Smart speaker DIY inspirado no Echo Dot, com ESP32-S3, integrado ao Home Assistant e OpenClaw (Quasar) para controle por voz de toda a casa.
O QuasarBox Γ© um speaker compacto com microfone always-listening que detecta uma wake word localmente, envia o Γ‘udio para processamento STT no servidor, interpreta o comando via IA (Claude/OpenClaw) e executa aΓ§Γ΅es no Home Assistant β tudo com resposta por voz.
βββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β QuasarBox (ESP32-S3) β
β Mic βββΊ Wake Word (local) βββΊ Streaming Γudio ββββββΌβββΊ Wi-Fi
β Speaker βββ Γudio TTS ββββββββββββββββββββββββββββββΌβββ Wi-Fi
βββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β
βΌ
βββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β DeskFelipeDell (Servidor) β
β β
β Home Assistant (Docker) βββΊ Voice Pipeline β
β βββ STT: Whisper (Docker) β
β βββ Conversation Agent: OpenClaw (Quasar/Claude) β
β βββ TTS: Piper (Docker) β
β βββ Wake Word: openWakeWord (Docker, backup) β
β β
β OpenClaw βββΊ Interpreta comando βββΊ Executa aΓ§Γ£o β
β βββ Home Assistant API (luzes, sensores, etc.) β
β βββ LG TV Controller (TV da sala) β
β βββ Qualquer integraΓ§Γ£o futura β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββ
- Arquitetura
- Hardware β Lista de Componentes
- Software β Stack Completo
- Sprint 0 β Setup do Servidor (HA Voice Pipeline)
- Sprint 1 β ProtΓ³tipo Hardware
- Sprint 2 β Firmware ESPHome
- Sprint 3 β Conversation Agent (OpenClaw)
- Sprint 4 β Wake Word
- Sprint 5 β Case e Acabamento
- Sprint 6 β Multi-Room
- Sprint 7 β IntegraΓ§Γ£o HA β Dispositivos
- Sprint 8 β PCB Custom
- Hardware Design (ReferΓͺncia)
- ReferΓͺncias
| Sprint | DescriΓ§Γ£o | Status |
|---|---|---|
| 0 | Setup servidor (HA Voice Pipeline via Docker) | π² NΓ£o iniciado |
| 1 | ProtΓ³tipo hardware (1 unidade, protoboard) | π² NΓ£o iniciado |
| 2 | Firmware ESPHome | π² NΓ£o iniciado |
| 3 | Conversation Agent (OpenClaw) | π² NΓ£o iniciado |
| 4 | Wake Word (provisΓ³ria + custom) | π² NΓ£o iniciado |
| 5 | Case 3D e acabamento | π² NΓ£o iniciado |
| 6 | Multi-room (5 cΓ΄modos) | π² NΓ£o iniciado |
| 7 | IntegraΓ§Γ£o HA β dispositivos smart | π² NΓ£o iniciado |
| 8 | PCB custom (design + fabricaΓ§Γ£o) | π² NΓ£o iniciado |
| Item | PreΓ§o Est. |
|---|---|
| ESP32-S3-DevKitC-1 N16R8 | ~R$40 |
| INMP441 breakout | ~R$15 |
| MAX98357A breakout | ~R$18 |
| Speaker 3W 4Ξ© 40mm | ~R$10 |
| WS2812B ring 8 LEDs | ~R$5 |
| Protoboard + jumpers | ~R$16 |
| Cabo USB-C | ~R$10 |
| Total protΓ³tipo | ~R$115 |
| Item | Qtd | Total |
|---|---|---|
| Componentes eletrΓ΄nicos (5x) | 5 | ~R$440 |
| PCB custom (JLCPCB, 5 un.) | 5 | ~R$50-120 |
| Fontes USB-C 5V 2A | 5 | ~R$100 |
| Filamento PLA (cases 3D) | 1 | ~R$60 |
| Dispositivos smart (lΓ’mpadas, plugs) | β | ~R$150-280 |
| Total estimado (5 unidades + devices) | ~R$800-1000 |
- Servidor: DeskFelipeDell (Ubuntu, headless, Docker)
- Home Assistant: Rodando em Docker,
localhost:8123 - OpenClaw (Quasar): InstΓ’ncia ativa com acesso a HA API
- Whisper: JΓ‘ instalado (
whisper-transcribe, modelo small, ~7s/3s Γ‘udio) - LG TV Controller: API local em
:8888 - Rede: Wi-Fi 2.4GHz em toda a casa
MIT