In this project I show you how to build a complete Local AI Assistant web app that runs 100% on your own machine—no API keys, no cloud, no monthly fees. I connect a local LLM (TinyLlama) running through Ollama, then I build a RAG (Retrieval Augmented Generation) pipeline so the assistant can answer based on the context and knowledge you upload. The main idea is simple: you give your own documents, the app stores them in a local vector database, and when the user asks a question the model reads that context and responds fast. I walk you through the full app experience: a landing page, a chat page with streaming/typing effect, and a knowledge section where you upload text that becomes your assistant’s “brain.” I also explain why a light model is perfect when you’re targeting a specific industry like health or finance—faster responses and lower system requirements. Then I break down the project architecture (Next.js App Router, API routes for chat/embeddings/upload, components, libs, and the local vector DB), and finally I do the full setup: install Ollama, pull the models, install dependencies, run the app, test chat, clear/upload knowledge, and see the assistant pull answers with references from your uploaded context.