Create app.py

app.py

@@ -0,0 +1,52 @@
+import os
+import gradio as gr
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+# Use your Hugging Face token from the Secrets
+HF_TOKEN = os.environ.get("HF_TOKEN")
+
+# Model (lightweight, public, works on CPU)
+MODEL_NAME = "SandLogicTechnologies/Gemma-3-270m-GGUF"
+
+# Load tokenizer and model
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, use_auth_token=HF_TOKEN)
+model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, use_auth_token=HF_TOKEN).to("cpu")
+
+history = []
+
+def chat_with_airi(user_msg):
+    global history
+    # build simple prompt
+    prompt = ""
+    for u, a in history[-5:]:
+        prompt += f"User: {u}\nAiri: {a}\n"
+    prompt += f"User: {user_msg}\nAiri:"
+
+    inputs = tokenizer(prompt, return_tensors="pt").to("cpu")
+    with torch.no_grad():
+        output = model.generate(
+            **inputs,
+            max_new_tokens=60,
+            do_sample=True,
+            top_p=0.9,
+            temperature=0.8,
+            pad_token_id=tokenizer.eos_token_id
+        )
+
+    reply = tokenizer.decode(output[0], skip_special_tokens=True)
+    reply = reply.split("Airi:", 1)[-1].strip()
+
+    history.append([user_msg, reply])
+    return history, ""
+
+# Gradio interface
+with gr.Blocks() as demo:
+    gr.HTML("<h2 style='text-align:center'>Airi — Mini Chat AI</h2>")
+    gr.HTML("<p style='text-align:center;color:#666;'>Small, Fast & Public Model</p>")
+
+    chat = gr.Chatbot()
+    msg = gr.Textbox(label="Talk to Airi…", placeholder="Write here…")
+    msg.submit(chat_with_airi, msg, [chat, msg])
+
+demo.launch()