convez.ai
PluginAutomaçãoAvançado
mikecutillo

Household Digest Engine

mikecutillo/household-digest

Síntese IA de múltiplas fontes em um briefing diário no Discord

Ver no GitHub
InicialPluginScore · de 0 a 100

Pipeline de síntese diária que integra Gmail, Google Calendar, Microsoft 365 e estado interno, processa tudo via Claude e entrega um briefing único no Discord toda manhã. Substitui 8 checagens de apps por uma leitura consolidada de 5 minutos.

Instalação

$npx convez-ai mikecutillo/household-digest

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
0
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 6 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Consolidar informações fragmentadas de email, calendário e notificações em um único resumo matinal
  • Eliminar ruído (marketing, notificações automáticas) mantendo apenas o que importa
  • Automatizar a composição estruturada de briefings com seções consistentes (agenda, overnight, pendências)

Veredito

Projeto maduro de síntese multi-fonte com engenharia de prompts sólida. Demonstra valor prático real — substitui comportamento diário repetitivo. Código privado limita customização, mas arquitetura é replicável e bem documentada.

README

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

Household Digest Engine

An AI-powered daily digest pipeline that ingests email, calendar, news, household events, and internal system state, summarizes everything via Claude, and delivers a single readable briefing to a Discord channel every morning.

This repo is a public overview. The running code is private.


What it is

A household has too many information channels — email across three accounts, shared calendars, school notifications, news, delivery updates, family schedules. This pipeline collapses all of them into a single AI-composed briefing so the day can start with one five-minute read instead of fifteen different app checks.

What it does

  • Ingests from multiple sources — Gmail (multi-account), Google Calendar, Microsoft 365 calendar, news RSS, AIOS internal state, weather APIs
  • Classifies and filters — drops noise (marketing, auto-notifications) and keeps signal (real messages, real events)
  • Composes the digest via Claude with per-section prompts — "today's schedule", "what changed overnight", "who needs a reply", "weekend preview"
  • Delivers to Discord at a scheduled time with a consistent house-style formatting
  • Archives past digests to SQLite for retrospective review

Architecture

graph LR
    Gmail[📧 Gmail × 3] --> Ingest
    Cal[📅 Google Calendar] --> Ingest
    M365[📨 Microsoft Graph] --> Ingest
    News[📰 RSS feeds] --> Ingest
    State[🏠 AIOS state] --> Ingest

    Ingest[🔍 Filter + classify] --> Compose
    Compose[🤖 Claude · per-section prompts] --> Sections

    Sections[📋 Today · Overnight · Replies · Weekend] --> Discord[💬 Discord channel]
    Sections --> Archive[(SQLite archive)]

Software

LayerTech
IngestionPython, Google APIs (Gmail, Calendar), Microsoft Graph, RSS parsers
Classification + summarizationAnthropic Claude
StorageSQLite (digest archive)
DeliveryDiscord webhooks
Schedulinglaunchd

What this demonstrates

  • Multi-source AI summarization — not a single-document summarizer, a cross-domain synthesis engine
  • Prompt engineering for structure — reliable section formatting day after day
  • Practical LLM utility — this replaces a real daily behavior (me checking 8 apps at 7am)

Stack

Python Anthropic Claude Google APIs Microsoft Graph SQLite Discord

Related in the AIOS Portfolio

  • Google AI Toolkit — Gmail + Drive AI toolkit; multi-account OAuth, email analysis, cleanup automation
  • M365 AI Toolkit — Microsoft 365 AI toolkit; mailbox analysis, OneDrive management via Graph API
  • BMO Discord Agent — Discord-native family AI companion; the delivery channel for the morning brief

Part of the AIOS portfolio. See the profile README for the full system map.

Ver o original no GitHub

Bom para

Relacionados