能力标签

🔌 MCP 🤖 Agent 🔄 工作流 👁 OCR 🌐 翻译 💻 CLI 🧬 Embedding 📚 RAG 🖼 视觉 🧠 Claude

⚙️

Agent工作流

论文阅读器

基于 Rust · 无代码搭建完整 AI 自动化流程

英文名：lumenfolio

⭐ 12 Stars 🍴 2 Forks 💻 Rust 📄 AGPL-3.0 🏷 AI 8.0分

8.0AI 综合评分

AI论文阅读Rust本地优先

⬇ 下载源码（GPL） ⚙️ 配置说明 📺 TG 频道

✦ AI Skill Hub 推荐

AI Skill Hub 强烈推荐：论文阅读器是一款优质的Agent工作流。AI 综合评分 8.0 分，在同类工具中表现稳健。如果你正在寻找可靠的Agent工作流解决方案，这是一个值得深入了解的选择。

📚 深度解析

论文阅读器是一套完整的 AI Agent 自动化工作流方案。随着 AI 能力的不断提升，基于 Agent 的自动化工作流正在成为提升个人和团队效率的核心方式。区别于传统的 RPA 自动化（模拟鼠标键盘操作），AI Agent 工作流通过理解任务意图、动态规划执行路径，能够处理更复杂的非结构化任务。

论文阅读器工作流的设计遵循"最小配置，最大复用"原则：核心逻辑已经封装好，用户只需配置自己的 API Key 和业务参数即可快速上手。工作流内置错误处理和重试机制，在网络波动或 API 限速等情况下仍能稳定运行，适合作为生产环境的自动化基础设施。

在实际部署时，建议先在测试环境中运行 3-5 次，验证各个环节的输出结果符合预期，再部署到生产环境。AI Skill Hub 评分 8.0 分，是同类 Agent 工作流中的精选推荐。

📋 工具概览

本地优先的AI论文阅读器，支持向量化PDF、页码/边界框引用、OCR/TS

论文阅读器是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。

GitHub Stars

⭐ 12

开发语言

Rust

支持平台

Windows / macOS / Linux

维护状态

轻量级项目，按需更新

开源协议

AGPL-3.0

AI 综合评分

8.0 分

工具类型

Agent工作流

Forks

📖 中文文档

以下内容由 AI Skill Hub 根据项目信息自动整理，如需查看完整原始文档请访问底部「原始来源」。

本地优先的AI论文阅读器，支持向量化PDF、页码/边界框引用、OCR/TS

📌 核心特色

可视化 Agent 工作流编排，无需编写复杂代码
支持多步骤自动化任务链，实现全流程无人值守
与外部 API、数据库和第三方服务无缝集成
内置错误处理与自动重试机制，保障稳定运行
提供可复用的自动化模板，快速在同类场景部署

🎯 主要使用场景

自动化日常重复性工作，将精力集中于创造性任务
构建数据采集 → 处理 → 输出的完整自动化管线
实现跨平台、跨系统的数据流转和业务协同

以下安装命令基于项目开发语言和类型自动生成，实际以官方 README 为准。

安装命令

# 方式一：cargo install（推荐）
cargo install lumenfolio

# 方式二：从源码编译
git clone https://github.com/tanghui315/lumenfolio
cd lumenfolio
cargo build --release
# 二进制在 ./target/release/lumenfolio

📋 安装步骤说明

访问 GitHub 仓库获取工作流文件
在对应平台（Dify / Flowise / Make 等）中找到「导入工作流」功能
上传工作流文件
按照提示配置必要的环境变量和 API Key
运行测试确认流程正常后投入使用

以下用法示例由 AI Skill Hub 整理，涵盖最常见的使用场景。

常用命令 / 代码示例

# 查看帮助
lumenfolio --help

# 基本运行
lumenfolio [options] <input>

# 详细使用说明请查阅文档
# https://github.com/tanghui315/lumenfolio

以下配置示例基于典型使用场景生成，具体参数请参照官方文档调整。

配置示例

# lumenfolio 配置说明
# 查看配置选项
lumenfolio --config-example > config.yml

# 常见配置项
# output_dir: ./output
# log_level: info
# workers: 4

# 环境变量（覆盖配置文件）
export LUMENFOLIO_CONFIG="/path/to/config.yml"

📑 README 深度解析真实文档完整度 55/100 查看 GitHub 原文 →

以下内容由系统直接从 GitHub README 解析整理，保留代码块、表格与列表结构。

简介

Lumenfolio

Local-first AI paper reader with verifiable page/bbox evidence.

Ask questions, translate PDFs, inspect tables/figures, and use local Codex / Claude Code agents without uploading papers by default.

Download · Watch 30s Demo · Star on GitHub · Give Feedback

Available for macOS Intel, macOS Apple Silicon, and Windows x86_64.

中文文档 (Chinese README)

Lumenfolio is a local-first desktop AI reading workspace for academic PDFs. It combines a focused PDF reader, vectorless agentic RAG, layout-aware translation, evidence-anchored notes, OCR / TSR visual evidence, and local Codex / Claude Code agents into one reading environment.

It is not just "chat over a PDF". Lumenfolio is built around local document evidence: pages, blocks, chunks, structure, tables, visual regions, citations, and bounding boxes that can point back to the original PDF. That evidence can also be exposed through local MCP tools so your signed-in Codex / Claude Code CLI can gather evidence and answer inside the reading workflow.

Lumenfolio demo

If Lumenfolio helps your paper-reading workflow, a star helps more researchers discover it.

Features

Three-pane reading workspace:
left: workspace folders and the PDFs in each folder
center: PDF reader, selection tools, translation controls
right: independent agent sessions, evidence chain, agent trace, and a floating notes drawer
Independent multi-session agent workspace (sessions decoupled from documents)
Local PDF indexing into SQLite
Agentic Q&A with citations, single- or cross-document
Workspace-aware retrieval: the agent can see and answer about your whole indexed library, with on-demand discovery for large libraries
Native tool-calling agent loop for capable models, with a rule-driven fallback for weaker/local models
Local agent providers: auto-detect Codex / Claude Code CLI and use them as no-extra-API-key local chat models
Local-agent MCP mode: let Codex / Claude call Lumenfolio read-only tools for pages, sections, tables, and visual evidence
Cross-document chat: @-mention up to 4 other indexed papers in one question
Knowledge precipitation (summary, entities, concepts, keywords) per document, fully local
Cross-document knowledge graph: reader-side concept-bridge graph + full-screen library graph with communities and insights
Trending papers feed (Hugging Face) with Daily/Weekly/Monthly scopes, one-click add into the workspace
Context-aware agent tools: search the whole library by concept, query the trending list, and route across related documents
Evidence chain and foldable agent trace in chat
Provider-based chat and translation configuration, with per-model context window auto-detected from the provider (manual override available)
Visual/table-aware retrieval path with rendered crops and TSR-ready table evidence
Local OCR for scanned/image-only PDFs on macOS Apple Silicon and Windows
PDFMathTranslate-based sidecar for layout-aware translation

Prerequisites

Node.js 18+ (LTS recommended)
npm 9+
Rust stable toolchain
Platform requirements for Tauri 2 (macOS/Linux/Windows build dependencies)

Build & Verification

Frontend build:

npm run build

Rust checks:

cd src-tauri
cargo check
cargo test

Suggested smoke sequence for most changes:

npm run build
cd src-tauri && cargo test

Additional project checks:

npm run check:translation-linking
npm run check:prod-no-testids

Trust, Data & Installation

Lumenfolio is local-first. PDF indexes, notes, chat history, and translation metadata are stored locally.
API keys are currently stored locally; migration to the system keychain is planned.
If a cloud chat or translation provider is configured, selected text, questions, page context, or translation content may be sent to that provider.
If you choose a local Codex / Claude Code provider, questions, conversation memory, and retrieved PDF evidence are passed to that local CLI; the actual model request is handled by the CLI and its signed-in account/subscription.
macOS builds are currently ad-hoc signed. Developer ID signing and notarization are planned to reduce first-run friction.
Release assets include SHA-256 checksums plus license, notice, AGPL sidecar license, and PDFMathTranslate source archive.