Document Parser GitHub: The 7 Repos You Need for Document AI
The best document parsers live on GitHub. Here are 7 essential repos with verified stars, from PDF parsing to OCR and formula extraction.
💡 What You Will Learn
The best document parsers live on GitHub. Here are 7 essential repos with verified stars, from PDF parsing to OCR and formula extraction.
Document AI starts with a good parser, and the best parsers are open source. This is the document parser GitHub shortlist we recommend after months of testing - each with verified star counts from August 2026.
The 7 Essential Repos
opendatalab/MinerU (76,942) - full PDF understanding with layout, formulas, and OCR. docling-project/docling (64,320) - structure-aware document conversion. datalab-to/marker (38,453) - fast PDF to markdown. Unstructured-IO/unstructured (15,268) - 25+ file formats. opendatalab/PDF-Extract-Kit (9,933) - formula and layout extraction. microsoft/table-transformer (2,935) - table detection. Byaidu/PDFMathTranslate (35,984) - PDF translation preserving formulas.
You do not need all seven. The pattern: MinerU or Docling as the main parser, marker for speed-critical conversion, PDF-Extract-Kit when formulas matter, table-transformer when tables dominate, Unstructured for mixed inputs, PDFMathTranslate for translating academic papers.
Comparison
| Repo | Job | Stars |
|---|---|---|
| MinerU | Full PDF understanding | 76,942 |
| Docling | Structure-aware conversion | 64,320 |
| Marker | Fast PDF to markdown | 38,453 |
| Unstructured | 25+ formats | 15,268 |
| PDF-Extract-Kit | Formulas and layout | 9,933 |
| Table Transformer | Table detection | 2,935 |
| PDFMathTranslate | PDF translation | 35,984 |
FAQ
Q: Are these production-ready?
A: Yes - MinerU, Docling and Unstructured all have stable APIs and are used in production pipelines at scale.
Q: Which has the best docs?
A: Unstructured has the most thorough documentation; MinerU has the largest community and most examples.
