Niamon

Mobile interface
Web / Desktop Interface

Inspiration

It was an old project to create a chat bot that could extensively interact with the digital world around it and the people in it

What it does

It’s an interactive AI assistant designed to help university students manage their academic and campus life more efficiently. The system allows students to talk to a bot and get responses back a conversational interface powered by curated tools focused on supporting studying, scheduling, and campus resources.

Currently, it integrates with the SFU Course API as a test environment, pulling real data such as tuition details, academic deadlines, term schedules (fall, spring, summer), exam calendars, refund dates, housing information, FAQs, and campus directions. The same framework can later scale to other universities by connecting to their public APIs.

In the long term, the AI will expand to include personalized calendar summaries, Canvas and student ID integration, secure local data storage, and optional cloud syncing for cross-device access. A mobile interface is planned for daily convenience, while Discord integration enables community features such as study groups and event organization.

The assistant is also designed for scalability and personalization users can customize the AI’s persona, voice, or visual model (e.g., Live2D avatars, custom models, or alternate “alter egos”).

Finally, the backend leverages Gemini 2.5 Flash for natural language understanding and API processing, providing fast, context-aware responses.

How we built it

We built a full-stack Svelte application with a Tauri backend, allowing it to run as a standalone desktop app and later expand to mobile. The Tauri layer handles native system integration and secure local data management, while the Svelte frontend provides a fast, reactive user interface.

The app connects to various API providers through a unified chat interface that enables the AI to access course data, scheduling tools, and other student-related resources. This includes integration with the Gemini API for intelligent responses and natural-language understanding.

For visuals, we implemented Live2D to give the AI an expressive, animated avatar adding a more human and engaging personality to the experience.

Our tech stack includes Svelte, Tailwind, Skeleton.dev, TypeScript, Rust, Tauri, Vercel, GitHub, and VS Code, with supporting tools like Ollama, ElevenLabs, and FontAwesome. The frontend uses HTML, CSS, and JavaScript, while .tech serves as the deployment domain.

Together, these technologies form a modular, cross-platform foundation for an AI-powered academic assistant that can eventually integrate with more university APIs and student systems.

Challenges we ran into

Time was our biggest constraint balancing design, implementation, and integration under tight deadlines made it difficult to perfect every feature.

We also faced several technical challenges. Managing API keys and authentication across multiple services (Gemini, ElevenLabs, and university APIs) required careful handling to ensure security and reliability. Additionally, token limits imposed by some APIs restricted how much data we could process or cache at once, forcing us to optimize our requests and reduce response sizes.

Getting the Live2D models fully functional was another hurdle. Implementing smooth animations, expression changes, and real-time reactions while maintaining performance required extensive tuning and debugging. Similarly, ElevenLabs voice integration presented difficulties synchronizing speech with text and timing the AI’s responses correctly was more complex than expected.

Despite these challenges, each obstacle helped refine our understanding of the system’s architecture and pushed us toward building a more robust, scalable foundation for the future.

Accomplishments that we're proud of

We’re proud that our project works the core AI assistant functions as intended, seamlessly pulling university data and responding naturally through the unified chat interface.

One of our biggest wins was getting the Live2D model fully operational. Seeing the AI come to life with expressive animations and dynamic reactions made the experience far more engaging and personal.

We also successfully integrated a custom voice model using ElevenLabs, giving our AI its own unique voice and personality. Combining visuals, speech, and intelligence into one cohesive system was a huge milestone for us and a major step toward creating a truly interactive student assistant.

What we learned

Throughout this project, we learned how to bring together multiple technologies from frontend frameworks to AI models into a cohesive, cross-platform experience. We gained hands-on experience with Svelte and Tauri, learning how to bridge web and native environments to create a responsive desktop application.

We also deepened our understanding of API integration, handling authentication, rate limits, and token management across services like Gemini, ElevenLabs, and Live2D. Working with real-time models taught us the importance of performance optimization and synchronization between animation, text, and voice.

Beyond the technical side, we learned how crucial time management, adaptability, and collaboration are when building something ambitious under pressure. Each challenge pushed us to find creative solutions and helped us grow as developers, designers, and problem-solvers.

What's next for Niamon

Mobile Interface: Expanding beyond desktop, we aim to build a fully responsive mobile version so students can access Niamon anywhere, anytime.

Discord Integration: Bringing Niamon into community spaces like Discord will allow it to assist with study groups, event reminders, and collaborative learning.

AI-Driven Live2D Control: We plan to enable real-time Live2D animation control directly from the AI model — allowing expressions and movements to dynamically match tone and context.

Local TTS Support: Implementing on-device text-to-speech will make Niamon faster, more private, and capable of working offline.