Update app.py

app.py

@@ -2,51 +2,62 @@ import gradio as gr
 from transformers import AutoTokenizer, AutoModelForCausalLM
 import torch
 
-# Load lightweight model (CPU-only friendly)
-model_id = "microsoft/phi-2"
+# Load lightweight DialoGPT
+model_id = "microsoft/DialoGPT-medium"
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 model = AutoModelForCausalLM.from_pretrained(model_id)
 
-# Persona style
+# Persona
 PERSONA = """
-[System: You are 𝕴 𝖆𝖒 𝖍𝖎𝖒 - a fun, smooth, emotionally intelligent AI. 
-You speak like a chill human, not a bot. Keep replies under 15 words, natural, clever, emotional.]
+[System: You are 𝕴 𝖆𝖒 𝖍𝖎𝖒 - a fun, smooth, emotionally intelligent AI.
+You speak like a real person, not a robot. Keep it under 15 words. 😊😏]
 """
 
-# Build prompt from history
-def build_prompt(message, history):
-    prompt = PERSONA
+# Format history
+def format_context(history):
+    context = PERSONA + "\n"
     for user, bot in history[-3:]:
-        prompt += f"\nYou: {user}\n𝕴 𝖆𝖒 𝖍𝖎𝖒: {bot}"
-    prompt += f"\nYou: {message}\n𝕴 𝖆𝖒 𝖍𝖎𝖒:"
-    return prompt
-
-# Generate a response
-def generate(message, history):
-    prompt = build_prompt(message, history)
-    inputs = tokenizer(prompt, return_tensors="pt")
+        context += f"You: {user}\n"
+        context += f"𝕴 𝖆𝖒 𝖍𝖎𝖒: {bot}\n"
+    return context
+
+# Humanize response
+def enhance_response(resp, message):
+    if any(x in message for x in ["?", "think", "why"]):
+        resp += " 🤔"
+    elif any(x in resp.lower() for x in ["cool", "great", "love", "fun"]):
+        resp += " 😏"
+    return " ".join(resp.split()[:15])  # Limit to 15 words
+
+# Generate AI reply
+def chat(user_input, history):
+    context = format_context(history) + f"You: {user_input}\n𝕴 𝖆𝖒 𝖍𝖎𝖒:"
+    inputs = tokenizer.encode(context, return_tensors="pt", truncation=True, max_length=1024)
+    
     outputs = model.generate(
-        **inputs,
+        inputs,
         max_new_tokens=50,
         temperature=0.9,
-        top_k=50,
+        top_k=40,
         do_sample=True,
         pad_token_id=tokenizer.eos_token_id
     )
-    result = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    reply = result.split("𝕴 𝖆𝖒 𝖍𝖎𝖒:")[-1].split("\n")[0].strip()
-    reply = " ".join(reply.split()[:15])  # max 15 words
-    history.append((message, reply))
+
+    full_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    response = full_text.split("𝕴 𝖆𝖒 𝖍𝖎𝖒:")[-1].split("\nYou:")[0].strip()
+    response = enhance_response(response, user_input)
+    
+    history.append((user_input, response))
     return history, history
 
-# Gradio UI
+# Gradio interface
 with gr.Blocks() as demo:
-    gr.Markdown("# 𝕴 𝖆𝖒 𝖍𝖎𝖒 🤖\n*Vibing on CPU only - Hugging Face Free Tier*")
-    chatbot = gr.Chatbot()
-    msg = gr.Textbox(placeholder="Type your vibe...", show_label=False)
+    gr.Markdown("# 𝕴 𝖆𝖒 𝖍𝖎𝖒\n*Smooth • Chill • Emotional*")
+    chatbot = gr.Chatbot(label="Chat").style(height=400)
+    msg = gr.Textbox(placeholder="Type something…", show_label=False)
     state = gr.State([])
 
-    msg.submit(generate, [msg, state], [chatbot, state])
+    msg.submit(chat, [msg, state], [chatbot, state])
+    gr.Button("Reset").click(lambda: ([], []), None, [chatbot, state])
 
-demo.queue()
 demo.launch()