Inspiration
I wanted to build something that honors my roots while exploring the future of software development. The Thirukkural is a masterpiece of classical Tamil literature containing 1,330 couplets of timeless ethical guidance. However, its ancient grammar and deep philosophical structure can be highly intimidating for modern readers. At the same time, when people try to use AI to learn or code, they often end up with messy, unorganized workflows. I wanted to solve both problems: using modern AI agents to build a clean, structural tool that breaks down ancient wisdom into relatable, real-world short stories.
What it does
Thirukkural Explainer is a bilingual interactive web application that acts as a modern bridge to ancient philosophy. A user can search for any Kural by its exact number, its starting words, or by chapter. The app instantly pulls up the original sacred Tamil verse using beautiful, distinct typography, alongside the authoritative 1886 English poetic translation by G.U. Pope. If a user wants to truly understand its moral lesson, they can trigger the AI Story engine. The app uses Google Gemini to generate a custom 2-to-3 paragraph real-world short story illustrating that exact couplet's wisdom. If a search comes up empty, the app handles it gracefully by serving randomized famous Kurals so the user never hits a dead end.
How we built it
Following the rules of the hackathon, I started from an absolute scratch, empty directory. I installed and initialized the Devpost Learn Skill Pack right inside my coding agent environment. The AI skills interviewed me about my feature goals and automatically generated the foundational planning files (scope.md, prd.md, and spec.md) before a single line of application code was written. • The Backend is built with FastAPI (Python 3.10+) for blazing-fast API endpoints. • The AI Integration connects to the new Google GenAI SDK using the gemini-2.5-flash model. • The Frontend uses clean HTML5, CSS3, and lightweight Vanilla JavaScript styled with a unique classical oil-paint glassmorphism theme. • To optimize speed and minimize API costs, I built an in-memory session cache that saves generated stories instantly for sub-millisecond repeat views.
Challenges we ran into
Handling complex Indic text layouts was a massive early hurdle. Tamil is a contextual, non-linear script. Without specific shaping engines and perfect Unicode handling, vowel markers and dots (pullis) easily break apart, overlap awkwardly, or render as empty boxes. I had to carefully map out the CSS typography using stable fonts like Noto Sans Tamil to keep the vertical reading stack flawless. Additionally, because Gemini generated stories so incredibly fast, my custom-built Trivia Loading Screen (which displays fun historical facts about Thiruvalluvar) flashed on the screen too quickly to be read! I had to implement asynchronous timing controls in the backend to ensure a smooth, premium visual experience during the presentation walkthrough
Accomplishments that we're proud of
I am incredibly proud of creating a product that balances aesthetics with technical reliability. The app doesn’t just output dry text; it creates a complete, coherent visual experience through its oil-paint glassmorphism theme. I am also proud of implementing the in-memory cache system and the fallback offline mode, which ensures that even if a user doesn't have a Gemini API key configured, the application naturally switches to pre-built contextual stories without crashing
What we learned
This hackathon completely changed my development habits. I learned the immense value of planning before building. Allowing the AI agent skills to interview me forced me to cut down unrealistic feature creep early on, keeping my project scope perfectly bounded for a 2-day timeline. I also learned how to handle asynchronous operations cleanly in FastAPI and gained a much deeper understanding of OpenType font behaviors across browsers.
What's next for Thirukkural Explainer
I want to take this proof-of-concept to the next level by turning it into a full Minimum Viable Product (MVP). The immediate next steps include: • Adding an audio synthesis layer so users can listen to the proper pronunciation of the Tamil verses aloud. • Implementing a user bookmarking feature using a lightweight database (like SQLite) so individuals can save their favorite Kurals and generated stories. • Opening up a community contribution tab where educators can submit their own real-world case studies to fine-tune the story generation system.
Log in or sign up for Devpost to join the conversation.