DojoGenesis Gateway：模块化Agentic AI运行时平台

DojoGenesis Gateway is a self-hosted Agentic AI runtime platform written in Go, featuring 19 independently version-controlled modules. It delivers enterprise-grade AI agent infrastructure with key capabilities including multi-provider LLM routing, DAG task orchestration, WASM sandbox execution, and real-time observability, all packaged as a single binary for simplified deployment.

The platform consists of over 140,000 lines of Go code distributed across 19 modules. Unlike many AI frameworks that only wrap basic APIs, it is designed from the ground up to meet enterprise deployment needs—including high availability, security, scalability, and observability—aiming to provide a complete, composable, production-ready AI agent infrastructure.

Gateway adopts a modular microkernel architecture with the core principle of 'independent version control, independent testing, independent import'. Each module has its own go.mod file and evolves independently while collaborating via Go workspace. This design enables independent evolution, fine-grained dependency management, flexible composition, and isolated testing.

Module Layers

• Infrastructure: Shared core data structures (Message, ToolCall) and SSE event system for observability.
• Service: HTTP server (OpenAI-compatible API, Chat UI), multi-provider plugin system (8 providers like OpenAI, Anthropic), and tool execution engine (33 built-in tools).
• Intelligence: Dialogue memory (semantic compression), MCP protocol integration, DAG orchestration engine, content-addressed skill system (89 skills), and agent behavior configuration.
• Runtime: Actor supervision tree, CAS storage, WASM sandbox, and event bus.
• Application: Workflow engine and integration test framework.

Key Features

• Multi-provider Routing: Intent-based model selection, explicit override, and failover.
• DAG Orchestration: Visual design, conditional branching, parallel execution, and error handling.
• WASM Sandbox: Isolated, high-performance execution of third-party code.
• Real-time Observability: SSE-based log streams, performance metrics, and distributed tracing.

Deployment Methods

• Source Build: Clone repo, configure .env, build and run.
• Docker: Use docker-compose for deployment.
• Precompiled Binary: Download and execute directly.

Application Scenarios

• Unified AI gateway for enterprise internal services.
• Agent platform for complex AI applications.
• Workflow automation for business processes.
• Edge AI deployment on lightweight devices.
• Multi-tenant SaaS with customized configurations.

Why Go?

• Performance and resource efficiency via goroutines.
• Simple deployment as a single static binary.
• CSP concurrency model for high-concurrency services.
• Rich ecosystem in cloud-native and microservices.

Modular vs Monolithic

Gateway uses a 'modular monolith' approach: single deployable unit with internal modularity, avoiding microservice complexity while maintaining code organization.

Comparison with Similar Tools

Feature	Gateway	LangChain	LlamaIndex	Haystack
Language	Go	Python	Python	Python
Deployment	Single binary	Library dependency	Library dependency	Library dependency
Multi-provider	Built-in 8	Config required	Config required	Config required
WASM Sandbox	Built-in	No	No	No
Edge Deployment	Friendly	Heavy	Heavy	Heavy

Community

• Open-source under MIT license.
• Contribution areas: New LLM providers, tool extensions, skill library enrichment, documentation, performance optimization.

Roadmap

• Support for more providers (Cohere, AI21).
• Distributed deployment mode.
• Enhanced RAG capabilities.
• Improved visual workflow editor.
• Enterprise security features (RBAC, audit logs).

DojoGenesis Gateway represents a systematic approach to building production-grade Agentic AI infrastructure. It is not just an API wrapper but a complete solution covering architecture to applications. For teams seeking self-hosted AI infrastructure, it offers a feature-rich, high-performance, and easy-to-deploy option. As AI moves from experimentation to production, such infrastructure projects will play an increasingly critical role in enterprise AI ecosystems.