Update app.py

app.py

@@ -1,13 +1,16 @@
+import llama_bootstrap  # noqa: F401 — MUST be first
+
 import gradio as gr
 import os
 import json
+import sys
 from pathlib import Path
 from typing import Optional
 
 MODEL_REPO = os.environ.get("MODEL_REPO", "DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF")
 MODEL_FILE = os.environ.get("MODEL_FILE", "Qwen3.6-27B-NEO-CODE-HERE-2T-OT-Q4_K_M.gguf")
 N_CTX = int(os.environ.get("N_CTX", "4096"))
-N_GPU = int(os.environ.get("N_GPU_LAYERS", "-1"))
+N_GPU = int(os.environ.get("N_GPU_LAYERS", os.environ.get("DAVIDAU_N_GPU_LAYERS", "-1")))
 
 SYSTEM_PROMPT = """You are codeMax, an uncensored 27B coding assistant. You excel at:
 - Writing production code in Python, TypeScript, JavaScript, Rust, Go, C/C++, SQL, shell
@@ -24,9 +27,9 @@ def _download():
     if _model_path is not None:
         return _model_path
     from huggingface_hub import hf_hub_download
-    print(f"[MODEL] Downloading {MODEL_FILE}...")
+    print(f"[MODEL] Downloading {MODEL_FILE}...", file=sys.stderr)
     _model_path = hf_hub_download(repo_id=MODEL_REPO, filename=MODEL_FILE)
-    print(f"[MODEL] Cached -> {_model_path}")
+    print(f"[MODEL] Cached -> {_model_path}", file=sys.stderr)
     return _model_path
 
 def _load():
@@ -35,7 +38,7 @@ def _load():
         return _llm
     from llama_cpp import Llama
     path = _download()
-    print(f"[MODEL] Loading (n_gpu_layers={N_GPU}, n_ctx={N_CTX})...")
+    print(f"[MODEL] Loading (n_gpu_layers={N_GPU}, n_ctx={N_CTX})...", file=sys.stderr)
     _llm = Llama(
         model_path=path,
         n_ctx=N_CTX,
@@ -44,10 +47,10 @@ def _load():
         verbose=False,
         seed=-1,
     )
-    print("[MODEL] Ready.")
+    print("[MODEL] Ready.", file=sys.stderr)
     return _llm
 
-print("[BOOT] Loading model at startup...")
+print("[BOOT] Loading model at startup...", file=sys.stderr)
 _load()
 
 def _read_text(path, enc="utf-8"):
@@ -160,7 +163,7 @@ def respond(message, history, uploaded_files):
 
     if file_context:
         messages.append({"role": "user", "content": "[Uploaded files]\n\n" + file_context})
-        messages.append({"role": "assistant", "content": "Got it! I've read through all the uploaded files."})
+        messages.append({"role": "assistant", "content": "Got it! I’ve read through all the uploaded files."})
 
     for entry in history:
         role = entry.get("role", "user")