Update app.py

app.py

@@ -1,6 +1,8 @@
 """
-CodeNyx – bullet-proof HF-Space chatbot
-Public model → no token required by default
+CodeNyx – 100 % ready-to-run Space
+- Model loads at startup, no button needed
+- 3 B StarCoder2 4-bit → < 8 GB RAM
+- Answers generated directly from The Stack v2 (code-only corpus)
 """
 
 import os
@@ -12,59 +14,51 @@ from transformers import (
     BitsAndBytesConfig,
     TextIteratorStreamer
 )
-from huggingface_hub import login
 from threading import Thread
 
 BOT_NAME = "CodeNyx"
-MODEL_ID = "bigcode/starcoder2-3b"           # PUBLIC
+MODEL_ID = "bigcode/starcoder2-3b"
+
 MAX_NEW_TOK = 1024
 TEMPERATURE = 0.2
 TOP_P = 0.9
 
 SYSTEM_PROMPT = (
-    f"You are {BOT_NAME}, an expert open-source coding assistant. "
-    "Always provide concise, runnable code snippets with short explanations."
+    f"You are {BOT_NAME}, an expert open-source coding assistant trained on "
+    "The Stack v2. Always return concise, runnable code snippets with brief explanations."
 )
 
 # ------------------------------------------------------------------
-# 1. Model loader with None-safe token
+# 1. Startup model loader (runs once)
 # ------------------------------------------------------------------
-def load_model(token: str | None):
-    """Return (tok, model) or raise."""
-    token = (token or "").strip() or None          # None-safe
-    if token:
-        login(token)
-
-    bnb_config = BitsAndBytesConfig(
-        load_in_4bit=True,
-        bnb_4bit_compute_dtype=torch.float16,
-        bnb_4bit_quant_type="nf4",
-        bnb_4bit_use_double_quant=True,
-    )
-
-    tok = AutoTokenizer.from_pretrained(
-        MODEL_ID,
-        use_auth_token=token,
-        trust_remote_code=True,
-    )
-    if tok.pad_token is None:
-        tok.pad_token = tok.eos_token
+print("🚀 CodeNyx – loading model …")
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_compute_dtype=torch.float16,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_use_double_quant=True,
+)
 
-    mdl = AutoModelForCausalLM.from_pretrained(
-        MODEL_ID,
-        quantization_config=bnb_config,
-        device_map="auto",
-        use_auth_token=token,
-        trust_remote_code=True,
-    )
-    return tok, mdl
+tokenizer = AutoTokenizer.from_pretrained(
+    MODEL_ID,
+    trust_remote_code=True,
+)
+if tokenizer.pad_token is None:
+    tokenizer.pad_token = tokenizer.eos_token
+
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID,
+    quantization_config=bnb_config,
+    device_map="auto",
+    trust_remote_code=True,
+)
+print("✅ CodeNyx model and tokenizer loaded.")
 
 # ------------------------------------------------------------------
 # 2. Chat helpers
 # ------------------------------------------------------------------
-tokenizer, model = None, None
-
 def build_prompt(history, user_input):
+    """Turn chat history into the model’s prompt format."""
     messages = [{"role": "system", "content": SYSTEM_PROMPT}]
     for human, ai in history:
         messages += [{"role": "user", "content": human},
@@ -101,13 +95,15 @@ def bot_turn(history):
 # 3. Gradio UI
 # ------------------------------------------------------------------
 with gr.Blocks(title=f"{BOT_NAME} – AI Pair-Programmer") as demo:
-    gr.Markdown(f"# 🤖 {BOT_NAME} – AI Pair-Programmer\n*3 B params, public model, zero-config*")
-
-    status_lbl = gr.Label(value="Loading model …")
-    chatbot = gr.Chatbot(height=500)
+    gr.Markdown(f"""
+    # 🤖 {BOT_NAME} – AI Pair-Programmer
+    Trained on **The Stack v2** (4 T tokens, permissive licences).  
+    Ask any coding question and receive **runnable code + short explanations**.
+    """)
+    chatbot = gr.Chatbot(height=500, label=f"{BOT_NAME} Chat")
     with gr.Row():
         msg = gr.Textbox(
-            placeholder="Ask me to write / debug / explain code …",
+            placeholder="Type your programming question here …",
             lines=2,
             scale=8,
             show_label=False,
@@ -116,25 +112,13 @@ with gr.Blocks(title=f"{BOT_NAME} – AI Pair-Programmer") as demo:
         send_btn = gr.Button("Send", scale=1, variant="primary")
     clear_btn = gr.Button("🗑️ Clear")
 
-    def _load():
-        global tokenizer, model
-        try:
-            tokenizer, model = load_model(None)   # None = no token
-            return "✅ Model loaded!"
-        except Exception as e:
-            return f"❌ {e}"
-
-    demo.load(_load, None, status_lbl)
-
-    def _send(user_msg, hist):
-        return user_turn(user_msg, hist)
-
-    def _bot(hist):
-        yield from bot_turn(hist)
-
-    msg.submit(_send, [msg, chatbot], [msg, chatbot]).then(_bot, chatbot, chatbot)
-    send_btn.click(_send, [msg, chatbot], [msg, chatbot]).then(_bot, chatbot, chatbot)
-    clear_btn.click(lambda: None, None, chatbot)
+    msg.submit(user_turn, [msg, chatbot], [msg, chatbot], queue=False).then(
+        bot_turn, chatbot, chatbot
+    )
+    send_btn.click(user_turn, [msg, chatbot], [msg, chatbot], queue=False).then(
+        bot_turn, chatbot, chatbot
+    )
+    clear_btn.click(lambda: None, None, chatbot, queue=False)
 
 # ------------------------------------------------------------------
 # 4. Launch