CoreSRE

A distributed AI agent orchestration and collaboration platform built on the A2A protocol. CoreSRE serves as the upstream orchestration layer for enterprise AI operations — managing agent registration, discovery, scheduling, workflow execution, and multi-agent collaboration.

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                     Frontend (React 19 + shadcn/ui)             │
│  Agent Registry · Workflow Designer · Tool Manager · Chat UI    │
├─────────────────────────────────────────────────────────────────┤
│                   API Gateway (ASP.NET Core 10)                 │
├─────────────────────────────────────────────────────────────────┤
│              Application Layer (CQRS / MediatR)                 │
├──────────┬──────────┬──────────┬──────────┬─────────────────────┤
│ Agent    │ Workflow │ Tool     │ Data     │ Sandbox             │
│ Registry │ Engine   │ Gateway  │ Sources  │ (K8s Pods)          │
├──────────┴──────────┴──────────┴──────────┴─────────────────────┤
│                      Domain Layer (DDD)                         │
├─────────────────────────────────────────────────────────────────┤
│                   Infrastructure Layer                          │
│  EF Core · A2A Protocol · MCP · External API · OTel             │
├─────────────────────────────────────────────────────────────────┤
│              .NET Aspire AppHost (Orchestration)                │
│            PostgreSQL (pgvector) · MinIO · Dashboard            │
└─────────────────────────────────────────────────────────────────┘

Tech Stack

Layer	Technology
Frontend	React 19, TypeScript 5.9, Vite 7, Tailwind CSS 4, shadcn/ui, AG-UI Protocol
Backend	.NET 10, ASP.NET Core Minimal API, MediatR (CQRS), EF Core 10
Database	PostgreSQL 17 + pgvector (semantic search)
Storage	MinIO (S3-compatible object storage)
Orchestration	.NET Aspire (local dev), Kubernetes (production)
Protocols	A2A (Agent-to-Agent), MCP (Model Context Protocol), AG-UI
Observability	OpenTelemetry, Prometheus, Loki, Jaeger, Alertmanager
AI	Microsoft.Extensions.AI, Microsoft Agent Framework

Quick Start

Prerequisites

• .NET 10 SDK
• Node.js 22+ with npm
• Docker Desktop with Kubernetes enabled
• PostgreSQL 17 (provided by Aspire) or a running instance

Development

# Clone and enter the project
git clone https://github.com/MadLongTom/CoreSRE.git
cd CoreSRE

# Install frontend dependencies
cd Frontend && npm install && cd ..

# Start everything (Aspire + Frontend)
.\dev.ps1

Demo Deployment (Kubernetes)

# Deploy observability stack + demo microservices + register data sources
.\deploy-demo.ps1

# Tear down everything
.\deploy-demo.ps1 -TearDown

Deploys to local Docker Desktop K8s:

• Observability namespace: Prometheus, Loki, Jaeger, Alertmanager
• demo-app namespace: order/payment/inventory services + traffic generator

Project Structure

CoreSRE/
├── Backend/
│   ├── CoreSRE/                    # Web API (endpoints, hubs, middleware)
│   ├── CoreSRE.AppHost/            # .NET Aspire orchestrator
│   ├── CoreSRE.Application/        # Use cases (CQRS commands & queries)
│   ├── CoreSRE.Domain/             # Entities, enums, value objects, interfaces
│   ├── CoreSRE.Infrastructure/     # EF Core, services, persistence, migrations
│   ├── CoreSRE.ServiceDefaults/    # Shared Aspire defaults (telemetry, health)
│   └── CoreSRE.*.Tests/            # Unit & integration tests
├── Frontend/                       # React SPA
├── k8s/                            # Kubernetes manifests
│   ├── demo-app/                   # Business microservices
│   └── observability/              # Monitoring stack
├── docs/                           # PRD, BRD, design docs, specs
├── specs/                          # Spec-driven development specs (per feature)
├── dev.ps1                         # One-command dev launcher
├── deploy-demo.ps1                 # K8s demo deployment script
└── CONSTITUTION.md                 # Project development constitution

Key Features

• Agent Registry — Register, discover, and manage AI agents with semantic search (pgvector)
• Workflow Engine — Visual workflow designer with data-flow execution, conditional branching, and real-time SignalR push
• Multi-Agent Chat — Team orchestration (MagneticOne pattern) with LLM selector group chat
• Tool Gateway — MCP tool discovery, OpenAPI import, and dynamic tool binding
• Data Sources — Integrate Prometheus, Loki, Jaeger, Alertmanager, Kubernetes, and Git repositories
• Sandbox — Isolated code execution in Kubernetes pods with terminal access (xterm.js + WebSocket)
• Agent Skills — Persistent sandbox-backed skill execution for agents
• AG-UI Protocol — Standard streaming protocol for agent-frontend communication

Development Principles

This project follows the Project Constitution:

1. TDD — Tests first, implementation second
2. DDD — Domain model is the single source of truth
3. SDD — Interface contracts before implementation

License

Private repository. All rights reserved.