English | 简体中文

LightningRAG is a Vue + Gin full-stack starter built around RAG (retrieval-augmented generation): knowledge bases (ingest, parse, chunk, vector retrieval), pluggable LLMs, embeddings, rerankers, and vector stores, Agent orchestration on a visual canvas (retrieval, LLM, tools, control flow), and optional webhook channel connectors (Feishu, DingTalk, Slack, Teams, and more). The admin surface unifies permissions, menus, and configuration alongside these RAG features.

• Frontend: Vue (Vite), Element Plus–style UI
• Backend: Go / Gin, GORM (MySQL), Redis, Swagger, Viper
• Platform: JWT, dynamic routes and menus, Casbin, form builder, code generation, and sample scaffolding
• RAG & agents: Hybrid / multi-path retrieval (vector, keyword, PageIndex, and related retriever types); conversation and streaming APIs with optional references; modular providers under server/rag/ (server/rag/README.md)
• Channels: Published agents can be bound to webhook endpoints for third-party platforms — see docs/THIRD_PARTY_CHANNEL_CONNECTORS.md

Requirements: Node.js > v18.16.0, Go ≥ 1.22. Full clone, dependency, and run steps:

• English: README.md
• 简体中文: README_zh.md

Demo: https://demo.LightningRAG.com — username admin, password 123456 (see main README for details).

Repository: https://github.com/LightningRAG/LightningRAG

Apache 2.0. Commercial use and attribution requirements are described in the “Important notices” section of the main README.

Short intro for GitHub org profile and similar contexts; for deployment, configuration (config.yaml → rag:), and contribution guidelines, use the repository root README.md / README_zh.md.