install-bonsai.sh

#!/bin/bash
# =============================================================================
# Bonsai-1.7B Installer - One-line setup for llama.cpp + model
# Usage: curl -sL https://gist.githubusercontent.com/madkoding/c7fc0ddb806ac47b56749cc30e1a5914/raw | bash
# =============================================================================

set -e

INSTALL_DIR="$HOME/llamaBonsai"
MODEL_URL="https://huggingface.co/prism-ml/Bonsai-1.7B-gguf/resolve/main/Bonsai-1.7B.gguf"
MODEL_NAME="Bonsai-1.7B.gguf"
PORT=8080

RED='\033[0;31m'
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
CYAN='\033[0;36m'
BOLD='\033[1m'
RESET='\033[0m'

clear
cat <<'EOF'
${CYAN}
  ██████╗ ██████╗ ██╗██████╗ ██╗███████╗███████╗
  ██╔══██╗██╔══██╗██║██╔══██╗██║██╔════╝██╔════╝
  ██║  ██║██████╔╝██║██║  ██║██║█████╗  ███████╗
  ██║  ██║██╔══██╗██║██║  ██║██║██╔══╝  ╚════██║
  ██████╔╝██║  ██║██║██████╔╝██║███████╗███████║
  ╚═════╝ ╚═╝  ╚═╝╚═╝╚═════╝ ╚═╝╚══════╝╚══════╝${RESET}
  ${BOLD}Bonsai-1.7B GGUF Q1_0 (1-bit)${RESET}
  ${YELLOW}prism-ml/Bonsai-1.7B-gguf${RESET}

  ${CYAN}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━${RESET}
  ${GREEN}Installing to:${RESET} $INSTALL_DIR
  ${CYAN}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━${RESET}
EOF

sleep 1

detect_cuda() {
    if command -v nvidia-smi &> /dev/null && nvidia-smi &> /dev/null; then
        echo "cuda"
    else
        echo "cpu"
    fi
}

check_deps() {
    local missing=()
    for cmd in git cmake curl; do
        if ! command -v $cmd &> /dev/null; then
            missing+=($cmd)
        fi
    done
    if [ ${#missing[@]} -ne 0 ]; then
        echo -e "${RED}Error: Missing dependencies: ${missing[*]}${RESET}"
        echo "Install with: sudo apt install ${missing[*]}"
        exit 1
    fi
}

mkdir -p "$INSTALL_DIR"
cd "$INSTALL_DIR"

echo -e "\n${YELLOW}[1/4]${RESET} Checking dependencies..."
check_deps

echo -e "${YELLOW}[2/4]${RESET} Setting up llama.cpp..."
if [ ! -d "llama.cpp" ]; then
    echo "  Cloning PrismML fork (includes Q1_0 kernels)..."
    git clone --depth 1 https://github.com/PrismML-Eng/llama.cpp 2>&1 | tail -2
else
    echo "  llama.cpp already present"
fi

BACKEND=$(detect_cuda)
if [ "$BACKEND" = "cuda" ]; then
    GPU_NAME=$(nvidia-smi --query-gpu=name --format=csv,noheader | head -1)
    echo -e "  ${GREEN}GPU detected:${RESET} $GPU_NAME"
    echo "  Building with CUDA support..."
    cd llama.cpp
    cmake -B build -DGGML_CUDA=ON 2>&1 | tail -3
    cmake --build build -j$(nproc) 2>&1 | tail -5
else
    echo -e "  ${YELLOW}No GPU/CUDA found, building CPU version...${RESET}"
    cd llama.cpp
    cmake -B build -DGGML_CUDA=OFF 2>&1 | tail -3
    cmake --build build -j$(nproc) 2>&1 | tail -5
fi

echo -e "\n${YELLOW}[3/4]${RESET} Downloading model (248 MB)..."
if [ ! -f "models/$MODEL_NAME" ]; then
    mkdir -p models
    curl -L "$MODEL_URL" -o "models/$MODEL_NAME" --progress-bar
else
    echo "  Model already downloaded"
fi

echo -e "\n${YELLOW}[4/4]${RESET} Starting server..."
./build/bin/llama-server \
    -m "models/$MODEL_NAME" \
    --host 0.0.0.0 \
    --port $PORT \
    -ngl 99 \
    --log-disable \
    &>/dev/null &

sleep 2

cat <<'EOF'

${CYAN}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━${RESET}
${GREEN}✓ Bonsai-1.7B is running!${RESET}

  ${BOLD}URL:${RESET} http://localhost:$PORT
  ${BOLD}API:${RESET} http://localhost:$PORT/v1/chat/completions

${YELLOW}Press Ctrl+C to stop${RESET}
${CYAN}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━${RESET}

EOF

trap "echo -e '\n${RED}Stopping server...${RESET}'; pkill -f llama-server 2>/dev/null; exit 0" SIGINT SIGTERM
wait