Update app.py

app.py

@@ -1,40 +1,41 @@
 import gradio as gr
-from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
-import torch
 import os
+import requests
 
-# Folder cache cu permisiune
-os.environ["TRANSFORMERS_CACHE"] = "/tmp/.cache"
-os.environ["HF_HOME"] = "/tmp/.cache"
-os.environ["HF_DATASETS_CACHE"] = "/tmp/.cache"
+# Alege un model public HF (gratuit)
+MODEL_ID = "google/flan-t5-small"  # poți folosi orice model HF
 
-# Model mic, gratuit
-MODEL_NAME = "google/flan-t5-small"
+HF_API_TOKEN = os.getenv("HF_API_TOKEN")
+HF_API_URL = f"https://api-inference.huggingface.co/models/{MODEL_ID}"
+HEADERS = {"Authorization": f"Bearer {HF_API_TOKEN}"} if HF_API_TOKEN else {}
 
-# Încarcă model și tokenizer
-tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
-model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_NAME)
-
-device = "cuda" if torch.cuda.is_available() else "cpu"
-model = model.to(device)
-
-# Funcția chatbot
-def chat_fn(question):
+def ask_ai(question):
     if not question.strip():
         return "Te rog scrie o întrebare."
     
-    inputs = tokenizer(question, return_tensors="pt").to(device)
-    outputs = model.generate(**inputs, max_new_tokens=150)
-    answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    return answer
+    payload = {
+        "inputs": question,
+        "parameters": {"max_new_tokens": 150, "return_full_text": False}
+    }
+    
+    try:
+        r = requests.post(HF_API_URL, headers=HEADERS, json=payload, timeout=30)
+        r.raise_for_status()
+    except Exception as e:
+        return f"Eroare la apel HF Inference API: {str(e)}"
+    
+    res = r.json()
+    # modelele HF returnează listă cu generated_text
+    if isinstance(res, list) and "generated_text" in res[0]:
+        return res[0]["generated_text"]
+    return str(res)
 
-# Gradio UI
 iface = gr.Interface(
-    fn=chat_fn,
+    fn=ask_ai,
     inputs=gr.Textbox(lines=2, placeholder="Întreabă ceva..."),
     outputs="text",
-    title="Chatbot simplu",
-    description="Chatbot minimal bazat pe Flan-T5-small (fără permisiuni la cache)."
+    title="Chatbot HF API",
+    description="Chatbot funcțional fără descărcarea modelelor, rulând HF Inference API."
 )
 
 iface.launch(server_name="0.0.0.0", server_port=7860)