Update Dockerfile

Dockerfile

@@ -1,3 +1,4 @@
+# We use the full image, not slim, to ensure better compatibility
 FROM python:3.10-slim
 
 # 1. Setup User (Required by Spaces)
@@ -8,19 +9,25 @@ ENV HOME=/home/user \
 
 WORKDIR $HOME/app
 
-# 2. BYPASS BUILD: Install the Pre-Compiled Engine
-# We use a specific URL to get the "CPU Wheel" (Binary) directly.
-# This takes 30 seconds instead of 30 minutes.
-RUN pip install --no-cache-dir --upgrade pip && \
-    pip install llama-cpp-python \
-    --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu
+# 2. INSTALL THE FACTORY TOOLS (Mandatory)
+# We need these. Without them, the build fails.
+USER root
+RUN apt-get update && apt-get install -y \
+    build-essential \
+    cmake \
+    git \
+    && rm -rf /var/lib/apt/lists/*
+USER user
+
+# 3. INSTALL WITH FALLBACK
+# We try to download the binary. If that fails, we now have the tools (gcc) to build it.
+# We set CMAKE_ARGS to make the build faster if it happens.
+ENV CMAKE_ARGS="-DLLAMA_BLAS=OFF"
 
-# 3. Install the Rest
 COPY --chown=user requirements.txt requirements.txt
-# We remove llama-cpp-python from requirements so it doesn't try to rebuild it
-RUN sed -i '/llama-cpp-python/d' requirements.txt && \
-    pip install --no-cache-dir -r requirements.txt
+RUN pip install --no-cache-dir --upgrade pip && \
+    pip install --no-cache-dir -r requirements.txt --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu
 
-# 4. Copy Code & Launch
+# 4. LAUNCH
 COPY --chown=user . .
 CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]