Globular Services
globulario/services
Plataforma distribuída gRPC-nativa com orchestração por IA para clusters self-hosted
Ver no GitHubPlataforma de microsserviços open-source para aplicações distribuídas auto-hospedadas. Construída em gRPC com Protocol Buffers, opera via um modelo de estado explícito de 4 camadas (Repository → Desired → Installed → Runtime) e inclui servidor MCP com 129+ ferramentas para operação assistida por IA.
Instalação
npx convez-ai globulario/services1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Implantar e operar clusters de microsserviços distribuídos
- Orquestração de fluxos de trabalho centralizados com convergência de estado
- Diagnóstico e auto-cura de clusters via AI operators
- Gerenciamento de pacotes, backups e DNS em ambientes distribuídos
- Integração com agentes Claude via MCP para operações autônomas
Veredito

Projeto maduro e ambicioso para quem precisa de infraestrutura distribuída real com operação assistida por IA. Requer expertise em sistemas distribuídos, gRPC e Go; ideal para times que buscam alternativa opensource a Kubernetes com integração nativa a agentes Claude.
README
A tradução para o português deste README ainda está na fila.
Globular Services
Microservices platform for self-hosted distributed applications. Built on gRPC with Protocol Buffers, running as native Linux binaries under systemd. etcd as the single source of truth. Workflow-driven convergence across a 4-layer state model (Repository → Desired → Installed → Runtime).
Architecture
Proto definitions (proto/)
↓ generateCode.sh
Go services (golang/) + TypeScript clients (typescript/)
↓ build-all-packages.sh
Package archives (.tgz)
↓ globular-installer
Running cluster (systemd + etcd + Envoy)
Repository Structure
services/
├── proto/ # 38 Protocol Buffer definitions (source of truth)
├── golang/ # All Go microservices (33 services + CLI + MCP)
│ ├── cluster_controller/ # Cluster management, desired state, workflows
│ ├── node_agent/ # Node-local executor, package management
│ ├── workflow/ # Centralized workflow execution engine
│ ├── repository/ # Package artifact registry (MinIO-backed)
│ ├── authentication/ # JWT token management
│ ├── rbac/ # Role-based access control
│ ├── dns/ # Authoritative DNS + zone management
│ ├── monitoring/ # Prometheus adapter
│ ├── backup_manager/ # Distributed backup orchestration
│ ├── cluster_doctor/ # Health analysis + auto-heal
│ ├── ai_memory/ # Persistent AI knowledge (ScyllaDB)
│ ├── ai_executor/ # AI diagnosis + remediation
│ ├── ai_watcher/ # Event-driven incident detection
│ ├── ai_router/ # Dynamic routing policies
│ ├── compute/ # Distributed batch job execution
│ ├── domain/ # External domain + ACME cert management
│ ├── globularcli/ # CLI tool
│ ├── mcp/ # Model Context Protocol server (129+ tools)
│ ├── globular_service/ # Shared service primitives
│ ├── globular_client/ # Shared client primitives
│ ├── interceptors/ # gRPC auth, RBAC, audit middleware
│ ├── config/ # etcd-backed configuration
│ ├── security/ # TLS, PKI, JWT, Ed25519 keystore
│ └── ... # 15+ additional services
├── typescript/ # TypeScript gRPC-Web client library
├── generated/ # Generated specs, packages, policies
├── generateCode.sh # Proto → Go/TypeScript code generation
└── build-all-packages.sh # Full package build pipeline
Quick Start
Build from Source
# Requires: Go 1.24+, protoc, protoc-gen-go
# Generate code and build all services
bash generateCode.sh
# Build all packages (infrastructure + services)
bash build-all-packages.sh
Run Tests
cd golang
go test ./... -race
Build a Specific Service
cd golang
go build ./echo/echo_server
Services
Control Plane
| Service | Port | Description |
|---|---|---|
| Cluster Controller | 12000 | Cluster management, desired state, membership, workflow dispatch |
| Node Agent | 11000 | Local executor, package tracking, service control |
| Workflow Service | 10004 | Centralized workflow execution and tracking |
| Cluster Doctor | 12005 | Health analysis, drift detection, auto-heal |
Infrastructure
| Service | Port | Description |
|---|---|---|
| etcd | 2379/2380 | Distributed configuration and state store |
| MinIO | 9000 | Object storage (packages, backups, artifacts) |
| Envoy Gateway | 443/8443 | TLS termination, xDS routing, gRPC-Web |
| Prometheus | 9090 | Metrics collection |
| Alertmanager | 9093 | Alert routing |
| ScyllaDB | 9042 | High-throughput data (AI memory, DNS storage) |
Core Services
| Service | Port | Description |
|---|---|---|
| Authentication | 10101 | JWT tokens, password management |
| RBAC | 10104 | Permission enforcement |
| Event | 10102 | Publish-subscribe event bus |
| File | 10103 | File management |
| DNS | 10006 | Authoritative DNS, zone management |
| Discovery | 10029 | Service discovery, install plans |
| Repository | — | Package artifact registry |
| Resource | — | Package descriptors, accounts, groups |
| Log | 10100 | Centralized logging |
AI Services
| Service | Port | Description |
|---|---|---|
| AI Memory | 10200 | Persistent knowledge store (ScyllaDB) |
| AI Watcher | 10210 | Event monitoring, incident detection |
| AI Router | 10220 | Dynamic routing policy computation |
| AI Executor | 10230 | Incident diagnosis and remediation |
Operational
| Service | Port | Description |
|---|---|---|
| Monitoring | 10019 | Prometheus API adapter |
| Backup Manager | 10040 | Backup orchestration |
| MCP Server | 10260 | AI agent interface (129+ diagnostic tools) |
| Domain Reconciler | — | External domain + ACME cert management (runs in controller) |
Application Services
| Service | Description |
|---|---|
| Persistence | MongoDB access layer |
| Storage | Key-value store (BadgerDB) |
| Search | Full-text search (Bleve) |
| Media | Audio/video management |
| Title | Metadata service |
| SMTP email | |
| LDAP | LDAP authentication provider |
| SQL | SQL database access |
| Blog | CMS engine |
| Conversation | Chat management |
| Catalog | Component catalog |
| Torrent | Torrent downloads |
Documentation
Full documentation is in docs/:
- Getting Started — From zero to running cluster
- Architecture — How components interact
- Day-0/1/2 Operations — Complete lifecycle
- Building from Source — Build process
- AI Layer — AI services, rules, and agent model
- Developer Guide — Local-first development
Key Design Principles
- etcd is the single source of truth — no environment variables, no hardcoded addresses
- Workflow-driven convergence — all state changes go through the workflow engine
- 4-layer state model — Repository → Desired → Installed → Runtime (never collapsed)
- Native binaries under systemd — no containers required
- Local-first — services run standalone with
go run, no cluster needed for development
License
See LICENSE for details.