#!/usr/bin/env bash

set -euo pipefail

CONTAINER_NAME="nim-llama-3.1-8b"
IMAGE="nvcr.io/nim/meta/llama-3.1-8b-instruct-dgx-spark:latest"
CACHE_DIR="/home/azurian/.cache/nim/llama-3.1-8b"
PORT="8000"
ENV_FILE="/root/.ngc-nim.env"

if [[ ! -f "$ENV_FILE" ]]; then
    echo "Errore: manca $ENV_FILE"
    echo "Il file deve contenere:"
    echo "NGC_API_KEY=la_tua_chiave"
    exit 1
fi

mkdir -p "$CACHE_DIR"
chmod -R a+rwX "$CACHE_DIR"

docker rm -f "$CONTAINER_NAME" 2>/dev/null || true

echo "Avvio Llama 3.1 8B NIM per DGX Spark..."

docker run -d \
  --name "$CONTAINER_NAME" \
  --restart unless-stopped \
  --runtime=nvidia \
  --gpus all \
  --ipc=host \
  --shm-size=16G \
  --env-file "$ENV_FILE" \
  -e NIM_GPU_MEM_FRACTION=0.20 \
  -v "$CACHE_DIR:/opt/nim/.cache" \
  -p "${PORT}:8000" \
  "$IMAGE"

echo
echo "Container avviato: $CONTAINER_NAME"
echo "Log:"
echo "docker logs -f $CONTAINER_NAME"
echo
echo "Health:"
echo "curl http://localhost:${PORT}/v1/health/ready"
echo
echo "Models:"
echo "curl -s http://localhost:${PORT}/v1/models | jq"