#!/bin/bash
# run.sh — convenience wrapper around docker compose
set -e

cd "$(dirname "$0")"

ENV_FILE="${ENV_FILE:-}"
COMPOSE_ARGS=()

if [ -n "$ENV_FILE" ]; then
    COMPOSE_ARGS+=(--env-file "$ENV_FILE")
fi

CMD="${1:-up}"
shift 2>/dev/null || true

case "$CMD" in
    up)
        docker compose "${COMPOSE_ARGS[@]}" up -d "${@}"
        echo "vLLM container started. Logs: docker compose logs -f"
        ;;
    down)
        docker compose "${COMPOSE_ARGS[@]}" down "${@}"
        ;;
    logs)
        docker compose "${COMPOSE_ARGS[@]}" logs -f "${@}"
        ;;
    rebuild)
        docker compose "${COMPOSE_ARGS[@]}" build --no-cache "${@}"
        docker compose "${COMPOSE_ARGS[@]}" up -d
        ;;
    exec)
        shift
        docker compose "${COMPOSE_ARGS[@]}" exec vllm "${@:-bash}"
        ;;
    test)
        echo "Waiting for server..."
        for i in $(seq 1 30); do
            if curl -s -o /dev/null -w "%{http_code}" http://localhost:8000/v1/models 2>/dev/null | grep -q 200; then
                echo "Ready!"
                curl -s http://localhost:8000/v1/models | jq .
                exit 0
            fi
            sleep 10
        done
        echo "Timeout waiting for server"
        exit 1
        ;;
    bench)
        curl -s http://localhost:8000/v1/chat/completions \
            -H 'Content-Type: application/json' \
            -d '{
                "model": "huihui-qwen36-27b-local",
                "messages": [{"role": "user", "content": "Hello, what is 2+2? Answer briefly."}],
                "max_tokens": 100,
                "temperature": 0
            }' | jq .
        ;;
    *)
        echo "Usage: $0 {up|down|logs|rebuild|exec|test|bench}"
        echo ""
        echo "  up       Start container (detached)"
        echo "  down     Stop container"
        echo "  logs     Tail logs"
        echo "  rebuild  Rebuild image and restart"
        echo "  exec     Execute command in container (default: bash)"
        echo "  test     Ping /v1/models until ready"
        echo "  bench    Quick inference smoke test"
        exit 1
        ;;
esac