A full-stack Vue + Gin starter kit, LightningRAG facilitates the development of Retrieval-Augmented Generation (RAG) applications. It targets developers and researchers seeking a robust, decoupled framework for managing knowledge bases, integrating Large Language Models (LLMs), and orchestrating complex agent workflows, offering an extensible RAG pipeline and a comprehensive feature set.

How It Works

The architecture features a decoupled Vue frontend and a Gin backend, built around a pluggable RAG system. It handles document ingestion, parsing, chunking, and vectorization into various vector stores. Retrieval methods include hybrid approaches like vector search, keyword matching, and PageIndex. Agent orchestration is visualized on a canvas, enabling multi-step automation via configurable nodes. The system supports pluggable LLMs, embeddings, and rerankers, with optional webhook connectors for seamless integration with chat platforms.

Quick Start & Requirements

• Installation: Backend: Navigate to the server directory, run go mod tidy, go generate, then go run .. Frontend: Navigate to the web directory, run npm install, then npm run serve. For an embedded UI, use make build-server-embed-local.
• Prerequisites: Node.js version > v18.16.0, Golang version >= v1.22, MySQL > 5.7 (InnoDB), and Redis are required. Optional cloud storage credentials (Qiniuyun, Aliyun, Tencent Cloud) are needed for file uploads.
• Links: Project GitHub: https://github.com/LightningRAG/LightningRAG. The online demo link is currently offline.

Highlighted Details

• Extensible RAG Pipeline: Manages knowledge bases, supports hybrid retrieval (vector, keyword, PageIndex), and integrates pluggable LLMs, embeddings, and rerankers.
• Visual Agent Orchestration: Features a canvas-based flow editor for building multi-step agent automation with an extensible tool registry.
• Multi-Platform Connectors: Includes built-in webhook support for integrating published agents with platforms such as Slack, Teams, Feishu, and DingTalk.
• Full-Stack Features: Provides JWT/Casbin authentication, user/role management, dynamic menus, a form builder, and code generation capabilities.
• Deployment Flexibility: Supports standard split-binary deployment (API + Nginx) or a single embedded binary. Cross-platform builds are facilitated via GoReleaser.

Maintenance & Community

The project utilizes GoReleaser for multi-platform builds, indicating active development. While contribution guidelines and a contributor list are available, specific community channels like Discord or Slack are not detailed in the provided README.

Licensing & Compatibility

Licensed under the Apache 2.0 license. Users must comply with its terms, including retaining work attribution. Commercial use is generally permitted under Apache 2.0, provided attribution requirements are met.

Limitations & Caveats

The project's online demo is currently unavailable. File uploads require users to provide their own credentials for specific cloud storage providers (Qiniuyun, Aliyun, Tencent Cloud). Multi-login functionality necessitates Redis configuration.