QueryDocs

Ask questions. Find answers in your documents.

QueryDocs is a full-stack document question-answering application built around retrieval-augmented generation (RAG). Users can upload documents, select the documents they want to search, ask questions in a conversation, and receive AI-generated answers with relevant sources.

Overview

QueryDocs is split into two applications:

• Frontend: Next.js interface for authentication, document management, chat, conversation history, profile, and settings.
• Backend: TypeScript and Express API for authentication, document processing, embeddings, vector search, answer generation, persistence, and Supabase Storage.

The backend uses PostgreSQL through Prisma for application data. Supabase provides private document storage and pgvector-based similarity search. Gemini provides document embeddings and the default answer-generation provider; Groq is also supported for answer generation.

Architecture

flowchart LR
	User[User] --> Web[Next.js frontend]
	Web --> API[Express API]
	API --> Auth[JWT authentication]
	API --> DB[(PostgreSQL + Prisma)]
	API --> Storage[Supabase Storage]
	API --> Vector[Supabase pgvector]
	API --> AI[Gemini or Groq]

Loading

RAG workflow

flowchart TD
	Upload[Upload TXT, PDF, or DOCX] --> Extract[Extract and clean text]
	Extract --> Chunk[Split text into chunks]
	Chunk --> Embed[Generate 1536-dimensional Gemini embeddings]
	Embed --> Store[(Store chunks and vectors)]
	Question[Ask a question] --> QueryEmbed[Embed the question]
	QueryEmbed --> Search[Similarity search]
	Store --> Search
	Search --> Context[Build relevant context]
	Context --> Answer[Generate an answer with sources]

Loading

Architecture diagrams

Diagram	Description
System architecture	Application components and external services
RAG pipeline	Document ingestion and question-answering flow
Database architecture	PostgreSQL entities and relationships

Features

• Account registration, login, and protected dashboard routes
• Upload and process TXT, PDF, and DOCX documents
• Semantic document search with vector embeddings
• AI-generated answers grounded in selected document content
• Source references returned with answers
• Conversation creation, history, renaming, and deletion
• Dashboard statistics for documents, questions, and storage
• Temporary local uploads with original files stored in private Supabase Storage
• Structured backend logging and validated environment configuration

Repository structure

.
├── architecture/       System, RAG, and database diagrams
├── backend/             Express API, Prisma schema, processing pipeline, and tests
├── frontend/            Next.js application and UI components
└── README.md            Project overview and navigation

Getting started

Prerequisites

• Node.js 20 or later
• Docker Desktop, or an existing PostgreSQL database
• A Supabase project with Storage and database access
• A Gemini API key

1. Start the backend

Follow the complete setup guide in backend/README.md. In brief:

cd backend
npm install
docker compose up -d postgres
npm run db:generate
npm run db:push
npm run dev

The backend runs at http://localhost:8080 by default. The backend README explains the required .env values and the one-time Supabase vector SQL setup.

2. Start the frontend

In a second terminal, follow frontend/README.md:

cd frontend
npm install
npm run dev

The frontend runs at http://localhost:3000 and uses http://localhost:8080 as the default API URL. To use another backend URL, create frontend/.env with:

NEXT_PUBLIC_API_URL=http://localhost:8080

The backend CLIENT_ORIGIN must match the frontend URL for browser requests to succeed.

Development commands

Run commands from the relevant application directory.

Area	Install	Development	Validation	Production
Backend	npm install	npm run dev	npm run typecheck and npm test	npm run build && npm start
Frontend	npm install	npm run dev	npm run lint	npm run build && npm start

API and application documentation

• Backend setup, environment, API routes, and database
• Frontend setup, routes, API integration, and project structure
• System architecture diagram
• RAG pipeline diagram
• Database architecture diagram

Security notes

• Keep backend .env files and API credentials out of version control.
• Never expose SUPABASE_SERVICE_ROLE_KEY, database credentials, or model API keys to the frontend.
• Use a strong, unique JWT_SECRET in production.
• Restrict the backend CORS origin to the deployed frontend URL.
• Keep the Supabase document bucket private.

Project context

QueryDocs is a learning project created to practice full-stack application development, RAG workflows, document processing, authentication, vector search, and AI integration. AI assistant agents were used during development for selected tasks, repetitive tasks, learning, exploration, implementation support, and documentation.

Educational Purpose

QueryDocs is primarily an educational and learning project. It may contain bugs, incomplete implementations, or dependencies on third-party services.

Users are solely responsible for reviewing, modifying, and deploying this project.

AI-Assisted Development

This project was developed with the assistance of AI coding tools and agents for code generation, debugging, refactoring, documentation, and development support.