mdify-cli 1.5.0__py3-none-any.whl → 2.0.0__py3-none-any.whl

mdify/__init__.py

@@ -1,3 +1,3 @@
 """mdify - Convert documents to Markdown via Docling container."""
 
-__version__ = "1.5.0"
+__version__ = "2.0.0"

mdify/cli.py

@@ -21,6 +21,8 @@ from urllib.error import URLError
 from urllib.request import urlopen
 
 from . import __version__
+from mdify.container import DoclingContainer
+from mdify.docling_client import convert_file
 
 # Configuration
 MDIFY_HOME = Path.home() / ".mdify"
@@ -29,7 +31,8 @@ PYPI_API_URL = "https://pypi.org/pypi/mdify-cli/json"
 CHECK_INTERVAL_SECONDS = 86400  # 24 hours
 
 # Container configuration
-DEFAULT_IMAGE = "ghcr.io/tiroq/mdify-runtime:latest"
+DEFAULT_IMAGE = "ghcr.io/docling-project/docling-serve-cpu:main"
+GPU_IMAGE = "ghcr.io/docling-project/docling-serve-cu126:main"
 SUPPORTED_RUNTIMES = ("docker", "podman")
 
 
@@ -37,10 +40,11 @@ SUPPORTED_RUNTIMES = ("docker", "podman")
 # Update checking functions
 # =============================================================================
 
+
 def _get_remote_version(timeout: int = 5) -> Optional[str]:
     """
     Fetch the latest version from PyPI.
-    
+
     Returns:
         Version string (e.g., "1.1.0") or None if fetch failed.
     """
@@ -56,16 +60,16 @@ def _get_remote_version(timeout: int = 5) -> Optional[str]:
 def _should_check_for_update() -> bool:
     """
     Determine if we should check for updates based on last check time.
-    
+
     Returns:
         True if check should be performed, False otherwise.
     """
     if os.environ.get("MDIFY_NO_UPDATE_CHECK", "").lower() in ("1", "true", "yes"):
         return False
-    
+
     if not LAST_CHECK_FILE.exists():
         return True
-    
+
     try:
         last_check = float(LAST_CHECK_FILE.read_text().strip())
         elapsed = time.time() - last_check
@@ -86,18 +90,18 @@ def _update_last_check_time() -> None:
 def _compare_versions(current: str, remote: str) -> bool:
     """
     Compare version strings.
-    
+
     Returns:
         True if remote version is newer than current.
     """
     try:
         current_parts = [int(x) for x in current.split(".")]
         remote_parts = [int(x) for x in remote.split(".")]
-        
+
         max_len = max(len(current_parts), len(remote_parts))
         current_parts.extend([0] * (max_len - len(current_parts)))
         remote_parts.extend([0] * (max_len - len(remote_parts)))
-        
+
         return remote_parts > current_parts
     except (ValueError, AttributeError):
         return False
@@ -106,15 +110,15 @@ def _compare_versions(current: str, remote: str) -> bool:
 def check_for_update(force: bool = False) -> None:
     """
     Check for updates and prompt user to upgrade if available.
-    
+
     Args:
         force: If True, check regardless of last check time and show errors.
     """
     if not force and not _should_check_for_update():
         return
-    
+
     remote_version = _get_remote_version()
-    
+
     if remote_version is None:
         if force:
             print(
@@ -124,19 +128,19 @@ def check_for_update(force: bool = False) -> None:
             )
             sys.exit(1)
         return
-    
+
     _update_last_check_time()
-    
+
     if not _compare_versions(__version__, remote_version):
         if force:
             print(f"mdify is up to date (version {__version__})")
         return
-    
-    print(f"\n{'='*50}")
+
+    print(f"\n{'=' * 50}")
     print(f"A new version of mdify-cli is available!")
     print(f"  Current version: {__version__}")
     print(f"  Latest version:  {remote_version}")
-    print(f"{'='*50}")
+    print(f"{'=' * 50}")
     print(f"\nTo upgrade, run:")
     print(f"  pipx upgrade mdify-cli")
     print(f"  # or: pip install --upgrade mdify-cli\n")
@@ -146,13 +150,18 @@ def check_for_update(force: bool = False) -> None:
 # Container runtime functions
 # =============================================================================
 
-def detect_runtime(preferred: str) -> Optional[str]:
+
+def detect_runtime(preferred: str, explicit: bool = True) -> Optional[str]:
     """
     Detect available container runtime.
-    
+
     Args:
         preferred: Preferred runtime ('docker' or 'podman')
-        
+        explicit: If True, warn when falling back to alternative.
+                  If False, silently use alternative without warning.
+                  Note: This only controls warning emission; selection order
+                  is always preferred → alternative regardless of this flag.
+
     Returns:
         Path to runtime executable, or None if not found.
     """
@@ -160,25 +169,28 @@ def detect_runtime(preferred: str) -> Optional[str]:
     runtime_path = shutil.which(preferred)
     if runtime_path:
         return runtime_path
-    
+
     # Try alternative
     alternative = "podman" if preferred == "docker" else "docker"
     runtime_path = shutil.which(alternative)
     if runtime_path:
-        print(f"Warning: {preferred} not found, using {alternative}", file=sys.stderr)
+        if explicit:
+            print(
+                f"Warning: {preferred} not found, using {alternative}", file=sys.stderr
+            )
         return runtime_path
-    
+
     return None
 
 
 def check_image_exists(runtime: str, image: str) -> bool:
     """
     Check if container image exists locally.
-    
+
     Args:
         runtime: Path to container runtime
         image: Image name/tag
-        
+
     Returns:
         True if image exists locally.
     """
@@ -196,18 +208,18 @@ def check_image_exists(runtime: str, image: str) -> bool:
 def pull_image(runtime: str, image: str, quiet: bool = False) -> bool:
     """
     Pull container image.
-    
+
     Args:
         runtime: Path to container runtime
         image: Image name/tag
         quiet: Suppress progress output
-        
+
     Returns:
         True if pull succeeded.
     """
     if not quiet:
         print(f"Pulling image: {image}")
-    
+
     try:
         result = subprocess.run(
             [runtime, "pull", image],
@@ -222,9 +234,9 @@ def pull_image(runtime: str, image: str, quiet: bool = False) -> bool:
 
 def format_size(size_bytes: int) -> str:
     """Format file size in human-readable format."""
-    for unit in ['B', 'KB', 'MB', 'GB']:
+    for unit in ["B", "KB", "MB", "GB"]:
         if size_bytes < 1024:
-            return f"{size_bytes:.1f} {unit}" if unit != 'B' else f"{size_bytes} {unit}"
+            return f"{size_bytes:.1f} {unit}" if unit != "B" else f"{size_bytes} {unit}"
         size_bytes /= 1024
     return f"{size_bytes:.1f} TB"
 
@@ -244,29 +256,33 @@ def format_duration(seconds: float) -> str:
 
 class Spinner:
     """A simple spinner to show progress during long operations."""
-    
+
     def __init__(self):
-        self.frames = ['⠋', '⠙', '⠹', '⠸', '⠼', '⠴', '⠦', '⠧', '⠇', '⠏']
+        self.frames = ["⠋", "⠙", "⠹", "⠸", "⠼", "⠴", "⠦", "⠧", "⠇", "⠏"]
         self.running = False
         self.thread = None
         self.start_time = None
-    
+
     def _spin(self):
         idx = 0
         while self.running:
             elapsed = time.time() - self.start_time
             frame = self.frames[idx % len(self.frames)]
-            print(f"\r{self.prefix} {frame} ({format_duration(elapsed)})", end="", flush=True)
+            print(
+                f"\r{self.prefix} {frame} ({format_duration(elapsed)})",
+                end="",
+                flush=True,
+            )
             idx += 1
             time.sleep(0.1)
-    
+
     def start(self, prefix: str = ""):
         self.prefix = prefix
         self.running = True
         self.start_time = time.time()
         self.thread = threading.Thread(target=self._spin, daemon=True)
         self.thread.start()
-    
+
     def stop(self):
         self.running = False
         if self.thread:
@@ -275,93 +291,45 @@ class Spinner:
         print(f"\r{' ' * 80}\r", end="", flush=True)
 
 
-def run_container(
-    runtime: str,
-    image: str,
-    input_file: Path,
-    output_file: Path,
-    mask_pii: bool = False,
-) -> Tuple[bool, str, float]:
-    """
-    Run container to convert a single file.
-    
-    Args:
-        runtime: Path to container runtime
-        image: Image name/tag
-        input_file: Absolute path to input file
-        output_file: Absolute path to output file
-        mask_pii: Whether to mask PII in images
-        
-    Returns:
-        Tuple of (success: bool, message: str, elapsed_seconds: float)
-    """
-    start_time = time.time()
-    
-    # Ensure output directory exists
-    output_file.parent.mkdir(parents=True, exist_ok=True)
-    
-    # Mount directories
-    input_dir = input_file.parent
-    output_dir = output_file.parent
-    
-    # Container paths
-    container_in = f"/work/in/{input_file.name}"
-    container_out = f"/work/out/{output_file.name}"
-    
-    cmd = [
-        runtime, "run", "--rm",
-        "-v", f"{input_dir}:/work/in:ro",
-        "-v", f"{output_dir}:/work/out",
-        image,
-        "--in", container_in,
-        "--out", container_out,
-    ]
-    
-    if mask_pii:
-        cmd.append("--mask")
-    
-    try:
-        result = subprocess.run(
-            cmd,
-            capture_output=True,
-            text=True,
-            check=False,
-        )
-        elapsed = time.time() - start_time
-        
-        if result.returncode == 0:
-            return True, "success", elapsed
-        else:
-            error_msg = result.stderr.strip() or result.stdout.strip() or "Unknown error"
-            return False, error_msg, elapsed
-            
-    except OSError as e:
-        elapsed = time.time() - start_time
-        return False, str(e), elapsed
-
-
 # =============================================================================
 # File handling functions
 # =============================================================================
 
 # Supported file extensions (based on Docling InputFormat)
 SUPPORTED_EXTENSIONS = {
-    '.pdf', '.docx', '.pptx', '.html', '.htm',
-    '.png', '.jpg', '.jpeg', '.gif', '.bmp', '.tiff', '.tif',  # images
-    '.asciidoc', '.adoc', '.asc',  # asciidoc
-    '.md', '.markdown',  # markdown
-    '.csv', '.xlsx',  # spreadsheets
-    '.xml',  # XML formats
-    '.json',  # JSON docling
-    '.mp3', '.wav', '.m4a', '.flac',  # audio
-    '.vtt',  # subtitles
+    ".pdf",
+    ".docx",
+    ".pptx",
+    ".html",
+    ".htm",
+    ".png",
+    ".jpg",
+    ".jpeg",
+    ".gif",
+    ".bmp",
+    ".tiff",
+    ".tif",  # images
+    ".asciidoc",
+    ".adoc",
+    ".asc",  # asciidoc
+    ".md",
+    ".markdown",  # markdown
+    ".csv",
+    ".xlsx",  # spreadsheets
+    ".xml",  # XML formats
+    ".json",  # JSON docling
+    ".mp3",
+    ".wav",
+    ".m4a",
+    ".flac",  # audio
+    ".vtt",  # subtitles
 }
 
 
 def get_files_to_convert(input_path: Path, mask: str, recursive: bool) -> List[Path]:
     """Get list of files to convert based on input path and options."""
     files = []
-    
+
     if input_path.is_file():
         files.append(input_path)
     elif input_path.is_dir():
@@ -369,19 +337,19 @@ def get_files_to_convert(input_path: Path, mask: str, recursive: bool) -> List[P
             files = list(input_path.rglob(mask))
         else:
             files = list(input_path.glob(mask))
-        
+
         # Filter to only files
         files = [f for f in files if f.is_file()]
     else:
         raise FileNotFoundError(f"Input path does not exist: {input_path}")
-    
+
     # Filter out hidden files and unsupported formats
     files = [
-        f for f in files
-        if not f.name.startswith('.')
-        and f.suffix.lower() in SUPPORTED_EXTENSIONS
+        f
+        for f in files
+        if not f.name.startswith(".") and f.suffix.lower() in SUPPORTED_EXTENSIONS
     ]
-    
+
     return files
 
 
@@ -414,7 +382,7 @@ def get_output_path(
             output_path = output_dir / relative_path.parent / output_name
         except ValueError:
             output_path = output_dir / output_name
-        
+
         return output_path
 
 
@@ -422,6 +390,7 @@ def get_output_path(
 # CLI argument parsing
 # =============================================================================
 
+
 def parse_args() -> argparse.Namespace:
     """Parse command line arguments."""
     parser = argparse.ArgumentParser(
@@ -436,74 +405,92 @@ Examples:
   mdify ./docs --runtime podman          Use Podman instead of Docker
 """,
     )
-    
+
     parser.add_argument(
         "input",
         type=str,
         nargs="?",
         help="Input file or directory to convert",
     )
-    
+
     parser.add_argument(
-        "-o", "--out-dir",
+        "-o",
+        "--out-dir",
         type=str,
         default="output",
         help="Output directory for converted files (default: output)",
     )
-    
+
     parser.add_argument(
-        "-g", "--glob",
+        "-g",
+        "--glob",
         type=str,
         default="*",
         help="Glob pattern for filtering files in directory (default: *)",
     )
-    
+
     parser.add_argument(
-        "-r", "--recursive",
+        "-r",
+        "--recursive",
         action="store_true",
         help="Recursively scan directories",
     )
-    
+
     parser.add_argument(
         "--flat",
         action="store_true",
         help="Disable directory structure preservation in output",
     )
-    
+
     parser.add_argument(
         "--overwrite",
         action="store_true",
         help="Overwrite existing output files",
     )
-    
+
     parser.add_argument(
-        "-q", "--quiet",
+        "-q",
+        "--quiet",
         action="store_true",
         help="Suppress progress messages",
     )
-    
+
     parser.add_argument(
-        "-m", "--mask",
+        "-m",
+        "--mask",
         action="store_true",
         help="Mask PII and sensitive content in document images",
     )
-    
+
+    parser.add_argument(
+        "--gpu",
+        action="store_true",
+        help="Use GPU-accelerated container image (docling-serve-cu126)",
+    )
+
+    parser.add_argument(
+        "--port",
+        type=int,
+        default=5001,
+        help="Port for docling-serve container (default: 5001)",
+    )
+
     # Container options
     parser.add_argument(
         "--runtime",
         type=str,
         choices=SUPPORTED_RUNTIMES,
-        default="docker",
-        help="Container runtime to use (default: docker)",
+        default=None,
+        help="Container runtime to use (auto-detects docker or podman if not specified)",
     )
-    
+
     parser.add_argument(
         "--image",
         type=str,
         default=DEFAULT_IMAGE,
         help=f"Container image to use (default: {DEFAULT_IMAGE})",
     )
-    
+
     parser.add_argument(
         "--pull",
         type=str,
@@ -511,20 +498,20 @@ Examples:
         default="missing",
         help="Image pull policy: always, missing, never (default: missing)",
     )
-    
+
     # Utility options
     parser.add_argument(
         "--check-update",
         action="store_true",
         help="Check for available updates and exit",
     )
-    
+
     parser.add_argument(
         "--version",
         action="version",
         version=f"mdify {__version__}",
     )
-    
+
     return parser.parse_args()
 
 
@@ -532,27 +519,30 @@ Examples:
 # Main entry point
 # =============================================================================
 
+
 def main() -> int:
     """Main entry point for the CLI."""
     args = parse_args()
-    
+
     # Handle --check-update flag
     if args.check_update:
         check_for_update(force=True)
         return 0
-    
+
     # Check for updates (daily, silent on errors)
     check_for_update(force=False)
-    
+
     # Validate input is provided
     if args.input is None:
         print("Error: Input file or directory is required", file=sys.stderr)
         print("Usage: mdify <input> [options]", file=sys.stderr)
         print("       mdify --help for more information", file=sys.stderr)
         return 1
-    
+
     # Detect container runtime
-    runtime = detect_runtime(args.runtime)
+    preferred = args.runtime if args.runtime else "docker"
+    explicit = args.runtime is not None
+    runtime = detect_runtime(preferred, explicit=explicit)
     if runtime is None:
         print(
             f"Error: Container runtime not found ({', '.join(SUPPORTED_RUNTIMES)})",
@@ -560,109 +550,166 @@ def main() -> int:
         )
         print("Please install Docker or Podman to use mdify.", file=sys.stderr)
         return 2
-    
+
     # Handle image pull policy
-    image = args.image
+    # Determine image based on --gpu flag
+    if args.gpu:
+        image = GPU_IMAGE
+    elif args.image:
+        image = args.image
+    else:
+        image = DEFAULT_IMAGE
+
     image_exists = check_image_exists(runtime, image)
-    
+
     if args.pull == "always" or (args.pull == "missing" and not image_exists):
         if not pull_image(runtime, image, args.quiet):
             print(f"Error: Failed to pull image: {image}", file=sys.stderr)
             return 1
     elif args.pull == "never" and not image_exists:
         print(f"Error: Image not found locally: {image}", file=sys.stderr)
-        print(f"Run with --pull=missing or pull manually: {args.runtime} pull {image}")
+        print(f"Run with --pull=missing or pull manually: {preferred} pull {image}")
         return 1
-    
+
     # Resolve paths
     input_path = Path(args.input).resolve()
     output_dir = Path(args.out_dir).resolve()
-    
+
     # Validate input
     if not input_path.exists():
         print(f"Error: Input path does not exist: {input_path}", file=sys.stderr)
         return 1
-    
+
     # Get files to convert
     try:
         files_to_convert = get_files_to_convert(input_path, args.glob, args.recursive)
     except Exception as e:
         print(f"Error: {e}", file=sys.stderr)
         return 1
-    
+
     if not files_to_convert:
         print(f"No files found to convert in: {input_path}", file=sys.stderr)
         return 1
-    
+
     total_files = len(files_to_convert)
     total_size = sum(f.stat().st_size for f in files_to_convert)
-    
+
     if not args.quiet:
         print(f"Found {total_files} file(s) to convert ({format_size(total_size)})")
         print(f"Using runtime: {runtime}")
         print(f"Using image: {image}")
         print()
-    
+
+    if args.mask:
+        print(
+            "Warning: --mask is not supported with docling-serve and will be ignored",
+            file=sys.stderr,
+        )
+
     # Determine input base for directory structure preservation
     if input_path.is_file():
         input_base = input_path.parent
     else:
         input_base = input_path
-    
-    # Convert files
+
     success_count = 0
     skipped_count = 0
     failed_count = 0
-    conversion_start = time.time()
-    spinner = Spinner()
-    
-    for idx, input_file in enumerate(files_to_convert, 1):
-        output_file = get_output_path(input_file, input_base, output_dir, args.flat)
-        file_size = input_file.stat().st_size
-        progress = f"[{idx}/{total_files}]"
-        
-        # Check if output exists and skip if not overwriting
-        if output_file.exists() and not args.overwrite:
-            if not args.quiet:
-                print(f"{progress} Skipped (exists): {input_file.name}")
-            skipped_count += 1
-            continue
-        
-        # Show spinner while processing
+    total_elapsed = 0.0
+
+    try:
         if not args.quiet:
-            spinner.start(f"{progress} Processing: {input_file.name} ({format_size(file_size)})")
-        
-        success, result, elapsed = run_container(
-            runtime, image, input_file, output_file, args.mask
-        )
-        
+            print(f"Starting docling-serve container...")
+            print()
+
+        with DoclingContainer(runtime, image, args.port) as container:
+            # Convert files
+            conversion_start = time.time()
+            spinner = Spinner()
+
+            for idx, input_file in enumerate(files_to_convert, 1):
+                output_file = get_output_path(
+                    input_file, input_base, output_dir, args.flat
+                )
+                file_size = input_file.stat().st_size
+                progress = f"[{idx}/{total_files}]"
+
+                # Check if output exists and skip if not overwriting
+                if output_file.exists() and not args.overwrite:
+                    if not args.quiet:
+                        print(f"{progress} Skipped (exists): {input_file.name}")
+                    skipped_count += 1
+                    continue
+
+                # Ensure output directory exists
+                output_file.parent.mkdir(parents=True, exist_ok=True)
+
+                # Show spinner while processing
+                if not args.quiet:
+                    spinner.start(
+                        f"{progress} Processing: {input_file.name} ({format_size(file_size)})"
+                    )
+
+                start_time = time.time()
+                try:
+                    # Convert via HTTP API
+                    result = convert_file(
+                        container.base_url, input_file, to_format="md"
+                    )
+                    elapsed = time.time() - start_time
+
+                    if not args.quiet:
+                        spinner.stop()
+
+                    if result.success:
+                        # Write result to output file
+                        output_file.write_text(result.content)
+                        success_count += 1
+                        if not args.quiet:
+                            print(
+                                f"{progress} {input_file.name} ✓ ({format_duration(elapsed)})"
+                            )
+                    else:
+                        failed_count += 1
+                        error_msg = result.error or "Unknown error"
+                        if not args.quiet:
+                            print(
+                                f"{progress} {input_file.name} ✗ ({format_duration(elapsed)})"
+                            )
+                            print(f"    Error: {error_msg}", file=sys.stderr)
+                except Exception as e:
+                    elapsed = time.time() - start_time
+                    failed_count += 1
+                    if not args.quiet:
+                        spinner.stop()
+                        print(
+                            f"{progress} {input_file.name} ✗ ({format_duration(elapsed)})"
+                        )
+                        print(f"    Error: {str(e)}", file=sys.stderr)
+
+            total_elapsed = time.time() - conversion_start
+
+        # Print summary
         if not args.quiet:
-            spinner.stop()
-        
-        if success:
-            success_count += 1
-            if not args.quiet:
-                print(f"{progress} {input_file.name} ✓ ({format_duration(elapsed)})")
-        else:
-            failed_count += 1
-            if not args.quiet:
-                print(f"{progress} {input_file.name} ✗ ({format_duration(elapsed)})")
-                print(f"    Error: {result}", file=sys.stderr)
-    
-    total_elapsed = time.time() - conversion_start
-    
-    # Print summary
-    if not args.quiet:
-        print()
-        print("=" * 50)
-        print("Conversion Summary:")
-        print(f"  Total files:     {total_files}")
-        print(f"  Successful:      {success_count}")
-        print(f"  Skipped:         {skipped_count}")
-        print(f"  Failed:          {failed_count}")
-        print(f"  Total time:      {format_duration(total_elapsed)}")
-        print("=" * 50)
-    
+            print()
+            print("=" * 50)
+            print("Conversion Summary:")
+            print(f"  Total files:     {total_files}")
+            print(f"  Successful:      {success_count}")
+            print(f"  Skipped:         {skipped_count}")
+            print(f"  Failed:          {failed_count}")
+            print(f"  Total time:      {format_duration(total_elapsed)}")
+            print("=" * 50)
+
+    except KeyboardInterrupt:
+        if not args.quiet:
+            print("\n\nInterrupted by user. Container stopped.")
+            if success_count > 0 or skipped_count > 0 or failed_count > 0:
+                print(
+                    f"Partial progress: {success_count} successful, {failed_count} failed, {skipped_count} skipped"
+                )
+        return 130
+
     # Return appropriate exit code
     if failed_count > 0:
         return 1

mdify/container.py

@@ -0,0 +1,128 @@
+"""Container lifecycle management for docling-serve."""
+
+import subprocess
+import time
+import uuid
+from typing import Optional
+
+from mdify.docling_client import check_health
+
+
+class DoclingContainer:
+    """Manages docling-serve container lifecycle.
+
+    Provides context manager support for automatic startup and cleanup.
+
+    Usage:
+        with DoclingContainer("docker", "ghcr.io/docling-project/docling-serve-cpu:main") as container:
+            # Container is running and healthy
+            response = requests.post(f"{container.base_url}/v1/convert/file", ...)
+        # Container automatically stopped and removed
+    """
+
+    def __init__(self, runtime: str, image: str, port: int = 5001):
+        """Initialize container manager.
+
+        Args:
+            runtime: Container runtime ("docker" or "podman")
+            image: Container image to use
+            port: Host port to bind (default: 5001)
+        """
+        self.runtime = runtime
+        self.image = image
+        self.port = port
+        self.container_name = f"mdify-serve-{uuid.uuid4().hex[:8]}"
+        self.container_id: Optional[str] = None
+
+    @property
+    def base_url(self) -> str:
+        """Return base URL for API requests."""
+        return f"http://localhost:{self.port}"
+
+    def start(self, timeout: int = 120) -> None:
+        """Start container and wait for health check.
+
+        Args:
+            timeout: Maximum seconds to wait for health (default: 120)
+
+        Raises:
+            subprocess.CalledProcessError: If container fails to start
+            TimeoutError: If health check doesn't pass within timeout
+        """
+        # Start container in detached mode
+        cmd = [
+            self.runtime,
+            "run",
+            "-d",  # Detached mode
+            "--rm",  # Auto-remove on stop
+            "--name",
+            self.container_name,
+            "-p",
+            f"{self.port}:5001",
+            self.image,
+        ]
+
+        try:
+            result = subprocess.run(cmd, capture_output=True, text=True, check=True)
+            self.container_id = result.stdout.strip()
+        except subprocess.CalledProcessError as e:
+            error_msg = e.stderr.strip() or e.stdout.strip() or "Unknown error"
+            raise subprocess.CalledProcessError(
+                e.returncode,
+                e.cmd,
+                output=e.stdout,
+                stderr=f"Failed to start container: {error_msg}",
+            )
+
+        # Wait for health check
+        self._wait_for_health(timeout)
+
+    def stop(self) -> None:
+        """Stop and remove container. Safe to call multiple times."""
+        if self.container_name:
+            subprocess.run(
+                [self.runtime, "stop", self.container_name],
+                capture_output=True,
+                check=False,
+            )
+
+    def is_ready(self) -> bool:
+        """Check if container is healthy.
+
+        Returns:
+            True if container is healthy, False otherwise
+        """
+        try:
+            return check_health(self.base_url)
+        except Exception:
+            return False
+
+    def _wait_for_health(self, timeout: int) -> None:
+        """Poll health endpoint until ready.
+
+        Args:
+            timeout: Maximum seconds to wait
+
+        Raises:
+            TimeoutError: If health check doesn't pass within timeout
+        """
+        start_time = time.time()
+        while time.time() - start_time < timeout:
+            try:
+                if check_health(self.base_url):
+                    return
+            except Exception:
+                pass
+            time.sleep(2)  # Poll every 2 seconds
+
+        raise TimeoutError(f"Container failed to become healthy within {timeout}s")
+
+    def __enter__(self):
+        """Context manager entry."""
+        self.start()
+        return self
+
+    def __exit__(self, exc_type, exc_val, exc_tb):
+        """Context manager exit - ensures cleanup."""
+        self.stop()
+        return False

mdify/docling_client.py

@@ -0,0 +1,224 @@
+"""HTTP client for docling-serve REST API."""
+
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Optional
+
+import requests
+
+
+@dataclass
+class ConvertResult:
+    """Result from document conversion."""
+
+    content: str
+    format: str
+    success: bool
+    error: Optional[str] = None
+
+
+@dataclass
+class StatusResult:
+    """Status of async conversion task."""
+
+    status: str  # "pending", "completed", "failed"
+    task_id: str
+    error: Optional[str] = None
+
+
+class DoclingClientError(Exception):
+    """Base exception for docling client errors."""
+
+    pass
+
+
+class DoclingHTTPError(DoclingClientError):
+    """HTTP error from docling-serve API."""
+
+    def __init__(self, status_code: int, message: str):
+        self.status_code = status_code
+        super().__init__(f"HTTP {status_code}: {message}")
+
+
+def check_health(base_url: str) -> bool:
+    """Check if docling-serve is healthy.
+
+    Args:
+        base_url: Base URL of docling-serve (e.g., "http://localhost:8000")
+
+    Returns:
+        True if healthy, False otherwise
+    """
+    try:
+        response = requests.get(f"{base_url}/health")
+        return response.status_code == 200
+    except requests.RequestException:
+        return False
+
+
+def convert_file(
+    base_url: str, file_path: Path, to_format: str = "md", do_ocr: bool = True
+) -> ConvertResult:
+    """Convert a file synchronously.
+
+    Args:
+        base_url: Base URL of docling-serve
+        file_path: Path to file to convert
+        to_format: Output format (default: "md")
+        do_ocr: Whether to perform OCR (default: True)
+
+    Returns:
+        ConvertResult with conversion output
+
+    Raises:
+        DoclingHTTPError: If HTTP request fails
+    """
+    try:
+        with open(file_path, "rb") as f:
+            response = requests.post(
+                f"{base_url}/v1/convert/file",
+                files={"files": (file_path.name, f, "application/pdf")},
+                data={"to_formats": to_format, "do_ocr": str(do_ocr).lower()},
+            )
+
+        if response.status_code != 200:
+            raise DoclingHTTPError(
+                response.status_code, response.text or "Conversion failed"
+            )
+
+        result_data = response.json()
+
+        # docling-serve returns results in a list format
+        if isinstance(result_data, list) and len(result_data) > 0:
+            first_result = result_data[0]
+            return ConvertResult(
+                content=first_result.get("content", ""), format=to_format, success=True
+            )
+        elif isinstance(result_data, dict):
+            return ConvertResult(
+                content=result_data.get("content", ""), format=to_format, success=True
+            )
+        else:
+            raise DoclingHTTPError(200, f"Unexpected response format: {result_data}")
+
+    except requests.RequestException as e:
+        return ConvertResult(content="", format=to_format, success=False, error=str(e))
+
+
+def convert_file_async(
+    base_url: str, file_path: Path, to_format: str = "md", do_ocr: bool = True
+) -> str:
+    """Start async file conversion.
+
+    Args:
+        base_url: Base URL of docling-serve
+        file_path: Path to file to convert
+        to_format: Output format (default: "md")
+        do_ocr: Whether to perform OCR (default: True)
+
+    Returns:
+        Task ID for polling
+
+    Raises:
+        DoclingHTTPError: If HTTP request fails
+    """
+    try:
+        with open(file_path, "rb") as f:
+            response = requests.post(
+                f"{base_url}/v1/convert/file/async",
+                files={"files": (file_path.name, f, "application/pdf")},
+                data={"to_formats": to_format, "do_ocr": str(do_ocr).lower()},
+            )
+
+        if response.status_code != 200:
+            raise DoclingHTTPError(
+                response.status_code, response.text or "Async conversion failed"
+            )
+
+        result_data = response.json()
+        task_id = result_data.get("task_id")
+
+        if not task_id:
+            raise DoclingHTTPError(200, f"No task_id in response: {result_data}")
+
+        return task_id
+
+    except requests.RequestException as e:
+        raise DoclingHTTPError(500, str(e))
+
+
+def poll_status(base_url: str, task_id: str) -> StatusResult:
+    """Poll status of async conversion task.
+
+    Args:
+        base_url: Base URL of docling-serve
+        task_id: Task ID from convert_file_async
+
+    Returns:
+        StatusResult with current status
+
+    Raises:
+        DoclingHTTPError: If HTTP request fails
+    """
+    try:
+        response = requests.get(f"{base_url}/v1/status/poll/{task_id}")
+
+        if response.status_code != 200:
+            raise DoclingHTTPError(
+                response.status_code, response.text or "Status poll failed"
+            )
+
+        result_data = response.json()
+
+        return StatusResult(
+            status=result_data.get("status", "unknown"),
+            task_id=task_id,
+            error=result_data.get("error"),
+        )
+
+    except requests.RequestException as e:
+        raise DoclingHTTPError(500, str(e))
+
+
+def get_result(base_url: str, task_id: str) -> ConvertResult:
+    """Get result of completed async conversion.
+
+    Args:
+        base_url: Base URL of docling-serve
+        task_id: Task ID from convert_file_async
+
+    Returns:
+        ConvertResult with conversion output
+
+    Raises:
+        DoclingHTTPError: If HTTP request fails or task not completed
+    """
+    try:
+        response = requests.get(f"{base_url}/v1/result/{task_id}")
+
+        if response.status_code != 200:
+            raise DoclingHTTPError(
+                response.status_code, response.text or "Result retrieval failed"
+            )
+
+        result_data = response.json()
+
+        # Similar to sync conversion, handle list or dict format
+        if isinstance(result_data, list) and len(result_data) > 0:
+            first_result = result_data[0]
+            return ConvertResult(
+                content=first_result.get("content", ""),
+                format=first_result.get("format", "md"),
+                success=True,
+            )
+        elif isinstance(result_data, dict):
+            return ConvertResult(
+                content=result_data.get("content", ""),
+                format=result_data.get("format", "md"),
+                success=True,
+            )
+        else:
+            raise DoclingHTTPError(200, f"Unexpected response format: {result_data}")
+
+    except requests.RequestException as e:
+        return ConvertResult(content="", format="md", success=False, error=str(e))

{mdify_cli-1.5.0.dist-info → mdify_cli-2.0.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mdify-cli
-Version: 1.5.0
+Version: 2.0.0
 Summary: Convert PDFs and document images into structured Markdown for LLM workflows
 Author: tiroq
 License-Expression: MIT
@@ -24,6 +24,9 @@ Classifier: Topic :: Utilities
 Requires-Python: >=3.8
 Description-Content-Type: text/markdown
 License-File: LICENSE
+Requires-Dist: requests
+Provides-Extra: dev
+Requires-Dist: pytest>=7.0; extra == "dev"
 Dynamic: license-file
 
 # mdify
@@ -98,15 +101,32 @@ Recursively convert files:
 mdify /path/to/documents -r -g "*.pdf"
 ```
 
-### Masking sensitive content
+### GPU Acceleration
 
-Mask PII and sensitive content in images:
+For faster processing with NVIDIA GPU:
 ```bash
-mdify document.pdf -m
-mdify document.pdf --mask
+mdify --gpu documents/*.pdf
 ```
 
-This uses Docling's content-aware masking to obscure sensitive information in embedded images.
+Requires NVIDIA GPU with CUDA support and nvidia-container-toolkit.
+
+### ⚠️ PII Masking (Deprecated)
+
+The `--mask` flag is deprecated and will be ignored in this version. PII masking functionality was available in older versions using a custom runtime but is not supported with the current docling-serve backend.
+
+If PII masking is critical for your use case, please use mdify v1.5.x or earlier versions.
+
+## Performance
+
+mdify now uses docling-serve for significantly faster batch processing:
+
+- **Single model load**: Models are loaded once per session, not per file
+- **~10-20x speedup** for multiple file conversions compared to previous versions
+- **GPU acceleration**: Use `--gpu` for additional 2-6x speedup (requires NVIDIA GPU)
+
+### First Run Behavior
+
+The first conversion takes longer (~30-60s) as the container loads ML models into memory. Subsequent files in the same batch process quickly, typically in 1-3 seconds per file.
 
 ## Options
 
@@ -119,9 +139,11 @@ This uses Docling's content-aware masking to obscure sensitive information in em
 | `--flat` | Disable directory structure preservation |
 | `--overwrite` | Overwrite existing output files |
 | `-q, --quiet` | Suppress progress messages |
-| `-m, --mask` | Mask PII and sensitive content in images |
+| `-m, --mask` | ⚠️ **Deprecated**: PII masking not supported in current version |
+| `--gpu` | Use GPU-accelerated container (requires NVIDIA GPU and nvidia-container-toolkit) |
+| `--port PORT` | Container port (default: 5001) |
 | `--runtime RUNTIME` | Container runtime: docker or podman (auto-detected) |
-| `--image IMAGE` | Custom container image (default: ghcr.io/tiroq/mdify-runtime:latest) |
+| `--image IMAGE` | Custom container image (default: ghcr.io/docling-project/docling-serve-cpu:main) |
 | `--pull POLICY` | Image pull policy: always, missing, never (default: missing) |
 | `--check-update` | Check for available updates and exit |
 | `--version` | Show version and exit |
@@ -175,19 +197,22 @@ The CLI:
 - Pulls the runtime container on first use
 - Mounts files and runs conversions in the container
 
-## Container Image
+## Container Images
+
+mdify uses official docling-serve containers:
 
-The runtime container is hosted at:
+**CPU Version** (default):
 ```
-ghcr.io/tiroq/mdify-runtime:latest
+ghcr.io/docling-project/docling-serve-cpu:main
 ```
 
-To build locally:
-```bash
-cd runtime
-docker build -t mdify-runtime .
+**GPU Version** (use with `--gpu` flag):
+```
+ghcr.io/docling-project/docling-serve-cu126:main
 ```
 
+These are official images from the [docling-serve project](https://github.com/DS4SD/docling-serve).
+
 ## Updates
 
 mdify checks for updates daily. When a new version is available:

mdify_cli-2.0.0.dist-info/RECORD

@@ -0,0 +1,12 @@
+assets/mdify.png,sha256=qUj7WXWqNwpI2KNXOW79XJwqFqa-UI0JEkmt1mmy4Rg,1820418
+mdify/__init__.py,sha256=s7XlWmH4zJ5jFiPjpd7mXrCaU8bD-S9RaPzT2VHUdeQ,90
+mdify/__main__.py,sha256=bhpJ00co6MfaVOdH4XLoW04NtLYDa_oJK7ODzfLrn9M,143
+mdify/cli.py,sha256=LY0q8NlnKuN5aFz_OpO5hGro-tQNCxoYO_M0qVd6FJY,21493
+mdify/container.py,sha256=AVIhiq_wO5id5hQ_s83lUPkAPCsAoTs25azRT6JmKII,3962
+mdify/docling_client.py,sha256=_9qjL5yOOeJahOg6an2P6Iii1xkeR6wmNJZG4Q6NRkk,6553
+mdify_cli-2.0.0.dist-info/licenses/LICENSE,sha256=NWM66Uv-XuSMKaU-gaPmvfyk4WgE6zcIPr78wyg6GAo,1065
+mdify_cli-2.0.0.dist-info/METADATA,sha256=92_uBI2nnKK-YEf39TB7gX1KHbZBHqIHxLZBe7-GOqY,7923
+mdify_cli-2.0.0.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+mdify_cli-2.0.0.dist-info/entry_points.txt,sha256=0Xki8f5lADQUtwdt6Eq_FEaieI6Byhk8UE7BuDhChMg,41
+mdify_cli-2.0.0.dist-info/top_level.txt,sha256=qltzf7h8owHq7dxCdfCkSHY8gT21hn1_E8P-VWS_OKM,6
+mdify_cli-2.0.0.dist-info/RECORD,,

{mdify_cli-1.5.0.dist-info → mdify_cli-2.0.0.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.9.0)
+Generator: setuptools (80.10.2)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

mdify_cli-1.5.0.dist-info/RECORD

@@ -1,10 +0,0 @@
-assets/mdify.png,sha256=qUj7WXWqNwpI2KNXOW79XJwqFqa-UI0JEkmt1mmy4Rg,1820418
-mdify/__init__.py,sha256=GxfVEOJLubSaiA0jNE2zgZq7sxiJMAr6Qn-cLPK8XJU,90
-mdify/__main__.py,sha256=bhpJ00co6MfaVOdH4XLoW04NtLYDa_oJK7ODzfLrn9M,143
-mdify/cli.py,sha256=D8_1_6NgWXkexGWqkgB0JO7c1r2T2_Va7J7iGwvewQA,20038
-mdify_cli-1.5.0.dist-info/licenses/LICENSE,sha256=NWM66Uv-XuSMKaU-gaPmvfyk4WgE6zcIPr78wyg6GAo,1065
-mdify_cli-1.5.0.dist-info/METADATA,sha256=LRKZupINA7w6HM9FyuDdmrLzWYovHqxTnQRHNohmRM0,6658
-mdify_cli-1.5.0.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-mdify_cli-1.5.0.dist-info/entry_points.txt,sha256=0Xki8f5lADQUtwdt6Eq_FEaieI6Byhk8UE7BuDhChMg,41
-mdify_cli-1.5.0.dist-info/top_level.txt,sha256=qltzf7h8owHq7dxCdfCkSHY8gT21hn1_E8P-VWS_OKM,6
-mdify_cli-1.5.0.dist-info/RECORD,,