Parse PDF, Word, and Excel documents into structured Markdown with three-level parser fallback.
npx skills add https://github.com/anbeime/skill --skill antinet-doc-parse
Please help me install this Agent Skill. Skill: antinet-doc-parse Source: https://github.com/anbeime/skill/tree/main/skills/antinet-doc-parse Command: npx skills add https://github.com/anbeime/skill --skill antinet-doc-parse
Converting heterogeneous documents (PDF, Word, Excel) into clean, machine-readable text for RAG pipelines and knowledge bases usually requires juggling multiple parsers and handling failures manually. This Skill automates multi-format document parsing with a three-level fallback chain, outputting structured Markdown, metadata, and a confidence score.
Run the script with python skills/doc-parse/scripts/run_doc_parse.py to parse the input documents and write the structured results to examples/snse_survey/skill_outputs/doc_parse.json.
High-intent search queries and answers about installing and using this skill.
Run the run_doc_parse.py script, which invokes a three-level parsing chain and outputs structured Markdown plus metadata as JSON. The result includes a confidence score so downstream retrieval stages can decide whether manual review is needed.
MinerU provides the strongest layout reconstruction and is tried first. PyMuPDF offers fast general PDF parsing as the second tier, while pdfplumber handles table and text extraction as the final fallback.
If all three parsers fail, the file is marked for manual intervention and no partial result is returned. For oversized single pages, the document is parsed in chunks and only failed chunks receive a low confidence flag.
It writes a JSON file containing the Markdown body, metadata such as title, page count, table count, and author, a 0-1 confidence score, and the name of the parser that ultimately succeeded.
Yes, the Skill accepts an optional formats whitelist parameter to limit supported input types. By default all formats are accepted, and python-magic is used for file type detection.
nathanabayUpload, retrieve, and manage permissions for files in Frappe workflows.
jg-chalk-ioBuild RAG pipelines with vector databases, embeddings, and LLM generation.
himanshu231204Route low-confidence regex extractions to LLM validation for structured text parsing.
NousResearchExtract text and structured data from PDFs, scanned documents, and images.
