AIStatusHub
CarltonYu/AIStatusHub
Hub local que centraliza status de múltiplas ferramentas de IA e roteia para displays
Ver no GitHubHub local que unifica atividade de ferramentas de IA (Claude Code, Codex, Kimi Code, VS Code) e roteia para dashboards, webhooks, MQTT, BLE e displays de hardware. Funciona como agregador de estado em tempo real com suporte a proxy compatível com OpenAI e Anthropic.
Instalação
npx convez-ai CarltonYu/AIStatusHub1 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
- Monitorar múltiplas ferramentas de IA simultaneamente em um só lugar
- Integrar status de IA com hardwares customizados (displays OLED, etc)
- Rotear eventos de IA para webhooks e sistemas de automação
Veredito

Solução prática para desenvolvedores que usam múltiplas ferramentas de IA e precisam de visibilidade centralizada. Bem arquitetado com suporte a diversos protocolos (HTTP, WebSocket, MQTT) e proxies compatíveis, mas requer conhecimento técnico para setup e customização de hardware.
README
A tradução para o português deste README ainda está na fila.
AIStatusHub
AIStatusHub is a lightweight local status hub for AI coding tools.
It is designed as a small Rust executable that reads one config.toml, keeps only current task state by default, and can forward compact state to another AIStatusHub instance.
Development
cargo run -- --config config.example.toml
Default address:
http://127.0.0.1:17888
If startup fails with another instance may already be running, AIStatusHub is
already bound to the configured port. Check or stop it on Windows:
Get-Process aistatushub
Get-Process aistatushub | Stop-Process
.\target\debug\aistatushub.exe --config config.toml
Current MVP Support
- Generic HTTP ingest for local AI tools and hooks.
- Current-state aggregation in memory.
- Minimal built-in web page.
- WebSocket state stream.
- JSON snapshot, when enabled in config.
stdoutoutput for test-stage printing.- IrisOLED UDP output for a Milk-V Duo + GC9A01 face display.
aistatushub reportfor Codex hooks, VSCode tasks, and manual test events.- OpenAI-compatible
/v1/chat/completions,/v1/models, and/v1/*proxy for Hermes Agent / Mimo-style clients. - Anthropic-compatible
/anthropic/v1/messagesproxy for VSCode ClaudeCode. - Kimi
/kimi/v1/chat/completionsproxy for VSCode Kimi Code. - Hub-to-Hub receive endpoints.
Codex/Claude/Kimi can be connected through hooks or local proxy routes. Current VSCode test setup uses:
- ClaudeCode extension:
ANTHROPIC_BASE_URL=http://127.0.0.1:17888/anthropic - Codex extension: user-level
~/.codex/config.tomlhooks, or the built-in local session monitor - Kimi Code extension:
KIMI_BASE_URL=http://127.0.0.1:17888/kimi/v1
Codex hooks are enabled by default, but Codex requires non-managed command hooks
to be reviewed once before they run. In the VSCode Codex panel, type /hooks,
review the AIStatusHub hook entries, then trust them. If no Codex callbacks
arrive, first check /hooks and /status to confirm the active config and
project trust state.
Some Codex desktop/VSCode builds do not expose /hooks in their chat UI. For
that case, enable the local monitor instead:
[monitors.codex_local]
enabled = true
source = "codex-local"
sessions_dir = ""
poll_interval_ms = 1500
idle_after_ms = 30000
max_scan_files = 5000
The monitor reads Codex's local sessions directory and treats recent session
file updates as AI activity. It only reports busy/idle transitions, so it stays
small and does not store conversation content.
Hermes Agent can be tested by enabling [proxy] in AIStatusHub config, then pointing Hermes at http://127.0.0.1:17888/v1:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:17888/v1
hermes config set model.default mimo-v2.5-pro
If AIStatusHub proxy.api_key is empty, Hermes must send a Bearer token, for example through OPENAI_API_KEY in ~/.hermes/.env.
Duo Face Output
AIStatusHub can drive the irisoled-face daemon running on the Duo board. When
any local AI task is active, it sends default angry; when all local AI tasks
finish, it sends normal.
[[outputs]]
id = "duo-face"
type = "irisoled-udp"
enabled = true
url = "udp://192.168.42.1:25250"
busy_expression = "angry"
idle_expression = "normal"
include_remote_states = false
The repository also includes .vscode/tasks.json for manual source events.