convez.ai
MCP ServerDevOps & InfraestruturaAvançado
bigduu

Zenith: Orquestrador de Monorepo para Bodhi AI

bigduu/Zenith

Orquestrador monorepo que sincroniza produto, UI, runtime, backend e docs

Ver no GitHub
InicialPluginScore · de 0 a 100

Monorepo umbrella que centraliza e coordena o Bodhi AI — um agente desktop local-first que transforma IA em workbench produtiva. Integra cinco submodules (runtime Rust, UI React, shell Tauri, backend Go, docs) com versionamento sincronizado e release train automatizado. Serve como single entry point arquitetural para todo o ecossistema.

Instalação

$npx convez-ai bigduu/Zenith

1 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

Estrelas
6
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 1 dia
Versão
—
Issues abertas
22
Contribuidores
3

Casos de uso

  • Manter múltiplos componentes (desktop, UI, runtime, backend) sincronizados em releases
  • Entender a arquitetura completa de um agente desktop local-first
  • Automatizar versionamento e pipeline de deployment coordenado entre repos

Veredito

Implementação sólida de orquestração monorepo com coordenação de releases inteligente e estrutura arquitetural clara. Mais relevante para quem trabalha com infraestrutura de CI/CD multi-repositório do que para uso direto como agente IA.

README

A tradução para o português deste README ainda está na fila.

Zenith

📖 中文版请看 README.zh-CN.md

In one line: the single entry point that ties a "get-things-done" AI desktop product together with the entire system behind it.

Bodhi AI turns AI from a chat box into a desktop workbench that actually does the work: you hand it a task, it uses tools, keeps memory, and produces results — and you can watch the whole thing happen, not just read a final answer. Zenith is the home base that holds the desktop product, the UI, the execution engine, the backend, and the docs together — and keeps their releases in sync.

First time here? Start with Bodhi AI.

Bodhi AI · Lotus · Bamboo · Bodhi Server · Pavilion · Architecture Overview


Key capabilities at a glance

CapabilityWhat it means
A map of the whole systemOne repo shows how product, UI, runtime, backend, and docs divide the work and fit together
Five submodules, one clonePull the full stack in a single recursive clone
Coordinated release trainBamboo → Lotus → Bodhi published in order, all driven by one config file
Daily nightly versioningCalendar-versioned (YYYY.M.N) auto-bump and nightly release
Submodule guardCI validates submodule pointers on every push and PR
Clear "start here" routingWhether you want the product, the frontend, or the runtime, there is a clear door in

Architecture

Zenith holds almost no business logic itself. It is a thin-shell monorepo: it pins five Git submodules, owns the root-level documentation, and orchestrates releases across repos. The real features live inside the submodules.

graph TD
  Z["Zenith (this repo)<br/>submodule pointers + release train"]

  Z --> B["Bodhi AI<br/>desktop product surface (Tauri shell)"]
  Z --> L["Lotus<br/>React + Vite UI layer"]
  Z --> R["Bamboo<br/>local-first Rust agent runtime"]
  Z --> S["Bodhi Server<br/>Go backend"]
  Z --> P["Pavilion<br/>website & docs"]

  B -. embeds .-> L
  L -. HTTP / SSE .-> R
  R -. auth · quota · LLM proxy .-> S
  P -. explains .-> B

Note —— The Bodhi shell hosts the UI and native integration; Lotus is the actual UI; Lotus talks to the Bamboo runtime over HTTP / SSE (not Tauri IPC); Bamboo defers account, quota, billing, and LLM-proxy concerns to the Go backend, Bodhi Server.

What each module does

ModulePathRoleStart here
Bodhi AIbodhi/Product surface: Tauri desktop shell hosting the UI, native integration, packagingBodhi AI
Lotuslotus/UI layer: React + Vite frontend, live event stream, view state, settingsLotus
Bamboobamboo/Execution engine: local-first Rust agent runtime — tasks, tools, memory, HTTP/SSE APIBamboo Agent
Bodhi Serverbodhi-server/Backend: Go server — auth, persistence, quota/billing, LLM proxyBodhi Server
Pavilionpavilion/Website & docs: download page, doc center, public narrativePavilion
Zenith (root).Coordinator: submodule pointers, root docs, release trainYou are here

Signature deep-dives

Start here routing

Zenith's biggest job is getting any person to the right door fast.

If you just want to understand the product

If you want to build

  • Desktop product / Tauri shell → bodhi/
  • Frontend interaction / React UI → lotus/
  • Agent runtime / Rust backend → bamboo/
  • Backend / Go backend → bodhi-server/
  • Website / docs / public content → pavilion/

The stack, organized on purpose

The five-way split exists so each layer can evolve on its own yet converge into one product at release time:

  • UI & experience live in Lotus (React/Vite) and iterate and test independently.
  • Execution lives in Bamboo (Rust), local-first, runnable as a standalone HTTP service.
  • Account, quota, billing, LLM proxy live in Bodhi Server (Go) — the server-trusted concerns, visible as real modules under bodhi-server/internal/ (auth, quota, pricing, proxy, database).
  • The desktop shell lives in Bodhi and only wraps the UI into an installable desktop app.
  • Public narrative lives in Pavilion, decoupled from code.

Coordinated release train

Shipping several repos at once, in dependency order, is error-prone. Zenith reduces it to one config plus one workflow.

The order is fixed:

  1. Bamboo → publish crate (publish-crate.yml in bigduu/Bamboo-agent)
  2. Lotus → publish npm package @bigduu/lotus (publish-npm.yml in bigduu/Lotus)
  3. Bodhi → build & ship desktop assets (deploy.yml in bigduu/Bodhi-AI)

Between steps, the train waits until the artifact is actually visible on crates.io / npm before continuing (see wait_for_crates_version / wait_for_npm_version in release-train.yml). All versions and refs default to .github/release-train.config.json (from_manifest), and can be overridden when dispatched manually.

Current config (.github/release-train.config.json):

{
  "refs":     { "bamboo": "main", "lotus": "main", "bodhi": "main" },
  "versions": { "release": "2026.6.2", "bamboo": "2026.6.2", "lotus": "2026.6.2", "bodhi": "2026.6.2" },
  "options":  { "lotus_skip_tests": false }
}

Related workflows (under .github/workflows/):

WorkflowPurpose
release-train.ymlManual coordinated release: Bamboo → Lotus → Bodhi
nightly-release.ymlDaily nightly auto-bump (YYYY.M.N) at 04:00 UTC
submodule-guard.ymlValidates submodule pointers on push & PR

Default policy —— Normal releases go through Zenith's release train; per-repo standalone flows are for recovery or special cases only.


Quick start / Development

Clone the full stack

git clone --recursive https://github.com/bigduu/Zenith.git
cd Zenith

Already cloned without submodules:

git submodule update --init --recursive

Run the desktop app

cd bodhi
npm install
npm run tauri:dev

Bodhi's web:dev / tauri:dev drive the Vite frontend in ../lotus; see bodhi/package.json.

Run the UI on its own

cd lotus
npm install
npm run dev

Run the agent runtime

cd bamboo
cargo run -- serve --port 9562

bamboo serve accepts optional --port / --bind / --data-dir / --static-dir / --workers overrides; defined in bamboo/src/bin/bamboo.rs. Without --port it uses the configured port. (Run bamboo --help for the other subcommands: config, -p headless, actor, broker, broker-agent.)

Manage submodule pointers

# show pinned revisions
git submodule status

# pull latest upstream commits
git submodule update --remote --recursive

# after submodule work, bump pointers from root
git add .gitmodules bamboo bodhi lotus pavilion bodhi-server
git commit -m "chore: bump submodule pointers"
git push

Workflow: Develop, commit, and push inside the submodule first, then bump and commit the pointer in Zenith. Full release steps live in AGENTS.md.


The rest of the stack

ModuleRepository
Bodhi AI — desktop product surfacehttps://github.com/bigduu/Bodhi-AI
Lotus — React UI layerhttps://github.com/bigduu/Lotus
Bamboo — Rust agent runtimehttps://github.com/bigduu/Bamboo-agent
Bodhi Server — Go backendhttps://github.com/bigduu/bodhi-server
Pavilion — website & docshttps://github.com/bigduu/Pavilion

Key docs


Only clicking one link? Open Bodhi AI. Want the why? Read the Zenith Architecture Overview.

Ver o original no GitHub

Bom para

Relacionados