Dockerfile

FROM ghcr.io/ggerganov/llama.cpp:full

RUN apt update && apt install wget -y

# Ganti ke model Dhanishtha GGUF (Q4_K_M)
RUN wget "https://huggingface.co/QuantFactory/finance-Llama3-8B-GGUF/resolve/main/finance-Llama3-8B.Q8_0.gguf?download=true" -O /Llama3finance.gguf

CMD ["--server", "-m", "/Llama3finance.gguf", "--port", "7860", "--host", "0.0.0.0", "-n", "4096"]