FROM ghcr.io/ggml-org/llama.cpp:server@sha256:a9b230ae078a6dd947ede5825f227af9fc5fc9e126c4717bebd4762c02601285

RUN useradd -m -u 1000 user
USER user

ENV HOME=/home/user
ENV PATH=/home/user/.local/bin:$PATH

WORKDIR /app

ADD --chown=user https://huggingface.co/mradermacher/Qwen3.5-0.8B-heretic-GGUF/resolve/main/Qwen3.5-0.8B-heretic.f16.gguf /app/model.gguf

EXPOSE 7860

CMD ["--port", "7860", "--no-mmap", "--flash-attn", "on", "--fit", "on", "--ctx-size", "8192", "--cache-ram", " 768", "--samplers", "top_k;temperature", "--temperature", "0.3", "--top-k", "40", "--backend-sampling", "--webui-mcp-proxy", "-m", "/app/model.gguf", "--chat-template-kwargs", "{ \"enable_thinking\": true }", "--no-mmproj"]