DRAGON-AI
π DRAGON-AI β is an open-source enterprise Agent platform for building, deploying, and managing AI applications. It combines AI agents, workflows, RAG and know
What is DRAGON-AI?
π DRAGON-AI β is an open-source enterprise Agent platform for building, deploying, and managing AI applications. It combines AI agents, workflows, RAG and knowledge management
Overview of DRAGON-AI
What it is
DRAGON-AI is an open-source enterprise Agent platform for building, deploying, and managing AI applications. It combines AI agents, workflows, RAG (Retrieval-Augmented Generation), and knowledge management into a single deployable system. The platform supports multi-tenancy and private deployment, giving users control over data, models, and permissions.
Key features
- Enterprise RBAC with resource-level ACL, audit logging, rate limiting, online session management, and ARQ pipeline monitoring.
- AI model gateway with OpenAI-compatible unified entry, supporting 12 model capabilities across 3 vendors (OpenAI-compatible, Tongyi Qianwen, Zhipu), including text generation, vector, rerank, image/video understanding, OCR, image generation, video generation, audio generation, and speech-to-text.
- Visual workflow orchestration with 21 node types (AI, control flow, data, external, approval, sub-workflow), draft/publish/versioning/rollback, and execution via WebSocket, SSE, or DB replay.
- Knowledge base and RAG with three library types (document, image-text, audio-video), hybrid retrieval (BM25 + dense vector kNN + graph-enhanced), and RAGAS evaluation.
- Deep exploration (Harness) agent with AgentScope ReAct, per-user cloud sandbox with persistent workspace, streaming timeline dialogue, human-in-the-loop AskUser, and automatic artifact rendering with online preview.
- Execution engine with arq + Redis sharded queues across four independent pipelines (workflow, ragflow, graphflow, harness), unified storage abstraction (local, NAS, OSS), and Nacos-based configuration.
Use cases
Building and deploying enterprise AI applications that require multi-tenant access control, integration of multiple AI models, complex workflow automation, knowledge retrieval from documents and media, and autonomous agents that execute tasks with sandboxed environments. Suitable for teams needing private deployment with full data sovereignty.
Getting started
See the repository README for full setup instructions. The quick start requires Python 3.11+, Node.js 20.12+/pnpm 10+, MySQL 8.x, Redis 5+ (recommended 7.x), and Nacos 2.x+. Example initialization command:mysql --default-character-set=utf8mb4 -h -u root -p < sql/v2_init.sql
License and pricing
Licensed under Apache-2.0. The project is open-source; pricing is not stated.
Based on the project's public README and repository metadata.