FROM python:3.12-slim

# Bharat-Tiny-LLM demo — runs the GGUF build via llama.cpp (lighter than fp16)
WORKDIR /app

# System dependencies
RUN apt-get update && apt-get install -y --no-install-recommends \
    git git-lfs ffmpeg libsm6 libxext6 cmake rsync libgl1 \
    && rm -rf /var/lib/apt/lists/* && git lfs install

# Pre-download the GGUF model at build time so the Space starts fast
RUN pip install --no-cache-dir "huggingface-hub<1.22.0" "llama-cpp-python>=0.3.0"
RUN python3 -c "\
from huggingface_hub import hf_hub_download; \
hf_hub_download(repo_id='eulogik/Bharat-Tiny-LLM-GGUF', filename='bharat-tiny-llm-q4_k_m.gguf'); \
"

# Copy app files
COPY app.py requirements.txt /app/

# Install app dependencies
RUN pip install --no-cache-dir -r requirements.txt

ENV GRADIO_SERVER_NAME=0.0.0.0
ENV GRADIO_SERVER_PORT=7860
ENV GRADIO_ANALYTICS_ENABLED=False
EXPOSE 7860

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