Update app.py

app.py

@@ -18,15 +18,14 @@ chunks = pickle.load(open("chunks.pkl", "rb"))
 metadata = pickle.load(open("metadata.pkl", "rb"))
 
 # -----------------------------
-# Load HF‑hosted small LLM
+# Load free HF small LLM
 # -----------------------------
-model_name = "NousResearch/Nous-Hermes-1.0-GPTQ"
+# Using distilgpt2 as it doesn't need a token
+model_name = "distilgpt2"
 
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
-    device_map="auto",  # Works on CPU or GPU
-    torch_dtype="auto"
 )
 
 generator = pipeline(
@@ -114,9 +113,11 @@ Answer in short and clear sentences.
 
     response = generator(prompt, max_new_tokens=150, do_sample=True, temperature=0.6)
     text = response[0]["generated_text"]
-    # Remove prompt if repeated
+
+    # Remove prompt repetition
     if prompt.strip() in text:
         text = text.split(prompt.strip())[-1].strip()
+
     return text
 
 # -----------------------------