Create app.py

app.py

@@ -0,0 +1,48 @@
+import gradio as gr
+from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
+
+# 1-line swap =  your own open-weights model here -----------------------------
+MODEL_NAME = "microsoft/DialoGPT-small"   # 356 MB; no HF token required
+# -----------------------------------------------------------------------------
+
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+model     = AutoModelForCausalLM.from_pretrained(MODEL_NAME)
+generator = pipeline(
+    "text-generation",
+    model=model,
+    tokenizer=tokenizer,
+    max_new_tokens=100,
+    do_sample=True,
+    top_p=0.92,
+    temperature=0.7,
+)
+
+def respond(message, history):
+    """
+    message: str   – latest user msg
+    history: list  – [(user, bot), …] accumulated by Gradio
+    returns (bot_reply, history)
+    """
+    # ---- assemble prompt from chat history ----
+    prompt = ""
+    for usr, bot in history:
+        prompt += f"User: {usr}\nBot: {bot}\n"
+    prompt += f"User: {message}\nBot:"
+
+    # ---- generate ----
+    output = generator(prompt, num_return_sequences=1)[0]["generated_text"]
+    reply  = output[len(prompt):].split("User:")[0].strip() or "…"
+
+    # ---- update history ----
+    history.append((message, reply))
+    return reply, history
+
+demo = gr.ChatInterface(
+    fn=respond,
+    title="🤖 Simple DialoGPT Chatbot",
+    description="100 % local model · no API key needed · swap MODEL_NAME to try other open-source LLMs.",
+    theme="soft",
+)
+
+if __name__ == "__main__":
+    demo.launch()