Update app.py

app.py

@@ -1,25 +1,13 @@
-# =============================================================================
-#  codeMax — Qwen 3.6 27B CoderBot on ZeroGPU (Gradio 6)
-#  Model: DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF
-# =============================================================================
-
-import spaces
+```python
 import gradio as gr
 import os
 import json
 from pathlib import Path
-from typing import List, Optional, Generator
+from typing import Optional
 
-# --- Config ----------------------------------------------------------------
-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", "8192"))
+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"))
 
 SYSTEM_PROMPT = """You are codeMax, an uncensored 27B coding assistant. You excel at:
@@ -27,14 +15,11 @@ SYSTEM_PROMPT = """You are codeMax, an uncensored 27B coding assistant. You exce
 - Debugging, refactoring, reviewing complex codebases
 - Explaining algorithms, architecture, system design
 - Reading shared files and answering about them
-
 Rules: put code in ```language blocks. Be precise and thorough."""
 
-# --- Model singleton -------------------------------------------------------
 _llm = None
 _model_path: Optional[str] = None
 
-
 def _download():
     global _model_path
     if _model_path is not None:
@@ -45,14 +30,13 @@ def _download():
     print(f"[MODEL] Cached -> {_model_path}")
     return _model_path
 
-
 def _load():
     global _llm
     if _llm is not None:
         return _llm
     from llama_cpp import Llama
     path = _download()
-    print(f"[MODEL] Loading (n_gpu_layers={N_GPU})...")
+    print(f"[MODEL] Loading (n_gpu_layers={N_GPU}, n_ctx={N_CTX})...")
     _llm = Llama(
         model_path=path,
         n_ctx=N_CTX,
@@ -64,10 +48,10 @@ def _load():
     print("[MODEL] Ready.")
     return _llm
 
+print("[BOOT] Loading model at startup...")
+_load()
 
-# --- File readers ----------------------------------------------------------
 def _read_text(path, enc="utf-8"):
-    """Safe text reader with encoding fallback."""
     for e in [enc, "utf-8-sig", "latin-1", "cp1252", "utf-16"]:
         try:
             with open(path, "r", encoding=e) as f:
@@ -77,13 +61,10 @@ def _read_text(path, enc="utf-8"):
     with open(path, "r", encoding="utf-8", errors="replace") as f:
         return f.read()
 
-
 def parse_file(file_path, file_name):
-    """Parse a single uploaded file into context-ready markdown."""
     suffix = Path(file_name).suffix.lower()
     name = Path(file_name).name
 
-    # ---- Code / text files ----
     if suffix in (
         ".txt", ".md", ".py", ".ts", ".tsx", ".js", ".jsx", ".mjs",
         ".css", ".scss", ".less", ".html", ".htm", ".xml", ".svg",
@@ -99,17 +80,12 @@ def parse_file(file_path, file_name):
     ):
         content = _read_text(file_path)
         lang = suffix.lstrip(".")
-        if suffix == ".md":
-            lang = "markdown"
-        elif suffix in (".yml",):
-            lang = "yaml"
-        elif suffix in (".tf", ".tfvars"):
-            lang = "hcl"
-        elif suffix in (".htm",):
-            lang = "html"
+        if suffix == ".md": lang = "markdown"
+        elif suffix in (".yml",): lang = "yaml"
+        elif suffix in (".tf", ".tfvars"): lang = "hcl"
+        elif suffix in (".htm",): lang = "html"
         return f"### `{name}`\n```{lang}\n{content}\n```\n\n---\n"
 
-    # ---- JSON ----
     if suffix == ".json":
         content = _read_text(file_path)
         try:
@@ -119,10 +95,8 @@ def parse_file(file_path, file_name):
             pass
         return f"### `{name}`\n```json\n{content}\n```\n\n---\n"
 
-    # ---- CSV ----
     if suffix == ".csv":
-        import csv
-        import io
+        import csv, io
         content = _read_text(file_path)
         rows = list(csv.reader(io.StringIO(content)))
         if len(rows) > 51:
@@ -137,7 +111,6 @@ def parse_file(file_path, file_name):
         table.insert(1, sep)
         return f"### `{name}`\n" + "\n".join(table) + "\n\n---\n"
 
-    # ---- Word documents ----
     if suffix == ".docx":
         try:
             import docx
@@ -147,28 +120,22 @@ def parse_file(file_path, file_name):
         except Exception as e:
             return f"### `{name}` (DOCX error: {e})\n\n---\n"
 
-    # ---- PDF ----
     if suffix == ".pdf":
         try:
             import pdfplumber
             with pdfplumber.open(file_path) as pdf:
-                text = "\n\n".join(
-                    page.extract_text() or "" for page in pdf.pages
-                )
+                text = "\n\n".join(page.extract_text() or "" for page in pdf.pages)
             return f"### `{name}`\n{text}\n\n---\n"
         except Exception as e:
             return f"### `{name}` (PDF error: {e})\n\n---\n"
 
-    # ---- Fallback ----
     try:
         content = _read_text(file_path)
         return f"### `{name}`\n```\n{content[:10000]}\n```\n\n---\n"
     except Exception as e:
         return f"### `{name}` (could not read: {e})\n\n---\n"
 
-
 def parse_all_files(files):
-    """Parse all uploaded files into a single context string."""
     if not files:
         return ""
     parts = []
@@ -188,36 +155,22 @@ def parse_all_files(files):
             parts.append(f"### `{name}`\nParse error: {e}\n\n---\n")
     return "\n".join(parts)
 
-
-# --- Generation (wrapped with @spaces.GPU for ZeroGPU) ---------------------
-@spaces.GPU
 def respond(message, history, uploaded_files):
-    """Generate a streaming response from the model."""
     file_context = parse_all_files(uploaded_files)
     messages = [{"role": "system", "content": SYSTEM_PROMPT}]
 
-    # Inject uploaded files as context
     if file_context:
-        messages.append({
-            "role": "user",
-            "content": "[Uploaded files — read carefully]\n\n" + file_context,
-        })
-        messages.append({
-            "role": "assistant",
-            "content": "Got it! I've read through all the uploaded files.",
-        })
+        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."})
 
-    # Append conversation history (Gradio 6 "messages" format)
     for entry in history:
         role = entry.get("role", "user")
         content = entry.get("content", "")
         if content:
             messages.append({"role": role, "content": content})
 
-    # Append the current user message
     messages.append({"role": "user", "content": message})
 
-    # Stream from the model
     llm = _load()
     stream = llm.create_chat_completion(
         messages=messages,
@@ -245,38 +198,26 @@ def respond(message, history, uploaded_files):
         {"role": "assistant", "content": partial},
     ]
 
-
-# --- Gradio UI ------------------------------------------------------------
 def create_demo():
     with gr.Blocks(title="codeMax — Qwen 3.6 27B Coder") as demo:
         gr.Markdown(
             "# codeMax\n"
-            "**Qwen 3.6 27B · Uncensored · Code-optimized · ZeroGPU**\n"
+            "**Qwen 3.6 27B · Uncensored · Code-optimized · Dedicated GPU**\n"
             "Drop code, docs, JSON, CSVs — chat with a 27B coding LLM."
         )
 
         with gr.Row(equal_height=True):
-            # ---- Left column: chat ----
             with gr.Column(scale=3):
                 chatbot = gr.Chatbot(
                     label="Chat",
                     height=580,
-                    avatar_images=(
-                        None,
-                        "https://huggingface.co/front/assets/huggingface_logo-noborder.svg",
-                    ),
+                    avatar_images=(None, "https://huggingface.co/front/assets/huggingface_logo-noborder.svg"),
                 )
                 with gr.Row():
-                    msg = gr.Textbox(
-                        placeholder="Ask about code, share files, or chat...",
-                        scale=8,
-                        show_label=False,
-                        container=False,
-                    )
+                    msg = gr.Textbox(placeholder="Ask about code, share files, or chat...", scale=8, show_label=False, container=False)
                     send = gr.Button(">", scale=1, variant="primary", min_width=48)
                 clear_btn = gr.Button("Clear", size="sm")
 
-            # ---- Right column: file upload ----
             with gr.Column(scale=1):
                 gr.Markdown("### Drop Files")
                 files = gr.File(
@@ -292,7 +233,6 @@ def create_demo():
                 )
                 uploaded_info = gr.Markdown("_No files uploaded._")
 
-        # ---- Event handlers ----
         def update_info(uploaded):
             if not uploaded:
                 return "_No files uploaded._"
@@ -308,23 +248,15 @@ def create_demo():
             for h in respond(message, history, current_files):
                 yield h
 
-        msg.submit(stream_response, [msg, chatbot, files], [chatbot]).then(
-            lambda: "", None, [msg]
-        )
-        send.click(stream_response, [msg, chatbot, files], [chatbot]).then(
-            lambda: "", None, [msg]
-        )
+        msg.submit(stream_response, [msg, chatbot, files], [chatbot]).then(lambda: "", None, [msg])
+        send.click(stream_response, [msg, chatbot, files], [chatbot]).then(lambda: "", None, [msg])
         clear_btn.click(lambda: [], None, chatbot, queue=False)
         files.change(update_info, files, uploaded_info)
 
     return demo
 
-
 if __name__ == "__main__":
     demo = create_demo()
     demo.queue(default_concurrency_limit=1, max_size=4)
-    demo.launch(
-        css="""
-        footer { display: none !important; }
-        """
-    )
+    demo.launch(css="""footer { display: none !important; }""")
+```