start.sh

#!/bin/bash

# 設定工作目錄
export WORK="/home/user/app"
cd "$WORK" || exit 1

# 建立 Python 虛擬環境 (可選)
python3 -m venv venv
source venv/bin/activate

# 安裝 Python 套件
pip install --upgrade pip
pip install gradio openai aiohttp

echo "Unzip and download model..."
unzip llama_cpp_avx512.zip > /dev/null 2>&1
#wget -O model.gguf https://huggingface.co/Intel/gpt-oss-20b-gguf-q4ks-AutoRound/resolve/main/gpt-oss-20b-32x2.4B-Q4_K_S.gguf > /dev/null 2>&1
wget -O model.gguf https://huggingface.co/unsloth/gpt-oss-20b-GGUF/resolve/main/gpt-oss-20b-Q4_K_M.gguf > /dev/null 2>&1

echo "Start llama.cpp server..."
nohup ./llama-server \
    -m model.gguf \
    --port 8000 \
    --host 0.0.0.0 \
    --threads 2 \
    --ctx-size 16384 \
    --mlock \
    --jinja \
    --temp 0.2 \
    --top-p 0.85 \
    -v \
    > llama.log 2>&1 &

# 持續監控 log，但不阻塞 python app.py
#tail -n 50 -f llama.log &
#sleep 200

echo "Start app.py..."
python app.py