docling-glm-ocr 0.2.0

A docling OCR plugin for GLM-OCR

docling-glm-ocr

A docling OCR plugin that delegates text recognition to a remote GLM-OCR model served by vLLM.

---

GitHub  |  PyPI

---

Overview

docling-glm-ocr is a docling plugin that replaces the built-in OCR stage with a call to a remote GLM-OCR model hosted on a vLLM server.

Each page crop is sent to the vLLM OpenAI-compatible chat completion endpoint as a base64-encoded image. The model returns Markdown-formatted text which docling merges back into the document structure.

The plugin registers itself under the "glm-ocr-remote" OCR engine key so it can be selected per-request through docling or docling-serve without changing application code.

Requirements

• Python 3.13+
• A running vLLM server hosting zai-org/GLM-OCR (or any compatible model)

Installation

# with uv (recommended)
uv add docling-glm-ocr

# with pip
pip install docling-glm-ocr

Usage

Python SDK

from docling.datamodel.base_models import InputFormat
from docling.datamodel.pipeline_options import PdfPipelineOptions
from docling.document_converter import DocumentConverter, PdfFormatOption

from docling_glm_ocr import GlmOcrRemoteOptions

pipeline_options = PdfPipelineOptions(
    allow_external_plugins=True,
    ocr_options=GlmOcrRemoteOptions(
        api_url="http://localhost:8001/v1/chat/completions",
        model_name="zai-org/GLM-OCR",
    ),
)

converter = DocumentConverter(
    format_options={
        InputFormat.PDF: PdfFormatOption(pipeline_options=pipeline_options)
    }
)
result = converter.convert("document.pdf")
print(result.document.export_to_markdown())

docling-serve

Select the engine per-request via the standard API:

curl -X POST http://localhost:5001/v1/convert/source \
  -H 'Content-Type: application/json' \
  -d '{
    "options": {
      "ocr_engine": "glm-ocr-remote"
    },
    "sources": [{"kind": "http", "url": "https://arxiv.org/pdf/2501.17887"}]
  }'

The server must have DOCLING_SERVE_ALLOW_EXTERNAL_PLUGINS=true set so the plugin is loaded automatically.

Configuration

Environment variables

Variable	Description	Default
GLMOCR_REMOTE_OCR_API_URL	vLLM chat completion URL	http://localhost:8001/v1/chat/completions
GLMOCR_REMOTE_OCR_PROMPT	Text prompt sent with each image crop	see below

GlmOcrRemoteOptions

All options can be set programmatically via GlmOcrRemoteOptions:

Option	Type	Description	Default
api_url	str	OpenAI-compatible chat completion URL	GLMOCR_REMOTE_OCR_API_URL env or http://localhost:8001/v1/chat/completions
model_name	str	Model name sent to vLLM	zai-org/GLM-OCR
prompt	str	Text prompt for each image crop	GLMOCR_REMOTE_OCR_PROMPT env or default prompt
timeout	float	HTTP timeout per crop (seconds)	120
max_tokens	int	Max tokens per completion	16384
scale	float	Image crop rendering scale	3.0
max_concurrent_requests	int	Max concurrent API requests	10
max_retries	int	Max retry attempts for HTTP errors	3
retry_backoff_factor	float	Exponential backoff factor for retries	2.0
lang	list[str]	Language hint (passed to docling)	["en"]

Default prompt:

Recognize the text in the image and output in Markdown format.
Preserve the original layout (headings/paragraphs/tables/formulas).
Do not fabricate content that does not exist in the image.

Architecture

flowchart LR
    subgraph docling
        Pipeline --> GlmOcrRemoteModel
    end

    subgraph vLLM
        GLMOCR["zai-org/GLM-OCR"]
    end

    GlmOcrRemoteModel -- "POST /v1/chat/completions\n(base64 image)" --> GLMOCR
    GLMOCR -- "Markdown text" --> GlmOcrRemoteModel

For each page the model:

1. Collects OCR regions from the docling layout analysis
2. Renders each region using the page backend (scale configurable, default 3×)
3. Encodes the crop as a base64 PNG data URI
4. POSTs concurrent chat completion requests to the vLLM endpoint (with retry logic)
5. Returns the recognised text as TextCell objects for docling to merge

Starting a GLM-OCR vLLM server

docker run -d \
  --rm --name ocr-glm \
  --gpus device=0 \
  --ipc=host \
  -p 8001:8000 \
  -v "${HOME}/.cache/huggingface:/root/.cache/huggingface" \
  -e "HF_TOKEN=${HF_TOKEN}" \
  --entrypoint /bin/bash \
  vllm/vllm-openai:latest \
  -c "uv pip install --system --upgrade transformers && \
      exec vllm serve zai-org/GLM-OCR \
        --served-model-name zai-org/GLM-OCR \
        --port 8000 \
        --trust-remote-code"

The plugin will connect to http://localhost:8001/v1/chat/completions by default.

Development

Setup

git clone https://github.com/DCC-BS/docling-glm-ocr.git
cd docling-glm-ocr
make install

Available commands

make install     Install dependencies and pre-commit hooks
make check       Run all quality checks (ruff lint, format, ty type check)
make test        Run tests with coverage report
make build       Build distribution packages
make publish     Publish to PyPI

Running tests

make test

Tests are in tests/ and use pytest. Coverage reports are generated at coverage.xml and printed to the terminal.

End-to-end tests

The e2e tests hit a real vLLM server and are skipped by default. To run them, set the server URL and use the e2e marker:

GLMOCR_REMOTE_OCR_API_URL=http://localhost:8001/v1/chat/completions pytest -m e2e

Code quality

This project uses:

• ruff – linting and formatting
• ty – type checking
• pre-commit – pre-commit hooks

Run all checks:

make check

Releasing

Releases are published to PyPI automatically. Update the version in pyproject.toml, then trigger the Publish workflow from GitHub Actions:

GitHub → Actions → Publish to PyPI → Run workflow

The workflow tags the commit, builds the package, and publishes to PyPI via trusted publishing.

License

MIT © Data Competence Center Basel-Stadt