NagaQuest

"NagaQuest Reload page "
"NagaQuest homepage — explore Nagaland's 17 tribes and their cultural heritage with AI-powered guides"
"Folklore Weaver — Gemini generates authentic tribe-specific folktales with moral wisdom and text-to-speech narration
"Nagamese Tutor — AI translations with phonetic breakdowns, usage notes, and tribal phrases for learning Nagaland's dialects"
"Heritage Interpreter powered by Gemini AI — decode traditional warrior shawls, customs, and festival symbolism with tribe-specific context"

Inspiration

I came to NIT Nagaland from Rajasthan and realized I barely understood the culture around me — Nagaland has 17 distinct tribes, each with their own languages, festivals, and textile traditions, but all this knowledge is scattered. I wanted to build something that brought it together in one place, using AI to make it accessible to anyone visiting or curious about the region.

What it does

NagaQuest is an AI-powered heritage guide with three core features: Folklore Weaver generates authentic tribe-specific folktales with cultural context, Heritage Interpreter explains traditional warrior textiles and Morung customs with historical depth, and Nagamese Tutor translates phrases into Nagamese and tribal dialects with phonetic guides. All features run on Google Gemini API with intelligent fallback — if the API fails, users still see real cultural content instead of errors.

How we built it

Frontend: React 19 + Vite + Tailwind CSS. Backend: Gemini API with a multi-model fallback system (trying models in order of highest quota first to avoid rate limits). Smart features: response caching with sessionStorage to reduce API calls, temperature tuning per feature (0.88 for creative folklore, 0.25 for accurate translation), and a hand-built offline database of tribe anchors (festivals, shawls, instruments) that gets injected into every prompt.

Challenges we ran into

Rate limiting was brutal — Gemini's free tier caps at 20 requests/day on some models. We solved this by profiling quota across all models and reordering our fallback array to prioritize high-quota models. Nagamese translation accuracy — since it's a low-resource creole, we hand-coded fallback translations for common phrases instead of relying on AI alone, so core functionality always works.

Accomplishments we're proud of

We built real prompt engineering, not just a chatbot wrapper. Each feature has a tailored system prompt with tribe-specific anchors (names, festivals, shawls) injected at runtime. The offline fallback system means the app never crashes — it degrades gracefully. And we shipped it working end-to-end with zero external dependencies beyond the Gemini SDK.

What we learned

Prompt engineering quality directly affects output quality — vague prompts get vague results. Caching and rate-limit optimization aren't afterthoughts, they're core to the architecture. And you can't fake cultural authenticity in AI — you have to research and encode real domain knowledge into your system prompts.

What's next for NagaQuest

Add text-to-speech for folktales and language phrases. Integrate with the Hornbill Festival's official schedule so users can get real-time festival guides tied to live events. Scale to include video tutorials of traditional crafts (weaving, carving) paired with AI narration. And eventually, collect feedback from actual Naga community members to continuously refine cultural accuracy.