feat: Add batch conversion API and CLI support

[jacken]

Adds convert_batch() to the MarkItDown class and corresponding CLI flags for converting multiple files concurrently.

Library API

A new convert_batch() method on MarkItDown accepts an iterable of sources and yields BatchConversionResult objects in completion order:

from markitdown import MarkItDown, BatchConversionResult

md = MarkItDown()
for result in md.convert_batch(["file1.pdf", "file2.docx", "file3.html"]):
    if result.success:
        print(result.result.markdown)
    else:
        print(f"Failed: {result.source} — {result.error}")

Each BatchConversionResult has:

• source — the original input
• result — a DocumentConverterResult, or None on failure
• error — the exception, or None on success
• success — True when no error is set

Error handling is configurable via on_error:

• "collect" (default) — wraps errors and continues
• "raise" — re-raises the first error immediately

Concurrency uses ThreadPoolExecutor (appropriate since heavy converters like pdfminer and lxml release the GIL). A pluggable executor parameter lets callers supply their own executor; if omitted, one is created and shut down automatically. The workers parameter controls thread count when using the default executor.

BatchConversionResult is exported from the top-level markitdown package.

CLI

# Convert multiple files to stdout (separated by banners)
markitdown file1.pdf file2.docx

# Write each file to an output directory
markitdown file1.pdf file2.docx --output-dir ./output/

# Control parallelism
markitdown file1.pdf file2.docx --workers 4

# Stop on first error instead of collecting all errors
markitdown file1.pdf file2.docx --fail-fast

Single-file and stdin modes are unchanged.

Tests

• tests/test_batch.py — unit and integration tests for BatchConversionResult and convert_batch()
• tests/test_cli_misc.py — CLI tests for multi-file stdout, --output-dir, --fail-fast, --workers, and error collection

[jacken]

@microsoft-github-policy-service agree