Update app.py

app.py

@@ -336,6 +336,7 @@ if "pdf_name"    not in st.session_state: st.session_state.pdf_name    = None
 if "pdf_pages"   not in st.session_state: st.session_state.pdf_pages   = 0
 if "pdf_chunks"  not in st.session_state: st.session_state.pdf_chunks  = 0
 if "q_count"     not in st.session_state: st.session_state.q_count     = 0
+if "last_q"      not in st.session_state: st.session_state.last_q      = ""
 
 # ─── MODEL LOADERS ────────────────────────────────────────────────────────────
 @st.cache_resource(show_spinner=False)
@@ -349,9 +350,13 @@ def load_llm():
     model_id  = "TinyLlama/TinyLlama-1.1B-chat-v1.0"
     tokenizer = AutoTokenizer.from_pretrained(model_id)
     model     = AutoModelForCausalLM.from_pretrained(
-        model_id, device_map="auto",
-        torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32
+        model_id,
+        torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
+        low_cpu_mem_usage=True,
+        device_map="cuda" if torch.cuda.is_available() else None
     )
+    if not torch.cuda.is_available():
+        model = model.to("cpu")
     pipe = pipeline(
         "text-generation", model=model, tokenizer=tokenizer,
         max_new_tokens=512, temperature=0.3, do_sample=True,
@@ -421,9 +426,9 @@ with st.sidebar:
     <div class="profile-card">
         <div class="profile-avatar">{avatar}</div>
         <div class="profile-name">SRIRAM SAI</div>
-        <div class="profile-role">// AI &amp; ML ENGINEER</div>
+        <div class="profile-role">AI &amp; ML ENGINEER</div>
         <div class="profile-links">
-            <a class="p-link" href="https://github.com/sriramsai18" target="_blank">💻 GitHub</a>
+            <a class="p-link" href="https://github.com/sriramsai18" target="_blank">💻 Github</a>
             <a class="p-link" href="https://www.linkedin.com/in/sriram-sai-laggisetti/" target="_blank">💼 LinkedIn</a>
         </div>
     </div>
@@ -502,7 +507,7 @@ st.markdown("""
 <div class="app-header">
     <div>
         <div class="app-title">QUERY<span>DOCS</span> AI 📚</div>
-        <div class="app-sub">// INTELLIGENT DOCUMENT Q&amp;A · RAG PIPELINE · TINYLLAMA 1.1B</div>
+        <div class="app-sub">INTELLIGENT DOCUMENT Q&amp;A · RAG PIPELINE </div>
     </div>
 </div>
 """, unsafe_allow_html=True)
@@ -530,7 +535,7 @@ if not st.session_state.vectorstore:
         <div class="wc-title">WELCOME TO QUERYDOCS AI</div>
         <div class="wc-sub">
             Upload any PDF document from the sidebar and start asking questions.<br>
-            Powered by RAG pipeline + TinyLlama 1.1B for accurate, context-aware answers.
+            Powered by RAG pipeline , context-aware answers.
         </div>
         <br>
         <span class="tip-chip">📋 Legal documents</span>
@@ -596,7 +601,7 @@ else:
 
     with col_q:
         question = st.text_input(
-            "", placeholder="// ask a question about your document...",
+            "", placeholder="ask a question about your document...",
             label_visibility="collapsed", key="question_input"
         )
     with col_btn:
@@ -604,8 +609,8 @@ else:
         ask_btn = st.button("▶ ASK", use_container_width=True)
         st.markdown('</div>', unsafe_allow_html=True)
 
-    # Generate answer
-    if (ask_btn or question) and question.strip():
+    # Generate answer — guard against infinite loop
+    if (ask_btn or question) and question.strip() and question.strip() != st.session_state.last_q:
         ts = time.strftime("%H:%M")
         st.session_state.messages.append({
             "role": "user", "content": question.strip(), "time": ts
@@ -644,4 +649,5 @@ else:
             })
 
         typing_slot.empty()
+        st.session_state.last_q = question.strip()
         st.rerun()