Skip to content

MiniCPM-OCR SDK v1.0.5

Latest

Choose a tag to compare

@huangfw huangfw released this 09 Nov 09:42
· 14 commits to main since this release

Support parsing multiple files at once, with each file getting its own result.

  • The parse command line supports passing multiple files or a directory at the same time (top level only, no recursion), and with -o DIR outputs a separate Markdown and JSON for each.

  • --split-images makes each individual image its own separate document instead of pages of the same document; PDFs are always merged by file.

  • The pipeline.image_document_scope configuration option controls the same behavior (collection / file).

  • input.file_paths supports listing a set of input files in a specified order in YAML.

  • assemble_documents_json() returns results grouped by input file, with the overall batch status being Success / Partial / Failed.

  • Page results' metadata adds new source (original file path) and image_id fields.