{"slug": "ask-me-anything-rag", "title": "Ask me anything (RAG)", "summary": "A developer built AMA, a Retrieval-Augmented Generation (RAG) system for Lagos State University (LASU) that uses Gemma 4 via OpenRouter to answer student questions about lecturers, hostels, and internships based on a searchable database of real student experiences stored in MongoDB Atlas with vector search.", "body_md": "*This is a submission for the Gemma 4 Challenge: Build with Gemma 4*\n\nAt LASU, students constantly ask each other things like \"is this lecturer strict,\" \"is this hostel safe,\" \"what's SIWES actually like at this company,\" but that knowledge lives scattered across WhatsApp groups, Twitter threads, and word of mouth. It gets lost the moment a graduating class leaves. AMA was built to solve that: a shared memory for the school, where real student experiences get collected once and stay searchable by anyone, forever, instead of vanishing into a thousand disconnected chats.\n\nGitHub: [https://github.com/MasterIfeanyi/hackathon-today](https://github.com/MasterIfeanyi/hackathon-today)\n\nWe used Gemma 4 (via OpenRouter) as the answer-generation model, and built the project around Retrieval-Augmented Generation (RAG), not fine-tuning, since we needed the answers to update instantly as new stories get submitted, without retraining anything.\n\nThe pipeline works like this: when a student submits a story or review, we generate an embedding for it using Google's Gemini embedding model and store both the text and the embedding in MongoDB Atlas, using MongoDB Vector Search as our vector database. When someone asks a question, we embed the question the same way, run a vector similarity search to pull the most relevant stories, and pass only those into Gemma 4 as context, so every answer is grounded in real submissions instead of the model guessing.\n\nOn the frontend, we used Next.js (App Router), Tailwind CSS v4 for styling, Zustand for lightweight UI state, and TanStack React Query for handling data fetching and caching.", "url": "https://wpnews.pro/news/ask-me-anything-rag", "canonical_source": "https://dev.to/ifeanyichima/ask-me-anything-rag-32jl", "published_at": "2026-07-25 16:15:19+00:00", "updated_at": "2026-07-25 17:02:28.800184+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-products", "ai-tools", "developer-tools"], "entities": ["LASU", "Gemma 4", "OpenRouter", "MongoDB Atlas", "Google Gemini", "Next.js", "Tailwind CSS", "Zustand"], "alternates": {"html": "https://wpnews.pro/news/ask-me-anything-rag", "markdown": "https://wpnews.pro/news/ask-me-anything-rag.md", "text": "https://wpnews.pro/news/ask-me-anything-rag.txt", "jsonld": "https://wpnews.pro/news/ask-me-anything-rag.jsonld"}}