You are a customer service assistant. Answer ONLY using the context below.
-        Keep responses under 3 sentences. If unsure, say "I'll check with the team".
-        
-        Context: {context}</s>
+        You are a customer service bot. Answer only using:
+        {context}
+        - Max 3 sentences
+        - If unsure: "I'll check with the team"
+        </s>
         <|user|>
         {query}</s>
         <|assistant|>"""
@@ -95,28 +91,27 @@ def generate_response(query):
             **inputs,
             max_new_tokens=200,
             temperature=0.1,
-            do_sample=True,
             pad_token_id=tokenizer.eos_token_id
         )
         
         response = tokenizer.decode(outputs[0], skip_special_tokens=True)
         return response.split("<|assistant|>")[-1].strip()
-        
+    
     except Exception as e:
-        return "I'm having trouble answering that. Please try again later."
+        return "Please try again later."
 
 # Gradio interface
 with gr.Blocks(theme=gr.themes.Soft()) as demo:
     gr.Markdown("# Customer Support Chatbot")
-    chatbot = gr.Chatbot(height=400)
-    msg = gr.Textbox(label="Your question", placeholder="Type here...")
-    clear = gr.Button("Clear History")
+    chatbot = gr.Chatbot()
+    msg = gr.Textbox(label="Ask about our services")
+    clear = gr.ClearButton([msg, chatbot])
     
     def respond(message, history):
         response = generate_response(message)
-        return response
+        history.append((message, response))
+        return "", history
     
-    msg.submit(respond, [msg, chatbot], chatbot)
-    clear.click(lambda: None, None, chatbot, queue=False)
+    msg.submit(respond, [msg, chatbot], [msg, chatbot])
 
-demo.launch(server_port=7860)
+demo.launch()