Multi-channel personal finance AI assistant built with Nuxt 3, Vercel AI SDK, Google Gemini Multimodal, Evolution API, and a custom anti-ban WhatsApp humanization engine.


Mahsoob is an intelligent, voice-first personal finance and expense intelligence platform designed to eliminate the friction of manual financial tracking through multi-channel conversational agents. Built full-stack with Nuxt 3, Nitro, and MySQL, Mahsoob transforms unstructured voice notes and text messages sent via WhatsApp and Telegram into categorized ledger transactions and actionable budgeting insights. Architected with the Vercel AI SDK and Google Gemini Multimodal models, the system seamlessly parses multilingual audio notes and receipts with dynamic model switching. To ensure robust WhatsApp messaging, it runs on a self-hosted Evolution API client paired with a custom-engineered Humanization Algorithm that dynamically simulates human typing rhythms, variable response delays, and realistic interaction pacing to evade rate limits and prevent automated platform blocks.
Multimodal Voice & Text Transaction Parsing: Ingests raw audio voice notes and text receipts from WhatsApp and Telegram, automatically extracting amounts, merchants, and currencies.
Automated Expense Classification & Budget Allocation: AI categorization engine distributing transactions into custom budgetary buckets and expense categories in real time.
Custom WhatsApp Anti-Ban Humanization Engine: Proprietary algorithm simulating realistic typing states and variable response latencies tailored to message length to avoid automated bans.
Vercel AI SDK & Model Portability: Model-agnostic AI pipeline leveraging Google Gemini Multimodal models with zero vendor lock-in for future LLM migrations.
Multi-Channel Ingestion (WhatsApp & Telegram): Self-hosted Evolution API client and Telegram Bot API synchronization enabling seamless on-the-go conversational logging.
Full-Stack Nuxt 3 & MySQL Architecture: Real-time analytics dashboard built with Nuxt Nitro backend, secure API routes, and structured relational MySQL storage.