RAG и базы знаний
opendataloader-pdf
Turn PDFs into AI-ready data: extract Markdown, JSON and tables — and auto-tag for accessibility
★ 29,5 тыс. звёздApache-2.0
Смотреть руководство →Marker — проект с открытым исходным кодом из категории «RAG и базы знаний» от datalab-to (datalab-to/marker на GitHub). У него 40 648 звёзд на GitHub и 2 945 форков, он распространяется по лицензии Apache-2.0 и последний раз обновлялся на GitHub 22 октября 2026.
Marker — converts PDFs to Markdown and JSON quickly and accurately, with layout, tables and math intact, Apache-2.0
The base package covers PDF conversion from PyPI.
Documents beyond PDFs need the full dependency set.
On an NVIDIA GPU use vLLM with Docker and the NVIDIA Container Toolkit; on CPU or Apple Silicon install the llama-server binary from llama.cpp.
Layout, tables and math survive as structured Markdown and JSON, not flattened text.
Balanced mode for maximum quality on GPU; fast mode for CPU-friendly conversion with targeted repair of bad blocks.
Layout detection and OCR come from dedicated vision models, with re-OCR of pages whose embedded text is unreliable.
Output is made to feed search, RAG and dataset tooling, not just to be read once.
The README install route from PyPI.
pip install marker-pdfAdds support for documents other than PDFs.
pip install marker-pdf[full]The llama-server binary Marker needs on non-GPU machines.
brew install llama.cppThe datalab-to team, built on the Surya vision models.
Yes — Apache-2.0 licensed.
No — fast mode is optimised for CPU, with llama.cpp as the backend; balanced mode is best on an NVIDIA GPU with vLLM.
Markdown and JSON with layout, tables and math preserved — structured output for search and RAG, not a plain text dump.
pip install marker-pdf (or marker-pdf[full] for non-PDF documents), set up the vLLM or llama.cpp backend, and convert.
Turn PDFs into AI-ready data: extract Markdown, JSON and tables — and auto-tag for accessibility
MinerU — turns PDFs and Office documents into LLM-ready Markdown and JSON for agents and RAG pipelines, Apache-2.0 plus terms
Render and deploy a folder of markdown as a site in one move.
A self-hostable markdown team wiki with a built-in MCP server.
Polyglot document intelligence in Rust — extract text, tables and structured data from 106 formats.
Knowledge-base — 本地优先的知识库桌面应用。markdown 编辑器 + 全文搜索（fts5） + 双向链接 / 知识图谱；多端, publicly avai...
Другие проекты из категории «RAG и базы знаний» с наибольшим числом звёзд в каталоге LibreAI — реалистичный список альтернатив, если Marker вам не подходит.
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or…
RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context…
Lightweight OCR toolkit turning PDFs and images into structured, LLM-ready data in 100+ languages
MinerU — turns PDFs and Office documents into LLM-ready Markdown and JSON for agents and RAG pipelines, Apache-2.0 plus terms
Stop renting your intelligence. Own it with AnythingLLM. Everything you need for a powerful local-first agent experience
Соседние проекты из категории «RAG и базы знаний» в каталоге, ближайшие по числу звёзд к Marker — продолжайте изучать полку по обе стороны от него.
Meta's library for efficient similarity search and clustering of dense vectors, in C++ and Python
Open-source engine turning continuous content streams into search, monitoring and alerts
Complete API layer for private AI apps on local models: RAG, skills, tools, MCP and text-to-SQL
PageIndex: vectorless, reasoning-based RAG that indexes documents like a human expert would
Ready-to-run templates for RAG, AI pipelines and enterprise search that stay in sync with live data
Local knowledge-base RAG and agent app built on LangChain with ChatGLM, Qwen and Llama models
Бесплатно и открыто — забирайте исходный код, ставьте звезду репозиторию и начинайте создавать.