start_server.py

import argparse
import atexit
import logging
import os
import signal
import socket
import subprocess
import sys
import time

import requests
import tomli
import uvicorn
from dotenv import load_dotenv

from core.config import get_settings
from core.logging_config import setup_logging

# Global variable to store the worker process
worker_process = None


def wait_for_redis(host="localhost", port=6379, timeout=20):
    """
    Wait for Redis to become available.

    Args:
        host: Redis host address
        port: Redis port number
        timeout: Maximum time to wait in seconds

    Returns:
        True if Redis becomes available within the timeout, False otherwise
    """
    logging.info(f"Waiting for Redis to be available at {host}:{port}...")
    t0 = time.monotonic()
    while time.monotonic() - t0 < timeout:
        try:
            with socket.create_connection((host, port), timeout=1):
                logging.info("Redis is accepting connections.")
                return True
        except (OSError, socket.error):
            logging.debug(f"Redis not available yet, retrying... ({int(time.monotonic() - t0)}s elapsed)")
            time.sleep(0.3)

    logging.error(f"Redis not reachable after {timeout}s")
    return False


def check_and_start_redis():
    """Check if the Redis container is running, start if necessary."""
    try:
        # Check if container exists and is running
        check_running_cmd = ["docker", "ps", "-q", "-f", "name=morphik-redis"]
        running_container = subprocess.check_output(check_running_cmd).strip()

        if running_container:
            logging.info("Redis container (morphik-redis) is already running.")
            return

        # Check if container exists but is stopped
        check_exists_cmd = ["docker", "ps", "-a", "-q", "-f", "name=morphik-redis"]
        existing_container = subprocess.check_output(check_exists_cmd).strip()

        if existing_container:
            logging.info("Starting existing Redis container (morphik-redis)...")
            subprocess.run(["docker", "start", "morphik-redis"], check=True, capture_output=True)
            logging.info("Redis container started.")
        else:
            logging.info("Creating and starting Redis container (morphik-redis)...")
            subprocess.run(
                ["docker", "run", "-d", "--name", "morphik-redis", "-p", "6379:6379", "redis"],
                check=True,
                capture_output=True,
            )
            logging.info("Redis container created and started.")

    except subprocess.CalledProcessError as e:
        logging.error(f"Failed to manage Redis container: {e}")
        logging.error(f"Stderr: {e.stderr.decode() if e.stderr else 'N/A'}")
        sys.exit(1)
    except FileNotFoundError:
        logging.error("Docker command not found. Please ensure Docker is installed and in PATH.")
        sys.exit(1)


def start_arq_worker():
    """Start the ARQ worker as a subprocess."""
    global worker_process
    try:
        logging.info("Starting ARQ worker...")

        # Ensure logs directory exists
        log_dir = os.path.join(os.getcwd(), "logs")
        os.makedirs(log_dir, exist_ok=True)

        # Worker log file paths
        worker_log_path = os.path.join(log_dir, "worker.log")

        # Open log files
        worker_log = open(worker_log_path, "a")

        # Add timestamp to log
        timestamp = subprocess.check_output(["date"]).decode().strip()
        worker_log.write(f"\n\n--- Worker started at {timestamp} ---\n\n")
        worker_log.flush()

        # Use sys.executable to ensure the same Python environment is used
        worker_cmd = [sys.executable, "-m", "arq", "core.workers.ingestion_worker.WorkerSettings"]

        # Start the worker with output redirected to log files
        worker_process = subprocess.Popen(
            worker_cmd,
            stdout=worker_log,
            stderr=worker_log,
            env=dict(os.environ, PYTHONUNBUFFERED="1"),  # Ensure unbuffered output
        )
        logging.info(f"ARQ worker started with PID: {worker_process.pid}")
        logging.info(f"Worker logs available at: {worker_log_path}")
    except Exception as e:
        logging.error(f"Failed to start ARQ worker: {e}")
        sys.exit(1)


def cleanup_processes():
    """Stop the ARQ worker process on exit."""
    global worker_process
    if worker_process and worker_process.poll() is None:  # Check if process is still running
        logging.info(f"Stopping ARQ worker (PID: {worker_process.pid})...")

        # Log the worker termination
        try:
            log_dir = os.path.join(os.getcwd(), "logs")
            worker_log_path = os.path.join(log_dir, "worker.log")

            with open(worker_log_path, "a") as worker_log:
                timestamp = subprocess.check_output(["date"]).decode().strip()
                worker_log.write(f"\n\n--- Worker stopping at {timestamp} ---\n\n")
        except Exception as e:
            logging.warning(f"Could not write worker stop message to log: {e}")

        # Send SIGTERM first for graceful shutdown
        worker_process.terminate()
        try:
            # Wait a bit for graceful shutdown
            worker_process.wait(timeout=5)
            logging.info("ARQ worker stopped gracefully.")
        except subprocess.TimeoutExpired:
            logging.warning("ARQ worker did not terminate gracefully, sending SIGKILL.")
            worker_process.kill()  # Force kill if it doesn't stop
            logging.info("ARQ worker killed.")

        # Close any open file descriptors for the process
        if hasattr(worker_process, "stdout") and worker_process.stdout:
            worker_process.stdout.close()
        if hasattr(worker_process, "stderr") and worker_process.stderr:
            worker_process.stderr.close()

    # Optional: Add Redis container stop logic here if desired
    # try:
    #     logging.info("Stopping Redis container...")
    #     subprocess.run(["docker", "stop", "morphik-redis"], check=False, capture_output=True)
    # except Exception as e:
    #     logging.warning(f"Could not stop Redis container: {e}")


# Register the cleanup function to be called on script exit
atexit.register(cleanup_processes)
# Also register for SIGINT (Ctrl+C) and SIGTERM
signal.signal(signal.SIGINT, lambda sig, frame: sys.exit(0))
signal.signal(signal.SIGTERM, lambda sig, frame: sys.exit(0))


def check_ollama_running(base_url):
    """Check if Ollama is running and accessible at the given URL."""
    try:
        api_url = f"{base_url}/api/tags"
        response = requests.get(api_url, timeout=2)
        return response.status_code == 200
    except requests.RequestException:
        return False


def get_ollama_usage_info():
    """Check if Ollama is required based on the configuration file and get base URLs."""
    try:
        with open("morphik.toml", "rb") as f:
            config = tomli.load(f)

        ollama_configs = []

        # Get registered Ollama models first
        ollama_models = {}
        if "registered_models" in config:
            for model_key, model_config in config["registered_models"].items():
                model_name = model_config.get("model_name", "")
                if "ollama" in model_name:
                    api_base = model_config.get("api_base")
                    if api_base:
                        ollama_models[model_key] = api_base

        # Check which components are using Ollama models
        components_to_check = ["embedding", "completion", "rules", "graph", "parser.vision"]

        for component in components_to_check:
            if component == "parser.vision":
                # Special handling for parser.vision
                if "parser" in config and "vision" in config["parser"]:
                    model_key = config["parser"]["vision"].get("model")
                    if model_key in ollama_models:
                        ollama_configs.append({"component": component, "base_url": ollama_models[model_key]})
            else:
                # Standard component check
                if component in config:
                    model_key = config[component].get("model")
                    if model_key in ollama_models:
                        ollama_configs.append({"component": component, "base_url": ollama_models[model_key]})

        # Add contextual chunking model check
        if (
            "parser" in config
            and config["parser"].get("use_contextual_chunking")
            and "contextual_chunking_model" in config["parser"]
        ):
            model_key = config["parser"]["contextual_chunking_model"]
            if model_key in ollama_models:
                ollama_configs.append(
                    {
                        "component": "parser.contextual_chunking",
                        "base_url": ollama_models[model_key],
                    }
                )

        return ollama_configs
    except Exception as e:
        logging.error(f"Error checking Ollama configuration: {e}")
        return []


def main():
    # Parse command line arguments
    parser = argparse.ArgumentParser(description="Start the Morphik server")
    parser.add_argument(
        "--log",
        choices=["debug", "info", "warning", "error"],
        default="info",
        help="Set the logging level",
    )
    parser.add_argument(
        "--skip-ollama-check",
        action="store_true",
        help="Skip Ollama availability check",
    )
    args = parser.parse_args()

    # Set up logging first with specified level
    setup_logging(log_level=args.log.upper())

    # Check and start Redis container
    check_and_start_redis()

    # Load environment variables from .env file
    load_dotenv()

    # Check if Ollama is required and running
    if not args.skip_ollama_check:
        ollama_configs = get_ollama_usage_info()

        if ollama_configs:
            # Group configs by base_url to avoid redundant checks
            base_urls = {}
            for config in ollama_configs:
                if config["base_url"] not in base_urls:
                    base_urls[config["base_url"]] = []
                base_urls[config["base_url"]].append(config["component"])

            all_running = True
            for base_url, components in base_urls.items():
                if not check_ollama_running(base_url):
                    print(f"ERROR: Ollama is not accessible at {base_url}")
                    print(f"This URL is used by these components: {', '.join(components)}")
                    all_running = False

            if not all_running:
                print("\nPlease ensure Ollama is running at the configured URLs before starting the server")
                print("Run with --skip-ollama-check to bypass this check")
                sys.exit(1)
            else:
                component_list = [config["component"] for config in ollama_configs]
                print(f"Ollama is running and will be used for: {', '.join(component_list)}")

    # Load settings (this will validate all required env vars)
    settings = get_settings()

    # Wait for Redis to be available (using environment variables or defaults)
    redis_host = os.environ.get("REDIS_HOST", "127.0.0.1")
    redis_port = int(os.environ.get("REDIS_PORT", "6379"))
    if not wait_for_redis(host=redis_host, port=redis_port):
        logging.error("Cannot start server without Redis. Please ensure Redis is running.")
        sys.exit(1)

    # Start ARQ worker in the background
    start_arq_worker()

    # Start server (this is blocking)
    logging.info("Starting Uvicorn server...")
    uvicorn.run(
        "core.api:app",
        host=settings.HOST,
        port=settings.PORT,
        loop="asyncio",
        log_level=args.log,
        # reload=settings.RELOAD # Reload might interfere with subprocess management
    )


if __name__ == "__main__":
    main()