Update Dockerfile

Dockerfile

@@ -1,11 +1,11 @@
-# Use a lightweight Python base (Debian-based)
+# Use a lightweight Python base
 FROM python:3.10-slim
 
 # Set working directory
 WORKDIR /app
 
-# 1. Install System Dependencies
-# 'cmake' and 'build-essential' are strictly required if wheel installation fails.
+# 1. Install Build Tools
+# strictly required for the fallback compilation
 RUN apt-get update && apt-get install -y \
     git \
     curl \
@@ -14,11 +14,15 @@ RUN apt-get update && apt-get install -y \
     libopenblas-dev \
     && rm -rf /var/lib/apt/lists/*
 
-# 2. Download Retro Font (VT323)
+# 2. CRITICAL: Upgrade pip
+# Old pip versions (default in 3.10-slim) often fail to find binary wheels,
+# forcing a slow source build. Upgrading fixes this.
+RUN pip install --upgrade pip setuptools wheel
+
+# 3. Download Retro Font (VT323)
 RUN curl -L -o /app/VT323.ttf https://github.com/google/fonts/raw/main/ofl/vt323/VT323-Regular.ttf
 
-# 3. Install Python Dependencies
-# We split the installation to ensure core libs are present before llama-cpp-python attempts to load.
+# 4. Install Python Dependencies (Split for caching)
 RUN pip install --no-cache-dir \
     torch \
     torchvision \
@@ -34,21 +38,19 @@ RUN pip install --no-cache-dir \
     safetensors \
     scipy
 
-# 4. Install Llama-CPP-Python (Corrected)
-# REMOVED: The '--prefer-binary' flag pointing to the wrong index.
-# CHANGED: We now let pip build from source using the installed CMake/GCC. 
-# This guarantees compatibility with the Debian container (glibc).
-# We set CMAKE_ARGS to ensure it builds a CPU-only version (no CUDA req).
-ENV CMAKE_ARGS="-DLLAMA_BLAS=ON -DLLAMA_NATIVE=OFF"
-RUN pip install --no-cache-dir llama-cpp-python
+# 5. Install Llama-CPP-Python (The Fix)
+# - CMAKE_ARGS="-DLLAMA_NATIVE=OFF": Prevents hanging on architecture detection.
+# - -v: Verbose mode so you can see the build scrolling instead of freezing.
+ENV CMAKE_ARGS="-DLLAMA_NATIVE=OFF -DLLAMA_BLAS=ON -DGGML_NATIVE=OFF"
+RUN pip install --no-cache-dir -v llama-cpp-python
 
-# 5. Create a non-root user
+# 6. Create a non-root user
 RUN useradd -m -u 1000 user
 USER user
 ENV HOME=/home/user \
     PATH=/home/user/.local/bin:$PATH
 
-# 6. Write the Monolith Application to disk
+# 7. Write the Monolith Application to disk
 COPY --chown=user <<'EOF' app.py
 import sys, os, io, base64, json, pickle, time
 import numpy as np
@@ -60,12 +62,13 @@ from flask import Flask, request, send_file, render_template_string
 from flask_sock import Sock
 from diffusers import StableDiffusionPipeline, AutoencoderTiny, LCMScheduler
 from PIL import Image, ImageDraw
-# Wrap Llama import to prevent crash if model not found
+
+# Graceful degradation if library fails (prevents crash loop)
 try:
     from llama_cpp import Llama
 except ImportError:
     Llama = None
-    print("[!] Warning: llama-cpp-python not loaded correctly.")
+    print("[!] CRITICAL: Llama-cpp-python failed to import. AI features will be limited.")
 
 # ============================================================================
 # 1. FRONTEND ASSET
@@ -506,6 +509,6 @@ if __name__ == '__main__':
     app.run(host='0.0.0.0', port=7860, threaded=True)
 EOF
 
-# 7. Launch the Monolith
+# 8. Launch the Monolith
 EXPOSE 7860
 CMD ["python", "app.py"]