Update app.py

app.py

@@ -1,22 +1,67 @@
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import AutoModelForCausalLM, AutoTokenizer
 
+# ID del modello Hugging Face (puoi cambiarlo con un altro compatibile)
 MODEL_ID = "INTERX/Qwen2.5-GenX-7B"
 
+# Caricamento tokenizer e modello
 tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
-model = AutoModelForCausalLM.from_pretrained(MODEL_ID, device_map="auto", torch_dtype="auto", trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID,
+    device_map="auto",
+    torch_dtype="auto",
+    trust_remote_code=True
+)
 
-def chat(user_input):
-    messages = [{"role": "user", "content": user_input}]
-    inputs = tokenizer.apply_chat_template(
-        messages,
+# Funzione di chat con storico
+def chat(user_input, chat_history):
+    if chat_history is None:
+        chat_history = []
+
+    # Aggiunge il nuovo input allo storico
+    chat_history.append({"role": "user", "content": user_input})
+
+    # Prepara l'input per il modello
+    tokenized_input = tokenizer.apply_chat_template(
+        chat_history,
         tokenize=True,
         add_generation_prompt=True,
-        return_tensors="pt"
+        return_tensors='pt'
     ).to(model.device)
 
-    output = model.generate(inputs, max_new_tokens=512)
+    # Genera la risposta
+    output = model.generate(tokenized_input, max_new_tokens=512, do_sample=True, temperature=0.7)
     response = tokenizer.decode(output[0], skip_special_tokens=True)
-    return response
 
-gr.Interface(fn=chat, inputs="text", outputs="text", title="Chat con Qwen2.5-GenX-7B").launch()
+    # Pulisce eventuali ripetizioni del prompt
+    response = response.split("user:")[-1].split("assistant:")[-1].strip()
+
+    # Aggiunge la risposta allo storico
+    chat_history.append({"role": "assistant", "content": response})
+
+    # Prepara lo storico per il componente gr.Chatbot
+    gr_history = [[entry["content"], None] if entry["role"] == "user" else [None, entry["content"]] for entry in chat_history]
+    merged_history = []
+    for i in range(0, len(gr_history), 2):
+        user_msg = gr_history[i][0] if i < len(gr_history) else ""
+        ai_msg = gr_history[i+1][1] if i+1 < len(gr_history) else ""
+        merged_history.append([user_msg, ai_msg])
+    
+    return merged_history, chat_history
+
+# UI con Gradio
+with gr.Blocks(title="Chat Qwen2.5-GenX-7B") as demo:
+    gr.Markdown("# 💬 Chat con Qwen2.5-GenX-7B")
+    chatbot = gr.Chatbot(label="Conversazione")
+    state = gr.State([])  # storico della conversazione
+    with gr.Row():
+        txt = gr.Textbox(show_label=False, placeholder="Scrivi un messaggio...").style(container=False)
+        send_btn = gr.Button("Invia")
+        clear_btn = gr.Button("❌ Reset")
+
+    # Azioni
+    send_btn.click(chat, [txt, state], [chatbot, state])
+    txt.submit(chat, [txt, state], [chatbot, state])
+    clear_btn.click(lambda: ([], []), None, [chatbot, state])
+
+demo.launch()