Dryade：自托管AI编排平台，数据主权时代的智能代理基础设施

Dryade is a self-hosted AI orchestration platform designed to address data privacy risks, vendor lock-in, and network dependency issues of cloud-based AI services. It supports local LLM deployment, multi-agent orchestration, RAG, visual workflow building, and edge hardware integration, empowering users to retain data sovereignty while leveraging AI capabilities.

Most AI applications rely on cloud services, leading to three core issues:

1. Data privacy risks (sensitive data exposure, compliance barriers for regulated industries)
2. Vendor lock-in (high migration costs, limited bargaining power)
3. Network dependency (unusable in offline/edge environments) Dryade was created to solve these by enabling self-hosted AI operations without external data transmission.

Key features include:

• Multi-model support: Local models (vLLM/Ollama: Llama, Qwen, Mistral), cloud APIs (OpenAI, Anthropic), custom endpoints
• Multi-agent orchestration: Chat (dialogue), Planner (task decomposition), Orchestrate (autonomous workflow) modes
• MCP integration: Connects to external tools/services via Model Context Protocol
• RAG: Built-in document processing, vectorization, semantic search for accurate responses
• Visual workflow: Drag-and-drop builder (ReactFlow) for low-code/no-code AI workflows
• Plugin ecosystem: Extensible via plugins (official market planned)

Architecture: Frontend (React/TypeScript), Backend (FastAPI), Orchestrator (ReAct loop), Tool Router (semantic+regex), LLM providers Agent adapters: MCP, CrewAI, ADK, LangChain, A2A Edge support: Optimized for NVIDIA Jetson (edge AI), DGX Spark (desktop AI), general GPU servers. Use cases: industrial sites, military/government, remote facilities, mobile platforms.

Deployment methods:

1. Docker Compose (recommended): Clone repo → copy .env → docker compose up -d (defaults to Ollama, configurable)
2. Manual: Use uv (Python) and npm (frontend) to start services
3. Edge hardware: Specialized guides for Jetson and DGX Spark

Dryade's key differentiators (vs Dify, n8n, Langflow):

• Full data sovereignty (zero telemetry)
• Native edge hardware support
• MCP server integration
• Multi-agent framework adapters
• Planned plugin market These make it stand out for privacy-focused, edge, and flexible AI deployment needs.

License: Dryade Source Use License (DSUL) — core features open-source free; enterprise features under separate terms Community: Discord, GitHub Discussions, contribution guides, examples, official docs (dryade.ai/docs) Conclusion: Dryade aligns with trends toward edge-distributed AI, data sovereignty, and transparent control. It's ideal for users prioritizing privacy, offline operation, or avoiding vendor lock-in.