Update app.py

app.py

@@ -1,56 +1,48 @@
 import gradio as gr
-import requests
+from transformers import AutoTokenizer, AutoModelForCausalLM
+from peft import PeftModel
+import torch
 
-def generate_response(user_input, chat_history, hf_token):
-    if not hf_token:
-        return chat_history, chat_history, "❌ Please enter your Hugging Face API token first."
+# --- Load tokenizer and model for CPU ---
+tokenizer = AutoTokenizer.from_pretrained("unsloth/Qwen3-1.7B")
 
+base_model = AutoModelForCausalLM.from_pretrained(
+    "unsloth/Qwen3-1.7B",
+    torch_dtype=torch.float32,
+    device_map={"": "cpu"},
+)
+
+model = PeftModel.from_pretrained(base_model, "khazarai/BioGenesis-ToT").to("cpu")
+
+
+# --- Chatbot logic ---
+def generate_response(user_input, chat_history):
     if not user_input.strip():
-        return chat_history, chat_history, ""
-
-    model_id = "khazarai/BioGenesis-ToT"  # Your hosted model
-
-    headers = {
-        "Authorization": f"Bearer {hf_token}"
-    }
-
-    # Combine chat history into a conversation string
-    conversation = ""
-    for msg in chat_history:
-        role = "User" if msg["role"] == "user" else "Assistant"
-        conversation += f"{role}: {msg['content']}\n"
-    conversation += f"User: {user_input}\nAssistant:"
-
-    # Send the request to HF Inference API
-    payload = {
-        "inputs": conversation,
-        "parameters": {
-            "max_new_tokens": 2200,
-            "temperature": 0.6,
-            "top_p": 0.95,
-            "top_k": 20,
-        }
-    }
-
-    response = requests.post(
-        f"https://api-inference.huggingface.co/models/{model_id}",
-        headers=headers,
-        json=payload,
+        return chat_history, chat_history
+
+    chat_history.append({"role": "user", "content": user_input})
+
+    text = tokenizer.apply_chat_template(
+        chat_history,
+        tokenize=False,
+        add_generation_prompt=True,
+        enable_thinking=True,
     )
 
-    if response.status_code != 200:
-        return chat_history, chat_history, f"⚠️ API Error: {response.text}"
+    inputs = tokenizer(text, return_tensors="pt").to("cpu")
 
-    result = response.json()
+    output_tokens = model.generate(
+        **inputs,
+        max_new_tokens=2200,
+        temperature=0.6,
+        top_p=0.95,
+        top_k=20,
+    )
 
-    # Extract model output
-    if isinstance(result, list) and len(result) > 0 and "generated_text" in result[0]:
-        reply = result[0]["generated_text"].split("Assistant:")[-1].strip()
-    else:
-        reply = "🤔 Sorry, I couldn’t generate a response."
+    response = tokenizer.decode(output_tokens[0], skip_special_tokens=True)
+    response = response.split(user_input)[-1].strip()
 
-    chat_history.append({"role": "user", "content": user_input})
-    chat_history.append({"role": "assistant", "content": reply})
+    chat_history.append({"role": "assistant", "content": response})
 
     gr_chat_history = [
         (m["content"], chat_history[i + 1]["content"])
@@ -58,7 +50,7 @@ def generate_response(user_input, chat_history, hf_token):
         if m["role"] == "user"
     ]
 
-    return gr_chat_history, chat_history, ""
+    return gr_chat_history, chat_history
 
 
 # --- UI Design ---
@@ -66,55 +58,45 @@ with gr.Blocks(theme=gr.themes.Soft(primary_hue="emerald", secondary_hue="slate"
     gr.HTML("""
     <div style="text-align: center; margin-bottom: 20px;">
         <h1 style="font-family: 'Inter', sans-serif; font-weight: 800; color: #047857; font-size: 2.2em;">
-            🧬 BioGenesis-ToT Chatbot (Hosted on Hugging Face)
+            🧬 BioGenesis-ToT Chatbot
         </h1>
         <p style="color: #4B5563; font-size: 1.05em; margin-top: -10px;">
-            Talk to your biology-trained LLM — no GPU needed, just your Hugging Face token ⚡
+            Your AI companion for biology, biochemistry, and life sciences.
         </p>
     </div>
     """)
 
     with gr.Row():
         with gr.Column(scale=6):
-            hf_token = gr.Textbox(
-                placeholder="Enter your Hugging Face API Token here...",
-                label="🔑 Hugging Face Token",
-                type="password",
-            )
-
             chatbot = gr.Chatbot(
                 label="BioGenesis Chat",
-                height=550,
-                bubble_full_width=False,
+                height=600,
+                bubble_full_width=True,
                 show_copy_button=True,
                 avatar_images=(
-                    "https://cdn-icons-png.flaticon.com/512/1077/1077012.png",
-                    "https://cdn-icons-png.flaticon.com/512/4140/4140048.png",
+                    "https://cdn-icons-png.flaticon.com/512/1077/1077012.png",  # user icon
+                    "https://cdn-icons-png.flaticon.com/512/4140/4140048.png",  # bot icon
                 ),
             )
-
             user_input = gr.Textbox(
                 placeholder="Ask me about cell biology, molecular structure, or biochemistry...",
                 label="💬 Your question",
                 lines=3,
                 autofocus=True,
             )
-
-            status_box = gr.Textbox(label="Status", interactive=False)
-
             with gr.Row():
                 send_btn = gr.Button("🚀 Send", variant="primary")
                 clear_btn = gr.Button("🧹 Clear Chat")
 
     state = gr.State([])
 
-    send_btn.click(generate_response, [user_input, state, hf_token], [chatbot, state, status_box])
-    user_input.submit(generate_response, [user_input, state, hf_token], [chatbot, state, status_box])
-    clear_btn.click(lambda: ([], [], ""), None, [chatbot, state, status_box])
+    send_btn.click(generate_response, [user_input, state], [chatbot, state])
+    user_input.submit(generate_response, [user_input, state], [chatbot, state])
+    clear_btn.click(lambda: ([], []), None, [chatbot, state])
 
     gr.HTML("""
     <div style="text-align: center; margin-top: 25px; color: #6B7280; font-size: 0.9em;">
-        Powered by <b>Hugging Face Inference API</b> | Built with ❤️ using Gradio
+        Powered by <b>Qwen3-1.7B + BioGenesis-ToT</b> | Built with ❤️ using Gradio
     </div>
     """)