Building an Open-Source, Multi-Agent Study Companion for My Sister

This submission is for the Hacktoberfest Weekend Challenge, where the creator built an open-source, multi-agent educational assistant called Bud AI. The assistant is designed specifically for college students and the creator's sister. Studying for college exams can be mentally exhausting, involve a lot of jumping between tasks, and present large amounts of information in one place. Bud AI aims to address these common issues students face.

Three major problems Bud AI solves for students are context window overhead, emotional and focus tracking, and on-demand tool discovery. Context window overhead refers to the difficulty of dealing with large study guides by breaking them down into smaller, easier-to-manage pieces using a specialized 'Teacher Persona.' Bud AI tracks students' emotions and focus levels in real-time through a high-contrast, glassmorphic UI that also displays encouraging visual mascot animations during late-night study sessions.

On-demand tool discovery allows the assistant to fetch relevant learning resources on the fly without leaving the workspace. The Bud AI live app can be accessed at https://bud-ai-rho.vercel.app. The code repository for Bud AI is located at jouzia / Bud-AI.

The project includes a stateful multi-agent AI tutor with memory, adaptive learning capabilities, and reward-driven educational benchmarking. Bud AI is designed to be a production-ready, multi-agent conversational assistant that runs on a stateful, reward-driven benchmark environment, specifically tailored for complex educational task execution.

The application layer consists of Bud AI built on a Next.js/Python custom web interface. The core engine is powered by OpenEnv Study Intelligence, which operates within a stateful agent environment. Currently, Bud AI is fully functional and deployable through Hugging Face Spaces and Vercel compatibility.

Bud AI presents a high-contrast, ultra-modern developer interface following a strict dark-mode layout. This design incorporates a brutalist architecture with deep near-black backgrounds (#0B0B0F) to reduce cognitive fatigue. The UI components are built as translucent, frosted glass panels with smooth backdrop blur filters and tiny white borders. Subtle neon color underglows signal different multi-agent execution states, like active routing or content streaming.

The core multi-agent logic and routing architecture move beyond monolithic single-prompt architectures. Bud AI utilizes an internal intent routing core that handles tool discovery, context optimization, and persona switching using a custom Multi-Agent Orchestration pipeline with Model Context Protocol (MCP) concepts. These components work together to fetch targeted learning resources, adjust the study focus, and switch between different study personas as needed.

Open innovation and open-weight models are crucial to Bud AI's development. Open innovation and open-weight models allow students to see precisely how their study data is handled and enable developers to fine-tune models for specific university syllabi. Using open frameworks also prevents lock-in to proprietary closed APIs that can change pricing or deprecate features without warning.

By swapping model backends (from open-weight local models to cloud APIs) seamlessly using structured Pydantic schemas, Bud AI benefits from community-driven extensions where students and developers can fork the repository, contribute new agent tools, or adapt the 'Teacher Persona' to their college curriculum.