Hacktutor is a multimodal AI agent designed to bridge the digital divide by bringing elite-level, personalized tutoring directly to WhatsApp, the world’s most accessible messaging platform. While AI tutoring exists, it often requires expensive apps or high-bandwidth web interfaces. Hacktutor allows students to simply snap a photo of their handwritten math homework or text a complex science question to get immediate, high-quality help. Powered by Google Gemini 2.5 Flash, the bot goes beyond simple Q&A. It uses a sophisticated "System Prompt" architecture to act as a proactive teacher, offering three distinct modes: Navigator (to assess learning styles), Tutor (for step-by-step problem solving and mistake analysis), and Roadmap (for generating long-term study plans). Technically, this project solves a major hurdle in building complex AI agents on serverless infrastructure. Deep reasoning tasks often exceed the 10-second timeout limits of platforms like Vercel. We engineered a custom Asynchronous Event Loop using Upstash Redis queues and Vercel Cron Jobs. This architecture decouples the AI processing from the HTTP response, allowing Gemini to "think" for as long as necessary to generate detailed, formatted lessons without timing out. Hacktutor proves that powerful, multimodal AI can be made accessible to anyone with a phone number.
Category tags: