# getomni-ai/zerox

OCR & Document Extraction using vision models

Repository: https://github.com/getomni-ai/zerox
Canonical: https://ross.abutalabs.com/products/zerox
Homepage: https://getomni.ai/ocr-demo
Language: TypeScript
License: MIT
License Family: permissive
Topics: ocr, pdf
Last push: 2025-05-20T00:39:18+00:00

## Health v2 (maintenance only)
Score: 35/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 22, release rhythm 40, longevity 55
- inputs: {"age_days": 773, "days_push": 471, "days_rel": 623, "gap_med": 2, "n_releases_24m": 8}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 12266, forks 848 (observed 2026-08-28T04:10:52.200031+00:00)

## What it is
Zerox is a library (Node.js and Python packages) that performs OCR and document extraction by converting files like PDFs, DOCX, and images into page images and sending them to vision models (GPT-4o, Gemini, Bedrock, etc.) to produce Markdown. It supports schema-based data extraction, concurrent page processing, and format maintenance across pages.

## Use cases
- convert pdfs to markdown for llm ingestion
- extract text from scanned documents using vision models
- ocr documents with weird layouts and tables
- parse docx and image files into markdown
- extract structured data from documents with a schema
- prepare documents for rag pipelines

## When to choose
- you need markdown output from complex document layouts, tables, or charts
- you want to leverage multimodal LLMs instead of traditional OCR engines
- you need support for multiple model providers (OpenAI, Azure, Bedrock, Gemini)
- you're building AI ingestion or RAG pipelines over documents

## When to avoid
- you need fully offline or free OCR without LLM API costs
- you require pixel-perfect text extraction where hallucination is unacceptable
- you need high-volume batch processing with minimal latency and cost
- you need a traditional OCR engine like Tesseract

## Facets
- artifact type: library
- maturity: active
- function: ocr, pdf, llm-inference, machine-learning, data-science
- domain: pdf, artificial-intelligence, large-language-models, developer-tools, files
- platform: python, cross-platform
- tags: vision-models, document-extraction, markdown-output, openai, gemini, aws-bedrock, document-ai, nodejs

## Member repositories
- getomni-ai/zerox (main) score 35

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:52.200031+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:14:12.497460+00:00, confidence not recorded.
  - readme: https://github.com/getomni-ai/zerox (fetched 2026-08-28T04:10:52.200031+00:00, sha f5731f1d5471)
  - registry_npm: https://registry.npmjs.org/zerox (fetched 2026-08-29T08:11:56.042460+00:00, sha 7b01d8633b96)
- Data as of 2026-08-30T08:39:29.467469+00:00.