#3434th of 50 in Data and scraping for AIFor data and documents, documents and ocr
indra
CPU-only OCR model that converts PDFs and images to markdown or JSON
- Stars
- 1.8k
- License
- Apache-2.0
- Last commit
- Nov 2026
- Last release
- Oct 2026
- Language
- Python
86 on 13 Oct 20261,760 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Datalab is hiring: 4 open roles
Overview
Indra OCR is a small ONNX Runtime OCR pipeline (about 76 MB of default models, no PyTorch) that converts PDFs and images into markdown, text, or JSON with line boxes, layout blocks and reading order. It covers 80+ languages through four script-specific recognizers, and for digital PDFs it reads the text layer and OCRs only what is missing. It ships a CLI, a Python API and a local web viewer.
Who it is for: Engineers needing fast self-hosted OCR of PDFs and scans without a GPU
Strengths
- Runs on CPU via ONNX Runtime; about 212 ms median per page on 4 vCPUs
- Scores 66.7 on OmniParseBench vs 58.4 for PP-StructureV3, per the README
- Outputs markdown, text, or JSON with boxes, layout labels and reading order
- Handles rotated, tilted and vertical pages; uses the PDF text layer when it passes checks
Weaknesses
- Handwriting scores 25.2% and degraded scans 29.5% on OmniParseBench
- Complex tables with merged cells or no ruling lines often misplace cells
- Math is plain text unless --math is set; it is slower and lowers table score
- Weights use modified OpenRAIL-M; commercial use above $20M funding or revenue needs a license
What it needs
- no GPU
- Needs onnxruntime, opencv, pypdfium2, Hugging Face (model download)
- Models: ONNX, indra-ocr
- port 8765
Quick answers
Is indra open source?
Yes. Its code is on GitHub under the Apache-2.0 license.
Does indra need a GPU?
No, it runs without one.
Also in Data and scraping for AI
See all 50| Rank | Project | Score |
|---|---|---|
| 1 | doclingPython library that converts PDFs, Office files and more into structured text | 92 out of 100 |
| 2 | mineruParses PDFs, images and Office files into Markdown and structured JSON | 91 out of 100 |
| 3 | opendataloader-pdfPDF parser that outputs Markdown, JSON with bounding boxes, and Tagged PDF | 90 out of 100 |
| #4 | agent-reachCLI that gives coding agents read access to login-gated and Chinese web platforms | 88 out of 100 |
| #5 | markitdownPython library and CLI that converts office files, PDFs and audio to Markdown | 87 out of 100 |