Update Dockerfile

Dockerfile

@@ -1,38 +1,45 @@
 FROM mambaorg/micromamba:1.5.3
 
-# Switch to root briefly to setup permissions (standard HF requirement)
+# 1. Switch to root to set permissions (Standard HF Space requirement)
 USER root
-
-# Create the working directory and ensure the user owns it
 WORKDIR /app
 RUN chown -R mambauser:mambauser /app
 
-# Switch back to the non-root user provided by the image
+# 2. Switch back to the non-root user
 USER mambauser
 
-# Initialize shell for the user
+# 3. Initialize the shell
 RUN micromamba shell init --shell=bash --root-prefix=~/micromamba
 
-# Copy requirements first to leverage caching
-COPY --chown=mambauser:mambauser requirements.txt /app/requirements.txt
-
-# Install Python and Pip dependencies via Micromamba to avoid system conflicts
-RUN micromamba install -y -n base python=3.10 pip -c conda-forge && \
+# 4. Install Python AND Compilers (gcc, g++, cmake)
+# We need these to build llama-cpp-python and for the AI to compile tools later
+RUN micromamba install -y -n base \
+    python=3.10 \
+    pip \
+    cmake \
+    make \
+    gcc_linux-64 \
+    gxx_linux-64 \
+    -c conda-forge && \
     micromamba clean --all --yes
 
-# Activate the environment and install pip packages
+# 5. Copy requirements
+COPY --chown=mambauser:mambauser requirements.txt /app/requirements.txt
+
+# 6. Install Python packages
+# We set CMAKE_ARGS to ensure it builds for CPU specifically
 ARG MAMBA_DOCKERFILE_ACTIVATE=1
+ENV CMAKE_ARGS="-DLLAMA_BLAS=OFF"
 RUN pip install --no-cache-dir -r requirements.txt
 
-# Download a fast, small CPU model (Qwen 2.5 1.5B Instruct)
-# We download it now so the space starts fast
+# 7. Download the Model (Qwen 2.5 1.5B - Fast CPU)
 RUN python -c "from huggingface_hub import hf_hub_download; hf_hub_download(repo_id='Qwen/Qwen2.5-1.5B-Instruct-GGUF', filename='qwen2.5-1.5b-instruct-q4_k_m.gguf', local_dir='./models')"
 
-# Copy the rest of the application
+# 8. Copy the application code
 COPY --chown=mambauser:mambauser . /app
 
-# Expose the single allowed port
+# 9. Expose the single allowed port
 EXPOSE 7860
 
-# Start the application
+# 10. Start the app
 CMD ["python", "main.py"]