Update app.py

app.py

@@ -1,45 +1,48 @@
 import gradio as gr
-import requests
-
-# Public model URL - no API key needed
-HF_API_URL = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.2"
-
-def query_hf_model(messages):
-    # Convert Gradio format (list of {"role":..., "content":...}) to prompt text
-    prompt = "\n".join(
-        f"{msg['role'].capitalize()}: {msg['content']}" for msg in messages if msg["role"] in {"user", "assistant"}
-    ) + "\nAssistant:"
-
-    payload = {
-        "inputs": prompt,
-        "parameters": {
-            "max_new_tokens": 128,
-            "temperature": 0.7,
-            "return_full_text": True
-        }
-    }
-
-    response = requests.post(HF_API_URL, json=payload)
-    if response.status_code == 200:
-        result = response.json()[0]["generated_text"]
-        reply = result[len(prompt):].strip()
-        return reply
-    else:
-        return f"⚠️ Error {response.status_code}: {response.reason}"
-
-def respond(user_input, chat_history):
-    # Add user message to history
-    chat_history.append({"role": "user", "content": user_input})
-    reply = query_hf_model(chat_history)
-    chat_history.append({"role": "assistant", "content": reply})
-    return "", chat_history
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
+
+# Load model directly from Hugging Face Hub
+model_name = "sshleifer/tiny-gpt2"
+
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(model_name)
+model.eval()
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model.to(device)
+
+def generate_reply(messages):
+    # Concatenate messages into a simple prompt
+    prompt = "\n".join([m["content"] for m in messages])
+    
+    inputs = tokenizer(prompt, return_tensors="pt").to(device)
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=100,
+        do_sample=True,
+        top_p=0.95,
+        temperature=0.7,
+        pad_token_id=tokenizer.eos_token_id
+    )
+
+    output_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    # Extract the newly generated part
+    reply = output_text[len(prompt):].strip().split("\n")[0]
+    return messages + [{"role": "assistant", "content": reply}]
 
 with gr.Blocks() as demo:
-    gr.Markdown("### 🤖 Mistral Chatbot — No Auth Required")
+    gr.Markdown("### 🤖 Tiny GPT-2 Chatbot (No API Key, Hugging Face Model)")
 
-    chatbot = gr.Chatbot(label="Chat with AI", type="messages", avatar_images=("👤", "🤖"))
-    msg = gr.Textbox(label="Your message", placeholder="Ask me anything...", scale=1)
+    chatbot = gr.Chatbot(label="Chat", type="messages")
+    user_input = gr.Textbox(label="Type your message here")
 
-    msg.submit(fn=respond, inputs=[msg, chatbot], outputs=[msg, chatbot])
+    def on_submit(message, chat_history):
+        chat_history.append({"role": "user", "content": message})
+        return "", chat_history
 
-demo.launch(share=True)
+    user_input.submit(on_submit, [user_input, chatbot], [user_input, chatbot]).then(
+        generate_reply, chatbot, chatbot
+    )
+
+demo.launch()