Story

QueryNest is an AI-powered knowledge and insight hub that lets you upload documents, search across them instantly, and ask questions in natural language — turning scattered files into connected knowledge. It combines Elasticsearch’s speed with the reasoning power of Google’s Gemini model to make information retrieval smarter and faster.

What Inspired Me

Elasticsearch truly got my attention when I realized how fast it can deliver results. When I saw its potential, I realized it could be the perfect foundation for AI-driven tools — especially ones that need to query large sets of unstructured data. That’s what inspired QueryNest.

Giving LLMs the ability to “see” across your documents — PDFs, text files, and more — feels like unlocking a new level of productivity. It’s useful for individuals, but it’s especially valuable for teams and organizations that deal with tons of information daily. This concept has so much potential that I’m already planning to expand QueryNest into a full-fledged SaaS knowledge base platform.

What I Learned

This project taught me how powerful fast search tools like Elasticsearch really are when combined with modern AI. I learned how important it is to bridge search and context — since current LLMs have limited memory, giving them access to a queryable database of your knowledge drastically improves their usefulness. It opened my eyes to the many ways this approach can simplify workflows and even create entirely new markets for AI tools.

How I Built It

QueryNest was built using:

• Next.js, TypeScript, and TailwindCSS for a clean, responsive frontend
• Supabase for storage and database management
• A Node.js + Express backend for file processing, OCR, and text extraction
• Elasticsearch for hybrid search (semantic + keyword)
• Google Gemini (Vertex AI) for intelligent, context-aware responses

All these components come together to create a seamless experience where users can upload files, search naturally, and receive AI-generated insights instantly.

Challenges I Faced

There were definitely a few challenges along the way. Initially, I debated whether to use a self-hosted or cloud-hosted version of Elasticsearch — I eventually went with self-hosting after some trial and error. Getting familiar with Elasticsearch’s APIs, indexing models, and integrating it smoothly with Gemini took some time, but once it clicked, everything came together beautifully.

Other than that, most things went fairly smoothly, aside from a few bumps setting up secure HTTP-only cookie authentication. Each hurdle ended up teaching me something new and making the system more robust overall.

Built With

• elasticsearch
• express.js
• gemini
• nextjs
• shadcnui
• supabase
• tailwindcss
• typescript