Update Dockerfile

Dockerfile

@@ -1,39 +1,37 @@
-FROM python:3.11-slim
+# Use a lightweight Python base image
+FROM python:3.10-slim
 
-# Install system dependencies
-RUN apt-get update && apt-get install -y --no-install-recommends gcc libc6-dev && rm -rf /var/lib/apt/lists/*
+# Install only curl to fetch the model (no build-essential or compiler tools needed)
+RUN apt-get update && apt-get install -y \
+    curl \
+    && rm -rf /var/lib/apt/lists/*
 
+# Set working directory
 WORKDIR /app
 
-# Install Python dependencies
-RUN pip install --no-cache-dir flask flask-cors Pillow litert-lm huggingface_hub
+# Upgrade pip and instruct pip to install the pre-compiled CPU wheels 
+RUN pip install --no-cache-dir --upgrade pip && \
+    PIP_EXTRA_INDEX_URL=https://github.io \
+    pip install --no-cache-dir "llama-cpp-python[server]"
 
-# 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
+# Download the model
+RUN curl -L -o meta-llama-3.1-8b-instruct-abliterated.Q4_K_M.gguf "https://huggingface.co/mlabonne/Meta-Llama-3.1-8B-Instruct-abliterated-GGUF/resolve/main/meta-llama-3.1-8b-instruct-abliterated.Q4_K_M.gguf"
 
-# 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
-# KV-cache size (prompt + generation combined). The model file supports up to 32k.
-ENV MAX_NUM_TOKENS=32768
+# Expose Hugging Face Space's mandatory port
 EXPOSE 7860
 
-CMD ["python3", "server.py"]
+# Set environment variables for the llama-cpp server
+# This enforces the API key requirement
+ENV API_KEY=sk-dummykey
+
+# Launch the server with optimized CPU and 16k context settings
+# --cache_type_k/v q8_0 drastically reduces the RAM footprint of the 16k context
+# Launch server
+CMD ["python3", "-m", "llama_cpp.server", \
+       "--model", "/app/meta-llama-3.1-8b-instruct-abliterated.Q4_K_M.gguf", \
+     "--host", "0.0.0.0", \
+     "--port", "7860", \
+     "--n_ctx", "16384", \
+     "--cache_type_k", "q8_0", \
+     "--cache_type_v", "q8_0", \
+     "--api_key", "sk-dummykey"]