feat: Add TIFF support to document and image generate routes

vlmrun/cli/_cli/chat.py

@@ -598,7 +598,9 @@ def chat(
         response_content = ""
         usage_data: Optional[Dict[str, Any]] = None
         response_id: Optional[str] = None
-        extra_body: Optional[Dict[str, Any]] = {"session_id": session_id} if session_id else None
+        extra_body: Optional[Dict[str, Any]] = (
+            {"session_id": session_id} if session_id else None
+        )
 
         start_time = time.time()
 
@@ -613,7 +615,7 @@ def chat(
                     ),
                     handle_api_errors(),
                 ):
-                    
+
                     response = client.agent.completions.create(
                         model=model,
                         messages=messages,

vlmrun/cli/_cli/generate.py

@@ -1,6 +1,7 @@
 """Generation API commands."""
 
 from pathlib import Path
+from typing import List
 
 import typer
 from PIL import Image
@@ -9,6 +10,28 @@
 from vlmrun.client import VLMRun
 from vlmrun.client.types import PredictionResponse
 from vlmrun.common.image import _open_image_with_exif
+from vlmrun.common.tiff import tiff_images
+
+
+def _is_tiff_file(path: Path) -> bool:
+    """Check if a file is a TIFF file."""
+    return path.suffix.lower() in (".tif", ".tiff")
+
+
+def _load_images_from_file(path: Path) -> List[Image.Image]:
+    """Load images from a file, handling multi-page TIFFs.
+
+    Args:
+        path: Path to the image file
+
+    Returns:
+        List of PIL Image objects
+    """
+    if _is_tiff_file(path):
+        return [page.image for page in tiff_images(path)]
+    else:
+        return [_open_image_with_exif(path)]
+
 
 app = typer.Typer(help="Generation operations", no_args_is_help=True)
 
@@ -28,8 +51,8 @@ def image(
     if not Path(image).is_file():
         raise typer.Abort(f"Image file does not exist: {image}")
 
-    img: Image.Image = _open_image_with_exif(image)
-    response: PredictionResponse = client.image.generate(images=[img], domain=domain)
+    images: List[Image.Image] = _load_images_from_file(image)
+    response: PredictionResponse = client.image.generate(images=images, domain=domain)
     rprint(response)
 
 
@@ -48,5 +71,11 @@ def document(
     if not Path(path).is_file():
         raise typer.Abort(f"Document file does not exist: {path}")
 
-    response = client.document.generate(file=path, domain=domain)
+    if _is_tiff_file(path):
+        images: List[Image.Image] = [page.image for page in tiff_images(path)]
+        response: PredictionResponse = client.image.generate(
+            images=images, domain=domain
+        )
+    else:
+        response = client.document.generate(file=path, domain=domain)
     rprint(response)

vlmrun/common/tiff.py

@@ -0,0 +1,49 @@
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Iterable, Union
+
+from PIL import Image
+
+from vlmrun.common.logging import logger
+
+
+@dataclass
+class TiffPage:
+    image: Image.Image
+    """The image of the page."""
+    page_number: int
+    """The page number of the page."""
+
+
+def tiff_images(
+    path: Path,
+) -> Iterable[Union[Image.Image, TiffPage]]:
+    """Extract images from a multi-page TIFF file.
+
+    Args:
+        path: Path to the TIFF file
+
+    Yields:
+        TiffPage objects containing the image and page number
+    """
+    if path.suffix.lower() not in (".tif", ".tiff"):
+        raise ValueError(
+            f"Unsupported file type: {path.suffix}. Supported types are .tif, .tiff"
+        )
+
+    logger.debug(
+        f"Extracting images from TIFF [path={path}, size={path.stat().st_size / 1024 / 1024:.2f} MB]"
+    )
+
+    img = Image.open(path)
+    n_frames = getattr(img, "n_frames", 1)
+
+    logger.debug(f"TIFF has {n_frames} frame(s)")
+
+    for i in range(n_frames):
+        img.seek(i)
+        frame = img.copy().convert("RGB")
+        yield TiffPage(image=frame, page_number=i)
+
+    img.close()
+    logger.debug(f"Closed TIFF file [path={path}]")

vlmrun/constants.py

@@ -18,8 +18,17 @@
 VLMRUN_TMP_DIR.mkdir(parents=True, exist_ok=True)
 
 SUPPORTED_VIDEO_FILETYPES = [".mp4", ".mov", ".avi", ".mkv", ".webm"]
-SUPPORTED_IMAGE_FILETYPES = [".jpg", ".jpeg", ".png", ".gif", ".bmp", ".tiff", ".webp"]
-SUPPORTED_DOCUMENT_FILETYPES = [".pdf", ".doc", ".docx"]
+SUPPORTED_IMAGE_FILETYPES = [
+    ".jpg",
+    ".jpeg",
+    ".png",
+    ".gif",
+    ".bmp",
+    ".tif",
+    ".tiff",
+    ".webp",
+]
+SUPPORTED_DOCUMENT_FILETYPES = [".pdf", ".doc", ".docx", ".tif", ".tiff"]
 SUPPORTED_AUDIO_FILETYPES = [".mp3", ".wav", ".m4a", ".flac", ".ogg"]
 
 # All supported file types for the chat CLI

vlmrun/version.py

@@ -1 +1 @@
-__version__ = "0.5.4"
+__version__ = "0.17.18"