Update app.py

app.py

@@ -1,6 +1,7 @@
 import streamlit as st
 import torch
 import threading
+
 from peft import PeftModel
 from transformers import (
     AutoModelForCausalLM,
@@ -10,7 +11,7 @@ from transformers import (
 
 # ---------------- CONFIG ----------------
 MODEL_ID = "Neon-AI/Niche"
-MAX_NEW_TOKENS = 16384
+MAX_NEW_TOKENS = 512
 TEMPERATURE = 0.7
 TOP_P = 0.9
 # ----------------------------------------
@@ -32,16 +33,19 @@ def load_model():
         device_map=None
     )
 
-    # If LoRA adapters exist, this will load them
+    # Load LoRA if present
     try:
         model = PeftModel.from_pretrained(base_model, MODEL_ID)
     except Exception:
-        model = base_model  # fallback if no LoRA
+        model = base_model
 
     model.to("cpu")
     model.eval()
     return tokenizer, model
 
+
+tokenizer, model = load_model()
+
 # -------- SESSION STATE --------
 if "history" not in st.session_state:
     st.session_state.history = []
@@ -53,18 +57,18 @@ if st.button("Send") and prompt.strip():
     st.session_state.history.append(("You", prompt))
 
     system_instructions = (
-    "You are Niche, a concise and intelligent AI. "
-    "Answer directly and naturally. "
-    "Do not use greetings, pleasantries, or offers of help. "
-    "Respond only with the requested information or explanation. "
-    "Keep responses short, clear, and focused. "
-    "Your owner is Neon. Mention Neon only if explicitly asked."
-    "Neon is a man the pronoun sghould always be 'him'"
+        "You are Niche, a concise and intelligent AI. "
+        "Answer directly and naturally. "
+        "Do not use greetings, pleasantries, or offers of help. "
+        "Respond only with the requested information or explanation. "
+        "Keep responses short, clear, and focused. "
+        "Your owner is Neon. Mention Neon only if explicitly asked. "
+        "Neon is a man; always use 'him'."
     )
 
     chat = [
-    {"role": "system", "content": system_instructions},
-    {"role": "user", "content": prompt}
+        {"role": "system", "content": system_instructions},
+        {"role": "user", "content": prompt}
     ]
 
     inputs = tokenizer.apply_chat_template(