📝 技術調査 - markitdown

pip install 'markitdown[all]'

pip install 'markitdown[pdf]'
pip install 'markitdown[docx]'
pip install 'markitdown[xlsx]'
pip install 'markitdown[pptx]'
pip install 'markitdown[audio-transcription]'
pip install 'markitdown[youtube-transcription]'
pip install 'markitdown[outlook]'
pip install 'markitdown[az-doc-intel]'
pip install 'markitdown[pdf,docx,pptx]'

uv venv --python=3.12 .venv
source .venv/bin/activate
uv pip install 'markitdown[all]'

pipx install 'markitdown[all]'

git clone git@github.com:microsoft/markitdown.git
cd markitdown
pip install -e 'packages/markitdown[all]'

docker build -t markitdown:latest .
docker run --rm -i markitdown:latest < ~/your-file.pdf > output.md

markitdown --version
python -c "import markitdown; print(markitdown.__version__)"

markitdown path-to-file.pdf > document.md
markitdown path-to-file.pdf -o document.md
cat path-to-file.pdf | markitdown

markitdown --list-plugins
markitdown --use-plugins path-to-file.pdf

markitdown path-to-file.pdf -o document.md -d -e "<document_intelligence_endpoint>"

from markitdown import MarkItDown

md = MarkItDown(enable_plugins=False)
result = md.convert("test.xlsx")
print(result.markdown)
print(result.text_content)
if result.title:
    print(result.title)

from markitdown import MarkItDown

md = MarkItDown()
with open("test.pdf", "rb") as f:
    result = md.convert_stream(f)
    print(result.markdown)

from markitdown import MarkItDown
from openai import OpenAI

md = MarkItDown(
    llm_client=OpenAI(),
    llm_model="gpt-4o",
    llm_prompt="optional custom prompt",
)
result = md.convert("example.jpg")
print(result.markdown)

from markitdown import MarkItDown

md = MarkItDown(docintel_endpoint="<document_intelligence_endpoint>")
result = md.convert("test.pdf")
print(result.markdown)

from markitdown import MarkItDown
from openai import OpenAI

md = MarkItDown(
    enable_plugins=True,
    llm_client=OpenAI(),
    llm_model="gpt-4o",
)
result = md.convert("document_with_images.pdf")

pip install markitdown-mcp

markitdown-mcp

markitdown-mcp --http --host 127.0.0.1 --port 3001

docker build -t markitdown-mcp:latest .
docker run -it --rm markitdown-mcp:latest
docker run -it --rm -v /home/user/data:/workdir markitdown-mcp:latest

{
  "mcpServers": {
    "markitdown": {
      "command": "markitdown-mcp"
    }
  }
}

{
  "mcpServers": {
    "markitdown": {
      "command": "docker",
      "args": ["run", "--rm", "-i", "-v", "/home/user/data:/workdir", "markitdown-mcp:latest"]
    }
  }
}

docker build -t markitdown:latest .
docker run --rm -i markitdown:latest < file.pdf > output.md
docker run --rm -v /path/to/files:/workdir markitdown:latest markitdown /workdir/file.pdf -o /workdir/output.md

import logging
logging.basicConfig(level=logging.DEBUG)

from markitdown import MarkItDown
md = MarkItDown()
result = md.convert("file.pdf")

from markitdown import MarkItDown, StreamInfo
import io

md = MarkItDown()
with open("file.pdf", "rb") as f:
    byte_stream = io.BytesIO(f.read())

stream_info = StreamInfo(
    mimetype="application/pdf",
    extension=".pdf",
    charset="utf-8",
)
result = md.convert_stream(byte_stream, stream_info=stream_info)
print(result.markdown)

from concurrent.futures import ThreadPoolExecutor
from markitdown import MarkItDown

md = MarkItDown(enable_plugins=False)

def convert_file(path):
    try:
        return path, md.convert(path).markdown
    except Exception as e:
        return path, f"ERROR: {e}"

files = ["a.pdf", "b.docx", "c.xlsx"]
with ThreadPoolExecutor(max_workers=4) as executor:
    results = list(executor.map(convert_file, files))

import zipfile

MAX_EXTRACTED_SIZE = 500 * 1024 * 1024

def safe_zip_check(zip_path):
    with zipfile.ZipFile(zip_path) as zf:
        total = sum(info.file_size for info in zf.infolist())
        if total > MAX_EXTRACTED_SIZE:
            raise ValueError(f"ZIP bomb suspected: {total} bytes")
        for info in zf.infolist():
            if ".." in info.filename or info.filename.startswith("/"):
                raise ValueError(f"Path traversal detected: {info.filename}")

from markitdown import MarkItDown
from langchain.text_splitter import MarkdownHeaderTextSplitter

md = MarkItDown()
result = md.convert("document.pdf")

headers_to_split_on = [
    ("#", "H1"),
    ("##", "H2"),
    ("###", "H3"),
]
splitter = MarkdownHeaderTextSplitter(headers_to_split_on=headers_to_split_on)
chunks = splitter.split_text(result.markdown)

__plugin_interface_version__ = 1

class MyConverter(DocumentConverter):
    def accepts(self, file_stream, stream_info, **kwargs) -> bool:
        return (stream_info.extension == ".myext" or
                stream_info.mimetype == "application/x-myformat")

    def convert(self, file_stream, stream_info, **kwargs) -> DocumentConverterResult:
        content = file_stream.read()
        return DocumentConverterResult(markdown=converted, title=None)

def register_converters(markitdown, **kwargs):
    markitdown.register_converter(MyConverter(), priority=0.0)

[project.entry-points."markitdown.plugin"]
my_plugin = "my_package.plugin_module"

import re

def clean_markdown_for_llm(md_text: str) -> str:
    md_text = re.sub(r'\n{3,}', '\n\n', md_text)
    md_text = re.sub(r'\n.*?Page \d+ of \d+.*?\n', '\n', md_text)
    md_text = re.sub(r'(\n---\n)+', '\n---\n', md_text)
    md_text = re.sub(r'<!--.*?-->', '', md_text, flags=re.DOTALL)
    return md_text.strip()

pip install markitdown-ocr
markitdown --use-plugins scan.pdf

pip install 'markitdown[az-doc-intel]'
markitdown scan.pdf -d -e "https://<your-endpoint>.cognitiveservices.azure.com/"

from markitdown import MarkItDown

def convert_with_ocr_fallback(path, docintel_endpoint=None):
    md = MarkItDown()
    result = md.convert(path)
    if result.markdown.strip():
        return result.markdown
    if docintel_endpoint:
        md_ocr = MarkItDown(docintel_endpoint=docintel_endpoint)
        return md_ocr.convert(path).markdown
    return ""

from markitdown import MarkItDown, StreamInfo

md = MarkItDown()
with open("sjis_file.txt", "rb") as f:
    stream_info = StreamInfo(
        extension=".txt",
        charset="shift_jis",
    )
    result = md.convert_stream(f, stream_info=stream_info)
    print(result.markdown)

markitdown file.pdf -o output.md

pip install 'markitdown[audio-transcription]'
brew install ffmpeg
sudo apt-get install ffmpeg
export FFMPEG_PATH=/usr/local/bin/ffmpeg

npx @modelcontextprotocol/inspector
markitdown-mcp --http --host 127.0.0.1 --port 3001