Update Dockerfile

Dockerfile

@@ -1,16 +1,19 @@
-FROM ghcr.io/ggml-org/llama.cpp:server
+FROM python:3.11-slim
 
-# Forces the server to use a writeable temporary directory for Hugging Face compatibility
-ENV HOME=/tmp
+WORKDIR /app
 
-# Expose the mandatory port for Hugging Face routing
+# System deps
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    git curl && rm -rf /var/lib/apt/lists/*
+
+# Python deps
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# App code
+COPY app.py .
+
+# HF Spaces listens on 7860
 EXPOSE 7860
 
-# Run the official, ultra-optimized C++ server binary directly
-CMD [ \
-    "--host", "0.0.0.0", \
-    "--port", "7860", \
-    "--model", "https://huggingface.co/bartowski/google_gemma-3-4b-it-GGUF/resolve/main/google_gemma-3-4b-it-Q4_K_M.gguf", \
-    "--ctx-size", "2048", \
-    "--threads", "2" \
-]
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860", "--workers", "1"]