Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.
Python ai4sciencedocument-analysisdocxextract-datalayout-analysisocrparserpdfpdf-converterpdf-extractor-llmpdf-extractor-pretrainpdf-extractor-ragpdf-parserpptxpythonxlsx
# Use the official vllm image for gpu with Volta、Turing、Ampere、Ada Lovelace、Hopper、Blackwell architecture (7.0 <= Compute Capability <= 12.1)
# The default base image uses vLLM 0.21.0 with CUDA 13.0. For CUDA 12.9 environments, switch to the commented cu129 image below.
# Compute Capability version query (https://developer.nvidia.com/cuda-gpus)
# support x86_64 architecture and ARM(AArch64) architecture
FROM vllm/vllm-openai:v0.21.0
# FROM vllm/vllm-openai:v0.21.0-cu129
# Install libgl for opencv support
RUN apt-get update && \
apt-get install -y \
libgl1 && \
apt-get clean && \
rm -rf /var/lib/apt/lists/*
# Install MinerU 4 with Torch; reuse vLLM from the base image.
RUN python3 -m pip install -U 'mineru[torch]>=4.0,<5' --break-system-packages && \
python3 -m pip cache purge
# Keep build-time model downloads and runtime selection aligned, including builds without a GPU.
ENV MINERU_MODEL_SMALL_BACKEND=torch \
MINERU_MODEL_VLM_ENGINE=vllm
# Download the Torch bundle and original VLM weights.
RUN /bin/bash -c "mineru-kit models download --tier standard --small-backend torch --vlm-engine vllm -s huggingface"
# Set the entry point to activate the virtual environment and run the command line tool
ENTRYPOINT ["/bin/bash", "-c", "export MINERU_MODEL_SOURCE=local && exec \"$@\"", "--"]