feat: picture description via langchain llm

langchain_docling/_plugins.py

@@ -0,0 +1,12 @@
+"""Register Docling plugins."""
+
+
+def picture_description():
+    """Picture description plugins."""
+    from langchain_docling.picture_description import PictureDescriptionLangChainModel
+
+    return {
+        "picture_description": [
+            PictureDescriptionLangChainModel,
+        ]
+    }

langchain_docling/picture_description.py

@@ -0,0 +1,88 @@
+"""Picture description model using LangChain primitives."""
+
+import base64
+import io
+from collections.abc import Iterable
+from pathlib import Path
+from typing import ClassVar, Literal, Optional, Type, Union
+
+from docling.datamodel.accelerator_options import AcceleratorOptions
+from docling.datamodel.pipeline_options import PictureDescriptionBaseOptions
+from docling.models.picture_description_base_model import PictureDescriptionBaseModel
+from docling.models.utils.hf_model_download import HuggingFaceModelDownloadMixin
+from langchain_core.language_models.chat_models import BaseChatModel
+from PIL import Image
+
+
+class PictureDescriptionLangChainOptions(PictureDescriptionBaseOptions):
+    """Options for the PictureDescriptionLangChainModel."""
+
+    kind: ClassVar[Literal["langchain"]] = "langchain"
+    llm: BaseChatModel
+    prompt: str = "Describe this document picture in a few sentences."
+    provenance: Optional[str] = None
+
+
+class PictureDescriptionLangChainModel(
+    PictureDescriptionBaseModel, HuggingFaceModelDownloadMixin
+):
+    """Implementation of a PictureDescription model using LangChain."""
+
+    @classmethod
+    def get_options_type(cls) -> Type[PictureDescriptionBaseOptions]:
+        """Define the option type for the factory."""
+        return PictureDescriptionLangChainOptions
+
+    def __init__(
+        self,
+        enabled: bool,
+        enable_remote_services: bool,
+        artifacts_path: Optional[Union[Path, str]],
+        options: PictureDescriptionLangChainOptions,
+        accelerator_options: AcceleratorOptions,
+    ):
+        """Initialize PictureDescriptionLangChainModel."""
+        super().__init__(
+            enabled=enabled,
+            enable_remote_services=enable_remote_services,
+            artifacts_path=artifacts_path,
+            options=options,
+            accelerator_options=accelerator_options,
+        )
+        self.options: PictureDescriptionLangChainOptions
+
+        if self.enabled:
+            self.llm = self.options.llm
+            self.provenance = "langchain"
+            if self.options.provenance:
+                self.provenance += f"-{self.options.provenance}"
+
+    def _annotate_images(self, images: Iterable[Image.Image]) -> Iterable[str]:
+        """Annotate the images with the LangChain model."""
+        # Create input messages
+        batch_messages = []
+
+        for image in images:
+            buffered = io.BytesIO()
+            image.save(buffered, format="PNG")
+            image_data = base64.b64encode(buffered.getvalue()).decode("utf-8")
+            batch_messages.append(
+                [
+                    {
+                        "role": "user",
+                        "content": [
+                            {"type": "text", "text": self.options.prompt},
+                            {
+                                "type": "image_url",
+                                "image_url": {
+                                    "url": f"data:image/png;base64,{image_data}"
+                                },
+                            },
+                        ],
+                    }
+                ]
+            )
+
+        responses = self.llm.batch(batch_messages)
+        for resp in responses:
+            yield resp.text()

pyproject.toml

@@ -63,6 +63,7 @@ dev = [
     "pytest~=8.3",
     "pytest-cov>=6.1.1",
     "python-semantic-release~=7.32",
+    "langchain-openai>=0.2.12",
 ]
 
 [tool.uv]
@@ -72,6 +73,9 @@ default-groups = "all"
 [tool.setuptools.packages.find]
 include = ["langchain_docling*"]
 
+[project.entry-points."docling"]
+langchain_docling = "langchain_docling._plugins"
+
 [tool.black]
 line-length = 88
 target-version = ["py39", "py310"]