Update app.py

app.py

@@ -2,34 +2,76 @@ import streamlit as st
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 from threading import Thread
 import torch
+from pypdf import PdfReader
+import io
 
-# 1. Page Configuration
-st.set_page_config(page_title="Qwen Chat", page_icon="🧠", layout="centered")
+# 1. Page Configuration & Aesthetic Injection
+st.set_page_config(page_title="Claude", page_icon="☁️", layout="wide")
 
-# Custom CSS for a cleaner "Claude-like" feel
+# CSS to override Streamlit's look to match Claude's 2026 UI
 st.markdown("""
     <style>
-        [data-testid="stSidebar"] {display: none;}
-        .stChatMessage { border-radius: 10px; margin-bottom: 5px; }
-        .stChatInputContainer { padding-bottom: 20px; }
+        /* Claude's specific background and fonts */
+        @import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600&display=swap');
+        
+        .stApp {
+            background-color: #ffffff;
+            font-family: 'Inter', sans-serif;
+        }
+
+        /* Sidebar styling: Off-white with a very thin border */
+        [data-testid="stSidebar"] {
+            background-color: #f9f9f8 !important;
+            border-right: 1px solid #e5e5e5;
+            width: 300px !important;
+        }
+
+        /* The Main Chat Container (Centered) */
+        .block-container {
+            max-width: 850px;
+            padding-top: 3rem;
+        }
+
+        /* Message Bubbles: Claude uses a clean, borderless look */
+        .stChatMessage {
+            border: none !important;
+            padding: 1.5rem 0 !important;
+            background-color: transparent !important;
+        }
+
+        /* Assistant Message (Bot) specific styling */
+        [data-testid="chatAvatarAssistant"] {
+            background-color: #d97757 !important; /* Claude's signature orange */
+            border-radius: 6px !important;
+        }
+
+        /* Floating, Centered Chat Input */
+        [data-testid="stBottom"] {
+            background-color: white !important;
+            border-top: none !important;
+        }
+        
+        .stChatInputContainer {
+            border: 1px solid #d1d1d1 !important;
+            border-radius: 14px !important;
+            box-shadow: 0 4px 24px rgba(0,0,0,0.06) !important;
+            max-width: 800px !important;
+            margin: 0 auto 20px auto !important;
+        }
+
+        /* Hide default Streamlit elements */
+        header {visibility: hidden;}
+        footer {visibility: hidden;}
     </style>
 """, unsafe_allow_html=True)
 
-st.title("Qwen 2.5 3B Chat 🚀")
-st.caption("A balanced, high-performance model for local CPU/GPU inference.")
-
-# 2. Optimized Model Loading
+# 2. Optimized 2026 Local Model (Qwen 2.5 3B)
 @st.cache_resourced
-def load_model():
-    # '3B' is the most feasible mid-point for modern laptops/PCs
+def load_llm():
     model_id = "Qwen/Qwen2.5-3B-Instruct"
     tokenizer = AutoTokenizer.from_pretrained(model_id)
-    
-    # Auto-detect device (Use GPU if available, else CPU)
     device = "cuda" if torch.cuda.is_available() else "cpu"
-    
-    # Use float16 for GPU or bfloat16 for modern CPUs to save memory
-    dtype = torch.float16 if device == "cuda" else torch.bfloat16
+    dtype = torch.bfloat16 
     
     model = AutoModelForCausalLM.from_pretrained(
         model_id, 
@@ -38,63 +80,69 @@ def load_model():
     )
     return model, tokenizer, device
 
-model, tokenizer, device = load_model()
+model, tokenizer, device = load_llm()
 
-# 3. Session State for Chat History
+# 3. Sidebar: "Projects" & Document Upload
+with st.sidebar:
+    st.markdown("<div style='font-size:1.1rem; font-weight:600; color:#1a1a1a; margin-bottom:1rem;'>Claude 3.5 Sonnet</div>", unsafe_allow_html=True)
+    if st.button("+ Start New Chat", use_container_width=True):
+        st.session_state.messages = []
+        st.session_state.pdf_text = ""
+        st.rerun()
+    
+    st.divider()
+    st.caption("UPLOAD DOCUMENTS")
+    uploaded_file = st.file_uploader("Drop a PDF here to analyze", type="pdf")
+    
+    if uploaded_file:
+        reader = PdfReader(uploaded_file)
+        text = "".join([page.extract_text() for page in reader.pages])
+        st.session_state.pdf_text = text
+        st.success(f"Attached: {uploaded_file.name}")
+
+    st.divider()
+    st.caption("RECENT CHATS")
+    st.markdown("📝 **LOS and LMS Training**")
+
+# 4. Main Chat Interface Logic
 if "messages" not in st.session_state:
     st.session_state.messages = []
+if "pdf_text" not in st.session_state:
+    st.session_state.pdf_text = ""
+
+# Empty State: Landing screen
+if not st.session_state.messages:
+    st.markdown("<div style='height: 12vh;'></div>", unsafe_allow_html=True)
+    st.markdown("<h1 style='text-align: center; font-weight: 500; color: #1a1a1a;'>How can I help you today?</h1>", unsafe_allow_html=True)
 
-# Display Chat History
-for message in st.session_state.messages:
-    with st.chat_message(message["role"]):
-        st.markdown(message["content"])
+# Render Chat History
+for msg in st.session_state.messages:
+    with st.chat_message(msg["role"]):
+        st.markdown(msg["content"])
 
-# 4. Chat Input & Streaming Logic
-if prompt := st.chat_input("How can I help you today?"):
-    # Add user message to history
+# 5. Chat Input & Generation
+if prompt := st.chat_input("Ask anything..."):
     st.session_state.messages.append({"role": "user", "content": prompt})
-    
     with st.chat_message("user"):
         st.markdown(prompt)
 
     with st.chat_message("assistant"):
-        # Setup Streamer
         streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
         
-        # Format conversation using the model's chat template
-        # Limit history to last 5 turns to prevent CPU slowdown
-        context_messages = st.session_state.messages[-10:] 
-        full_prompt = [{"role": "system", "content": "You are Qwen, a helpful and concise AI assistant."}] + context_messages
+        # Build context (System Prompt + PDF Content + History)
+        sys_msg = "You are a world-class AI assistant. Be helpful and professional."
+        if st.session_state.pdf_text:
+            sys_msg += f"\n\nContext from uploaded document:\n{st.session_state.pdf_text[:5000]}" # Limit to first 5k chars for speed
+            
+        messages = [{"role": "system", "content": sys_msg}] + st.session_state.messages[-8:]
         
-        model_inputs = tokenizer.apply_chat_template(
-            full_prompt, 
-            tokenize=True, 
-            add_generation_prompt=True, 
-            return_tensors="pt"
-        ).to(device)
-
-        # Generation Arguments
-        generation_kwargs = dict(
-            input_ids=model_inputs,
-            streamer=streamer,
-            max_new_tokens=1024,
-            do_sample=True,
-            temperature=0.7,
-            top_p=0.8,
-            repetition_penalty=1.1,
-            pad_token_id=tokenizer.eos_token_id
-        )
-
-        # Start thread
-        thread = Thread(target=model.generate, kwargs=generation_kwargs)
-        thread.start()
+        inputs = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt").to(device)
 
-        # Stream the output
-        response_container = st.empty()
-        full_response = ""
+        gen_kwargs = dict(input_ids=inputs, streamer=streamer, max_new_tokens=1024, do_sample=True, temperature=0.7, top_p=0.9, pad_token_id=tokenizer.eos_token_id)
+        
+        thread = Thread(target=model.generate, kwargs=gen_kwargs)
+        thread.start()
         
-        # Use st.write_stream for a native feel
         full_response = st.write_stream(streamer)
             
-    # Save assistant response to history
     st.session_state.messages.append({"role": "assistant", "content": full_response})