ok

Dockerfile

@@ -8,25 +8,26 @@ ENV PYTHONDONTWRITEBYTECODE=1 \
 
 WORKDIR /app
 
-# Minimal dependencies
+# Install build dependencies
 RUN apt-get update && apt-get install -y \
+    build-essential \
+    cmake \
     curl \
+    git \
     && rm -rf /var/lib/apt/lists/*
 
 # Create user
 RUN useradd -m -u 1000 user
 RUN mkdir -p /app/cache /app/models && chown -R user:user /app
 
-USER user
-
-# Copy the wheel file you downloaded
-COPY --chown=user:user llama_cpp_python-0.3.16-cp310-cp310-manylinux_2_17_x86_64.whl /tmp/
-
-# Install pip
+# Install pip as root
 RUN pip install --no-cache-dir --upgrade pip
 
-# Install the wheel directly
-RUN pip install --no-cache-dir /tmp/llama_cpp_python-0.3.16-cp310-cp310-manylinux_2_17_x86_64.whl
+USER user
+
+# Build and install llama-cpp-python with proper flags
+RUN CMAKE_ARGS="-DGGML_BLAS=OFF -DGGML_NATIVE=OFF" \
+    pip install --no-cache-dir llama-cpp-python==0.3.2
 
 # Install other dependencies
 COPY --chown=user:user requirements.txt .