fix: move all heavy work to runtime, minimal Dockerfile for fast build

Dockerfile

@@ -1,34 +1,15 @@
 FROM python:3.11-slim
 
-# Install build + runtime dependencies in one layer
+# Install all dependencies (build + runtime)
 RUN apt-get update && apt-get install -y \
     git cmake clang build-essential curl libgomp1 \
     && rm -rf /var/lib/apt/lists/*
 
-# Clone bitnet.cpp (shallow clone for speed)
-RUN git clone --depth 1 --recursive https://github.com/microsoft/BitNet.git /opt/BitNet
-
-# Install BitNet Python deps
-RUN pip install --no-cache-dir -r /opt/BitNet/requirements.txt
-
-# Build bitnet.cpp WITHOUT downloading the model (just compile the binary)
-# We use cmake directly instead of setup_env.py to avoid the model download
-RUN cd /opt/BitNet && \
-    cmake -B build -DCMAKE_BUILD_TYPE=Release \
-        -DGGML_BITNET_ARM_TL1=OFF \
-        -DGGML_BITNET_X86_TL2=OFF && \
-    cmake --build build --config Release -j$(nproc) --target llama-server llama-cli
-
 # Create non-root user
 RUN useradd -ms /bin/bash user
 WORKDIR /home/user/app
 
-# Copy just the binaries we need
-RUN cp /opt/BitNet/build/bin/llama-server /home/user/app/ && \
-    cp /opt/BitNet/build/bin/llama-cli /home/user/app/ && \
-    rm -rf /opt/BitNet
-
-# Install Python app deps
+# Install Python deps first (cached layer)
 COPY requirements.txt .
 RUN pip install --no-cache-dir -r requirements.txt
 
@@ -36,9 +17,7 @@ RUN pip install --no-cache-dir -r requirements.txt
 COPY app.py start.sh ./
 RUN chmod +x start.sh
 
-# Create model directory
-RUN mkdir -p /home/user/app/models && \
-    chown -R user:user /home/user/app
+RUN chown -R user:user /home/user/app
 
 USER user