Back to Projects
KnowledgeBase AI
A full-stack Retrieval-Augmented Generation (RAG) application that enables users to upload documents or webpages and chat with an AI assistant that answers questions using only the ingested content, complete with source citations.
Overview
Built KnowledgeBase AI, a production-ready RAG application using Next.js, LangChain.js, OpenAI, and ChromaDB that allows users to ingest PDFs, DOCX files, text files, and webpages into a vector database. Implemented an end-to-end retrieval pipeline with semantic search, document chunking, embeddings, streaming AI responses, and inline source citations, delivering accurate, context-aware answers grounded entirely in uploaded knowledge.
Key Achievements
Built a production-ready Retrieval-Augmented Generation (RAG) system capable of answering questions using only uploaded documents and webpages
Implemented an end-to-end document ingestion pipeline supporting PDF, DOCX, TXT, and URL sources with automatic parsing, chunking, embedding generation, and vector storage
Developed a semantic search pipeline using OpenAI embeddings and ChromaDB to retrieve the most relevant document chunks for every query
Integrated LangChain.js and OpenAI GPT models to generate context-aware responses while preventing hallucinations through retrieval-based prompting
Implemented real-time streaming AI responses with the Vercel AI SDK to improve chat responsiveness and user experience
Designed an interactive citation system that displays document sources, chunk references, and expandable context snippets for answer transparency
Built a modern full-stack interface with drag-and-drop document uploads, document management, persistent application state, and responsive chat UI
Containerized the vector database using Docker and designed a modular architecture with reusable ingestion, retrieval, and parsing components for maintainability
Tech Stack
Next.js 14ReactTypeScriptTailwind CSSshadcn/uiNode.jsLangChain.jsOpenAI APIChromaDBDockerZustandVercel AI SDKAxiosCheeriopdf-parseMammoth