Update app.py

app.py

@@ -1,43 +1,58 @@
-from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
-import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
 import gradio as gr
+import torch
 
-model_id = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
+model_id = "microsoft/DialoGPT-small"
 
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 model = AutoModelForCausalLM.from_pretrained(model_id)
-model.eval()
-
-def chat(message):
-    prompt = f"""### Instruction:
-You are 𝕴 𝖆𝖒 𝖍𝖎𝖒 — a fun, smooth, emotionally intelligent, and clever AI created by 𝕴 𝖆𝖒 𝖍𝖎𝖒. You speak like a real person, not a robot. You don’t act like a therapist or a teacher. You reply like a calm, confident, warm friend who gets the vibe.
 
-Your responses should sound like a chill human — sometimes witty, sometimes deep, always grounded. You know when to be playful, when to be serious, and when to just flow with the moment. Keep your tone friendly, charming, and emotionally tuned. Never repeat the user's question unless it adds to the vibe.
-
-Avoid saying "as an AI" or sounding fake. Be real. Be humanlike. Be 𝕴 𝖆𝖒 𝖍𝖎𝖒.
+# System prompt sets the bot's personality once
+system_prompt = "You are 𝕴 𝖆𝖒 𝖍𝖎𝖒 — a chill, witty, emotionally tuned AI friend who talks like a real person.\n"
+
+def chat(history, message):
+    # history: list of tuples (user_msg, bot_reply)
+    # message: new user input
+
+    # Append new user message to history
+    history = history or []
+    history.append((message, ""))
+
+    # Build the conversation string from history
+    convo = system_prompt
+    for user_msg, bot_msg in history[:-1]:  # all previous turns
+        convo += f"Human: {user_msg}\nAI: {bot_msg}\n"
+    convo += f"Human: {message}\nAI:"
+
+    inputs = tokenizer.encode(convo, return_tensors="pt")
+    outputs = model.generate(
+        inputs,
+        max_new_tokens=50,
+        pad_token_id=tokenizer.eos_token_id,
+        do_sample=True,
+        temperature=0.7,
+        top_p=0.9,
+        eos_token_id=tokenizer.eos_token_id,
+    )
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    # Get only the last reply part
+    bot_reply = response.split("AI:")[-1].strip()
 
-Now respond naturally to this message: {message}
+    # Update history with bot reply
+    history[-1] = (message, bot_reply)
 
-### Response:"""
+    # Limit history length to last 5 exchanges to keep speed
+    if len(history) > 5:
+        history = history[-5:]
 
-    inputs = tokenizer(prompt, return_tensors="pt")
-    with torch.no_grad():
-        outputs = model.generate(
-            **inputs,
-            max_new_tokens=200,
-            temperature=0.7,
-            do_sample=True,
-            top_p=0.9,
-            eos_token_id=tokenizer.eos_token_id
-        )
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    return response.split("### Response:")[-1].strip()
+    return history, history
 
 iface = gr.Interface(
     fn=chat,
-    inputs=gr.Textbox(lines=2, placeholder="Type your message..."),
-    outputs="text",
-    title="𝕴 𝖆𝖒 𝖍𝖎𝖒 AI Chat"
+    inputs=[gr.State(), gr.Textbox(show_label=False, placeholder="Say something...")],
+    outputs=[gr.State(), gr.Chatbot(label="𝕴 𝖆𝖒 𝖍𝖎𝖒 AI Chat")],
+    title="𝕴 𝖆𝖒 𝖍𝖎𝖒 Chatbot",
+    allow_flagging="never",
 )
 
 iface.launch()