Summary

Resolves #2182.

This PR adds a lightweight --info flag (and a public API method get_metadata()) that outputs structured JSON metadata (path, size, mime type, and detected converter) for a given file or stream without executing the full conversion pipeline.

This is highly useful for triage, cost/duration estimation, and routing in AI agent pipelines, where agents want to know what a document is before committing to full, potentially expensive conversion paths.

Examples

$ markitdown --info example.docx
{
  "mime_type": "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
  "detected_converter": "DocxConverter",
  "size_bytes": 135824,
  "path": "example.docx"
}

Changes

• Implemented public get_metadata(), get_metadata_local(), and get_metadata_stream() methods in the MarkItDown class (_markitdown.py) to resolve and return metadata by querying the accepts() method of registered converters.
• Added a --info option to the CLI (__main__.py) which triggers get_metadata() and outputs the results as JSON.
• Added comprehensive unit tests in test_module_misc.py (test_get_metadata()) and test_cli_misc.py (test_info_flag()).
• Updated existing CLI tests in test_cli_misc.py to use sys.executable instead of the hardcoded "python" command, ensuring they run correctly on systems (like macOS) that only have python3 on their PATH.