"Not just a Reader, but a Cognitive Recorder."

Welcome to Read it DEEP, the AI-driven platform designed to transform how you interact with research papers. Move beyond passive reading into active knowledge construction with our "Deep Read" philosophy.

Read it DEEP is a dual-engine platform:

1. Cognitive Recorder: It tracks your reading path, highlighting, and thought process.
2. Research Asset Factory: It refines raw papers into structured assets—methods, datasets, and inspirations.

Powered by LangGraph and state-of-the-art LLMs, we turn your library into a Dynamic Knowledge Graph.

Efficiently bringing knowledge into your system.

The journey begins with our Smart Ingestion pipeline.

• Drag & Drop: Simply drag your PDF into the upload area.
• Mineru Parsing: Our integration with Mineru V4 ensures high-fidelity parsing, preserving layout, formulas, and images as Markdown.
• Real-time Feedback: Watch as your paper goes from Uploading → Parsing → Indexing.

Your organized research headquarters.

Once ingested, papers appear in your Library.

• Auto-Metadata: We automatically fetch titles, authors, and publication dates.
• Visual Cards: Papers are presented as cards with key details, making retrieval instant.
• Search & Filter: Quickly find papers by keywords or topics.

Focus, connect, and think.

Clicking "Start Deep Reading" activates our signature 3-Column Layout:

• Interactive Citations: Hover over a citation [1] to see the reference instantly without losing your place.
• Translation: Seamlessly switch between original and translated text with a single click.

Where information becomes an asset.

This is the heart of "Deep Reading".

• Method Alchemy (方法炼金台): Select a method description in the text, and the AI extracts parameters, loss functions, and even generates PyTorch pseudocode.
• Data Warehouse (资产仓库): Automatically validates dataset URLs and licenses.
• Idea Canvas (灵感画板): Record your hypotheses and link them directly to the evidence in the text.

Visualizing your second brain.

As you read, the graph evolves.

• Citation Links: See what influenced this paper.
• Similarity Connections: Discover papers in your library with similar concepts, powered by vector embeddings.

• Local-First AI: Powered by local LLMs (vLLM/Ollama compatible) for privacy and speed.
• LangGraph Agents: sophisticated loops for self-correcting extraction and verification.
• Vector Database: pgvector integration for semantic search and graph construction.
• Modern Stack: Built with React, Vite, Tailwind, Python FastAPI, and SQLite/PostgreSQL.

Ready to dive deep?

1. Upload your first paper.
2. Open it in the Reader.
3. Activate the Workbench.
4. Build your Knowledge Graph.

# 配置环境变量
cp .env.example .env
# 编辑 .env 填写 API Keys

# 启动开发服务器
./start.sh

cp .env.docker.example .env
./docker-start.sh              # 默认: Frontend 3000, Backend 8080
./docker-start.sh 80 8080      # 使用 80 端口

GitHub Actions 会在每次推送时自动构建镜像到 GHCR。

# 1. 上传配置到服务器
scp docker-compose.ghcr.yml .env user@server:/opt/readitdeep/

# 2. 在服务器上拉取并启动
cd /opt/readitdeep
docker compose -f docker-compose.ghcr.yml pull
FRONTEND_PORT=80 GITHUB_OWNER=oMygpt docker compose -f docker-compose.ghcr.yml up -d

docker compose -f docker-compose.ghcr.yml pull
docker compose -f docker-compose.ghcr.yml up -d

所有数据存储在 ./readit_data/ 目录:

• db/ - 数据库、papers.json、workbench.json
• uploads/ - PDF 和解析结果
• redis/ - 缓存数据

📚 详细部署文档: docs/DOCKER_DEPLOYMENT.md

Read it DEEP — Where reading meets thinking.

This project is licensed under the GNU Affero General Public License v3.0 (AGPL v3).

• Open Source: You are free to use, modify, and distribute this software under the terms of the AGPL v3.
• Copyleft: If you modify this software and distribute it (web service included), you must make your modifications open source under the same license.
• Dual Licensing: For commercial use cases, proprietary integration, or exemptions from AGPL conditions, please contact CHUNLIN@Readit DEEP for a commercial license.

Copyright (C) 2025 CHUNLIN@Readit DEEP.