Dockerfile

FROM ghcr.io/ggml-org/llama.cpp:full

WORKDIR /app

# প্রয়োজনীয় প্যাকেজ ইনস্টল
RUN apt update && apt install -y python3-pip
RUN pip install -U huggingface_hub

# Hugging Face থেকে মডেল ডাউনলোড
RUN python3 -c 'from huggingface_hub import hf_hub_download; \
    repo="LiquidAI/LFM2-24B-A2B-GGUF"; \
    hf_hub_download(repo_id=repo, filename="LFM2-24B-A2B-Q4_K_M.gguf", local_dir="/app");'

# সার্ভার কনফিগারেশন (Thinking মোড ডিজেবল করা হয়েছে)
ENTRYPOINT ["/app/llama-server"]
CMD ["-m", "/app/LFM2-24B-A2B-Q4_K_M.gguf", \
     "--host", "0.0.0.0", \
     "--port", "7860", \
     "-t", "2", \
     "--cache-type-k", "f16", \
     "--cache-type-v", "f16", \
     "-c", "50000", \
     "-n", "4096", \
     "--jinja"]