Description

Granite Docling is a family of instruction-tuned models designed for document understanding tasks. These models are fine-tuned on a diverse set of tasks including document classification, information extraction, and question answering. The models are optimized for performance on document-centric tasks and can handle a variety of document formats and layouts.

Usage

PDF Conversion

#!/usr/bin/env -S uv run --script
# /// script
# requires-python = ">=3.12"
# dependencies = ["docling>=2.60.0", "requests>=2.32.5"]
# ///
import tempfile

import requests
from docling.datamodel.base_models import InputFormat
from docling.datamodel.pipeline_options import VlmPipelineOptions
from docling.datamodel.pipeline_options_vlm_model import ApiVlmOptions, ResponseFormat
from docling.document_converter import DocumentConverter, PdfFormatOption
from docling.pipeline.vlm_pipeline import VlmPipeline
from pydantic import AnyUrl

pdf_url = "https://arxiv.org/pdf/1706.03762.pdf"
with tempfile.NamedTemporaryFile(suffix=".pdf") as f:
    response = requests.get(pdf_url)
    f.write(response.content)
    f.flush()

    pipeline_options = VlmPipelineOptions(
        enable_remote_services=True,
        vlm_options=ApiVlmOptions(
            url=AnyUrl("http://localhost:11434/v1/chat/completions"),
            params={"model": "danchev/granite-docling"},
            prompt="Convert this page to docling.",
            temperature=0.0,
            response_format=ResponseFormat.DOCTAGS,
        ),
    )

    doc_converter = DocumentConverter(
        format_options={
            InputFormat.PDF: PdfFormatOption(
                pipeline_options=pipeline_options, pipeline_cls=VlmPipeline
            )
        }
    )

    markdown = doc_converter.convert(f.name).document.export_to_markdown()
    print(markdown)

PNG Conversion

#!/usr/bin/env -S uv run --script
# /// script
# requires-python = ">=3.12"
# dependencies = ["docling-core>=2.50.0", "Pillow>=11.3.0", "requests>=2.32.5"]
# ///
import base64
import io

import requests
from docling_core.types.doc.document import DoclingDocument, DocTagsDocument
from PIL import Image
from requests import Response

img_url = "https://ibm.biz/docling-page-with-list"
response: Response = requests.get(img_url)
png_bytes: bytes = response.content

response = requests.post(
    url="http://localhost:11434/api/chat",
    json={
        "messages": [
            {
                "role": "user",
                "content": "Convert this image to docling.",
                "images": [base64.b64encode(png_bytes).decode("utf-8")],
            }
        ],
        "model": "danchev/granite-docling",
        "stream": False,
    },
)

doctags: str = response.json()["message"]["content"]

doc: DoclingDocument = DoclingDocument.load_from_doctags(
    doctag_document=DocTagsDocument.from_doctags_and_image_pairs(
        doctags=[doctags], images=[Image.open(io.BytesIO(png_bytes))]
    ),
)

markdown: str = doc.export_to_markdown()
print(markdown)

Reference:

Hugging Face