monalisa@menezes — portfolio — zsh

Jarbas

TypeScriptNode.js 24BaileysSQLiteOpenAI

Jarbas is a digital butler that lives on WhatsApp: instead of several separate bots, a single core handles the connection (via Baileys) and intent routing, and each new capability plugs in as an independent module — without rebuilding the whole foundation. Written in TypeScript on Node.js 24, it uses the runtime's native SQLite for persistence, cron for scheduled reminders and reports, and OpenAI models (GPT-4o and Whisper) for intent detection, audio transcription, link summarization, and image analysis. The skills under development include saving and summarizing links, transcribing audio, creating reminders, logging expenses with automatic categorization, estimating calories from meal photos, and generating weekly financial reports. The project is under active development: the foundational architecture is in place and skills are being built incrementally.