Update Dockerfile

Dockerfile

@@ -1,37 +1,58 @@
-FROM python:3.11-slim
+FROM python:3.12-slim
 
-# Install system dependencies
-RUN apt-get update && apt-get install -y --no-install-recommends gcc libc6-dev && rm -rf /var/lib/apt/lists/*
+ENV DEBIAN_FRONTEND=noninteractive
+ENV PYTHONUNBUFFERED=1
+ENV PIP_NO_CACHE_DIR=1
+ENV LD_LIBRARY_PATH=/opt/llama
 
 WORKDIR /app
 
-# Install Python dependencies
-RUN pip install --no-cache-dir flask flask-cors Pillow litert-lm huggingface_hub
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    git \
+    cmake \
+    build-essential \
+    curl \
+    ca-certificates \
+    libgomp1 \
+    && rm -rf /var/lib/apt/lists/*
+
+RUN git clone --depth 1 https://github.com/ggml-org/llama.cpp.git /tmp/llama.cpp \
+    && cmake -S /tmp/llama.cpp -B /tmp/llama.cpp/build \
+        -DGGML_NATIVE=OFF \
+        -DGGML_OPENMP=ON \
+        -DLLAMA_BUILD_SERVER=ON \
+        -DLLAMA_BUILD_TESTS=OFF \
+        -DLLAMA_BUILD_EXAMPLES=ON \
+    && cmake --build /tmp/llama.cpp/build \
+        --config Release \
+        --target llama-server \
+        -j2 \
+    && mkdir -p /opt/llama \
+    && cp /tmp/llama.cpp/build/bin/llama-server /opt/llama/ \
+    && cp /tmp/llama.cpp/build/bin/*.so* /opt/llama/ \
+    && rm -rf /tmp/llama.cpp
+
+RUN pip install huggingface_hub
+
+RUN python3 -c "\
+from huggingface_hub import hf_hub_download; \
+hf_hub_download( \
+    repo_id='LiquidAI/LFM2.5-2.6B-GGUF', \
+    filename='LFM2.5-2.6B-QAD-Q4_0.gguf', \
+    local_dir='/app/models' \
+)"
 
-# Create dummy Vulkan library to satisfy litert-lm on CPU
-RUN printf 'void vkGetInstanceProcAddr(){}\nvoid vkCreateInstance(){}\nvoid vkDestroyInstance(){}\nvoid vkEnumeratePhysicalDevices(){}\nvoid vkGetDeviceProcAddr(){}' > /tmp/vk.c \
-    && gcc -shared -fPIC -o /app/libvulkan.so.1 /tmp/vk.c \
-    && rm /tmp/vk.c
-
-# Create model directory
-RUN mkdir -p /app/models/gemma
-
-# --- KEY FIX: Updated 'hf download' to fetch the 12B model ---
-RUN hf download litert-community/gemma-4-E2B-it-litert-lm gemma-4-E2B-it.litertlm --local-dir /app/models/gemma
-# Copy the server code
-COPY server.py .
-
-# Setup permissions for Hugging Face Spaces (requires user 1000)
-RUN useradd -m -u 1000 user
-RUN chown -R user:user /app
-USER user
-
-# Set environment variables
-ENV PORT=7860
-ENV LD_LIBRARY_PATH=/app
-
-# --- KEY FIX: Updated path to point to the 12B model file ---
-ENV GEMMA_MODEL_PATH=/app/models/gemma/gemma-4-E2B-it.litertlm
 EXPOSE 7860
 
-CMD ["python3", "server.py"]
+CMD ["/opt/llama/llama-server", \
+     "--model", "/app/models/LFM2.5-2.6B-QAD-Q4_0.gguf", \
+     "--host", "0.0.0.0", \
+     "--port", "7860", \
+     "--alias", "LFM2.5-2.6B", \
+     "--threads", "2", \
+     "--ctx-size", "32768", \
+     "--cache-type-k", "q8_0", \
+     "--cache-type-v", "q8_0", \
+     "--parallel", "1", \
+     "--cont-batching", \
+     "--flash-attn", "auto"]