FROM python:3.10-slim

WORKDIR /app

ENV PYTHONUNBUFFERED=1

# تثبيت أدوات + build خفيف
RUN apt-get update && apt-get install -y \
    git \
    build-essential \
    cmake \
    wget \
    && rm -rf /var/lib/apt/lists/*

# تحميل llama.cpp (نسخة صغيرة)
RUN git clone --depth 1 https://github.com/ggerganov/llama.cpp.git

WORKDIR /app/llama.cpp

# build بخيوط قليلة لتجنب OOM 🔥
RUN cmake -B build -DLLAMA_BUILD_EXAMPLES=ON \
 && cmake --build build -j2

WORKDIR /app

# تحميل النموذج
RUN wget -O model.gguf "https://huggingface.co/mradermacher/LFM2.5-1.2B-Thinking-Kimi-V2-Heretic-Uncensored-DISTILL-GGUF/resolve/main/LFM2.5-1.2B-Thinking-Kimi-V2-Heretic-Uncensored-DISTILL.Q4_K_M.gguf"

# Gradio
RUN pip install gradio

COPY app.py .

CMD ["python", "app.py"]