Update app.py

app.py

@@ -1,33 +1,32 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
 
-# 🔑 Tu modelo gratuito y cliente HF
-MODEL_ID = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
+# Modelo gratuito y público
+MODEL_ID = "google/gemma-2b-it"
 client = InferenceClient(model=MODEL_ID)
 
-def responder(mensaje, historial):
+def chat(message, history):
+    history = history or []
+    # Convierte el historial a formato texto
+    prompt = ""
+    for user_msg, bot_msg in history:
+        prompt += f"User: {user_msg}\nAssistant: {bot_msg}\n"
+    prompt += f"User: {message}\nAssistant:"
+
     try:
-        respuesta = ""
-        for salida in client.chat_completion(
-            model=MODEL_ID,
-            messages=[{"role": "user", "content": mensaje}],
-            max_tokens=200,
-            stream=True,
-        ):
-            if salida.choices[0].delta.content:
-                respuesta += salida.choices[0].delta.content
-        historial.append((mensaje, respuesta))
-        return "", historial
+        response = ""
+        for chunk in client.text_generation(prompt, max_new_tokens=128, stream=True):
+            response += chunk.token.text
+        history.append((message, response.strip()))
+        return history, history
     except Exception as e:
-        return f"⚠️ Error interno: {e}", historial
+        return history + [(message, f"⚠️ Error interno: {e}")], history
 
-# 🎨 Interfaz de Gradio
-demo = gr.ChatInterface(
-    responder,
-    title="Chat Gradio con HF (TinyLlama)",
-    description="Modelo gratuito usando TinyLlama desde Hugging Face Inference API",
-    theme="soft",
+chatbot = gr.ChatInterface(
+    fn=chat,
+    title="Chat FastAPI con HF",
+    description="Chat simple usando google/gemma-2b-it en Hugging Face"
 )
 
 if __name__ == "__main__":
-    demo.launch()
+    chatbot.launch()