Update app.py

app.py

@@ -1,6 +1,8 @@
 """
-CodeNyx – zero-config Gradio chatbot
-Auto-handles Hugging Face tokens without stdin prompts.
+CodeNyx – HF-Space-ready chatbot
+- 3 B parameters
+- 4-bit quant → < 8 GB VRAM
+- Accepts HF token only for gated models (none required here)
 """
 
 import os
@@ -15,40 +17,26 @@ from transformers import (
 from huggingface_hub import login
 from threading import Thread
 
-MODEL_ID      = "bigcode/starcoder2-3b-instruct"
+# ------------------------------------------------------------------
+# 1. Configuration
+# ------------------------------------------------------------------
 BOT_NAME      = "CodeNyx"
-SYSTEM_PROMPT = (f"You are {BOT_NAME}, an expert open-source coding assistant. "
-                 "Always provide concise, runnable code snippets with short explanations.")
-
+MODEL_ID      = "bigcode/starcoder2-3b"        #  <- public, no token needed
+SYSTEM_PROMPT = (
+    f"You are {BOT_NAME}, an expert open-source coding assistant. "
+    "Always provide concise, runnable code snippets with short explanations."
+)
 MAX_NEW_TOK   = 1024
 TEMPERATURE   = 0.2
 TOP_P         = 0.9
 
 # ------------------------------------------------------------------
-# 1. Token helper (Gradio popup instead of stdin)
-# ------------------------------------------------------------------
-def ensure_token(token_value):
-    """
-    token_value comes from the Gradio UI the first time.
-    We cache it in environment variable HF_TOKEN and login once.
-    Returns True on success.
-    """
-    token = token_value.strip()
-    if not token:
-        return False
-    os.environ["HF_TOKEN"] = token
-    login(token)
-    return True
-
-# ------------------------------------------------------------------
-# 2. Lazy model loader (once token is ready)
+# 2. Lightweight 4-bit loader
 # ------------------------------------------------------------------
-model, tokenizer = None, None
-
-def load_model():
-    global model, tokenizer
-    if model is not None:
-        return True  # already loaded
+def load_model(token: str = None):
+    """Return (tokenizer, model) or raise."""
+    if token:
+        login(token)  # only if provided
 
     bnb_config = BitsAndBytesConfig(
         load_in_4bit=True,
@@ -59,7 +47,7 @@ def load_model():
 
     tokenizer = AutoTokenizer.from_pretrained(
         MODEL_ID,
-        use_auth_token=os.getenv("HF_TOKEN"),
+        use_auth_token=token or None,
         trust_remote_code=True,
     )
     if tokenizer.pad_token is None:
@@ -69,13 +57,13 @@ def load_model():
         MODEL_ID,
         quantization_config=bnb_config,
         device_map="auto",
-        use_auth_token=os.getenv("HF_TOKEN"),
+        use_auth_token=token or None,
         trust_remote_code=True,
     )
-    return True
+    return tokenizer, model
 
 # ------------------------------------------------------------------
-# 3. Chat logic
+# 3. Chat helpers
 # ------------------------------------------------------------------
 def build_prompt(history, user_input):
     messages = [{"role": "system", "content": SYSTEM_PROMPT}]
@@ -118,48 +106,50 @@ def bot_turn(history):
         yield history
 
 # ------------------------------------------------------------------
-# 4. Gradio UI flow
+# 4. Gradio UI
 # ------------------------------------------------------------------
+tokenizer, model = None, None  # lazy load
+
 with gr.Blocks(title=f"{BOT_NAME} – AI Pair-Programmer") as demo:
     gr.Markdown(f"""
     # 🤖 {BOT_NAME} – AI Pair-Programmer  
-    *3 B params, 100 % free-tier friendly.*  
-    Paste your Hugging Face token **once** if asked, then chat away.
+    *Public model — no token needed.*  
+    Ask any coding question and get **runnable code + short explanations**.
     """)
 
-    token_box  = gr.Textbox(label="🤗 Hugging Face Token (only first time)", type="password")
-    load_btn   = gr.Button("Authorize")
-    status_lbl = gr.Label(value="Waiting for token …")
-
-    # --- main chat controls (hidden until token OK) ---
-    with gr.Column(visible=False) as chat_col:
-        chatbot = gr.Chatbot(height=450)
-        with gr.Row():
-            msg = gr.Textbox(
-                placeholder="Ask me to write / debug / explain code …",
-                lines=2,
-                scale=8,
-                show_label=False,
-                container=False,
-            )
-            send_btn = gr.Button("Send", scale=1, variant="primary")
-        clear_btn = gr.Button("🗑️ Clear")
+    # Optional token box (for future gated models)
+    token_box = gr.Textbox(label="🤗 HF Token (optional)", type="password", visible=False)
+    load_btn  = gr.Button("Load / Reload", visible=False)
+    status_lbl = gr.Label(value="Loading model …")
+
+    # Chat area
+    chatbot = gr.Chatbot(height=450)
+    with gr.Row():
+        msg = gr.Textbox(
+            placeholder="Ask me to write, debug, or explain code …",
+            lines=2,
+            scale=8,
+            show_label=False,
+            container=False,
+        )
+        send_btn = gr.Button("Send", scale=1, variant="primary")
+    clear_btn = gr.Button("🗑️ Clear")
 
     # ------------------------------------------------------------------
-    # 5. Event wiring
+    # 5. Events
     # ------------------------------------------------------------------
-    def _auth(token):
-        ok = ensure_token(token)
-        if ok:
-            try:
-                load_model()
-                return gr.update(visible=False), gr.update(visible=True), "✅ Ready! Start coding."
-            except Exception as e:
-                return gr.update(visible=True), gr.update(visible=False), f"❌ Error: {e}"
-        else:
-            return gr.update(visible=True), gr.update(visible=False), "❌ Invalid token."
-
-    load_btn.click(_auth, token_box, [token_box, chat_col, status_lbl])
+    def _load(token):
+        global tokenizer, model
+        try:
+            tokenizer, model = load_model(token if token.strip() else None)
+            return "✅ Model loaded!"
+        except Exception as e:
+            return f"❌ {e}"
+
+    load_btn.click(_load, token_box, status_lbl)
+
+    # auto-load once on start
+    demo.load(_load, token_box, status_lbl)
 
     def _send(user_msg, hist):
         return user_turn(user_msg, hist)
@@ -171,4 +161,8 @@ with gr.Blocks(title=f"{BOT_NAME} – AI Pair-Programmer") as demo:
     send_btn.click(_send, [msg, chatbot], [msg, chatbot]).then(_bot, chatbot, chatbot)
     clear_btn.click(lambda: None, None, chatbot)
 
-demo.queue().launch(server_name="0.0.0.0", server_port=7860, share=True)
+# ------------------------------------------------------------------
+# 6. Launch
+# ------------------------------------------------------------------
+if __name__ == "__main__":
+    demo.queue().launch(server_name="0.0.0.0", server_port=7860, share=True)