Undocx

Undocx is a collaborative, AI-powered document workspace. Teams create workspaces, write and organize rich Markdown documents ("artifacts"), and use an AI assistant to generate content, ask questions, and edit sections in place. Artifacts can be shared with members, organized into folders and collections, and published as read-only public links.

Demo

Features

• Workspaces & collaboration — invite members by email with editor/viewer roles; per-artifact access control on top of workspace roles.
• AI assistant — generate new documents, ask Q&A grounded in your content, and insert or replace sections. Relevant artifacts are pulled in automatically via semantic (vector) search.
• Rich editor — Lexical-based editor with Markdown, tables, code syntax highlighting, callouts, footnotes, MathML, and inline SVG.
• Organization — folders, collections (tags), favorites, and an archive.
• Sharing — token-based email invitations and public read-only views at /share/[artifactId].
• Notifications — activity tracking across workspaces and artifacts.
• Auth — email/password, Google OAuth, two-factor (TOTP), email verification, and password reset.

Tech stack

• Framework: Next.js 16 (App Router), React 19, TypeScript
• Styling/UI: Tailwind CSS 4, shadcn/ui, Lucide icons
• Editor: Lexical
• Database: PostgreSQL (with the pgvector extension) via Drizzle ORM
• Auth: better-auth
• AI: Vercel AI SDK with OpenAI (gpt-4o-mini for generation, text-embedding-3-small for embeddings)
• Email: Nodemailer (SMTP)
• Data fetching: TanStack Query
• Package manager: pnpm

Getting started

Prerequisites

• Node.js
• pnpm
• Docker (for the local Postgres + mail services)

1. Install dependencies

pnpm install

2. Configure environment

Copy the example file and fill in the secrets:

cp example.env .env

Variable	Description
DATABASE_URL	PostgreSQL connection string (matches the docker-compose Postgres service by default).
BETTER_AUTH_URL	Base URL of the app, used for auth callbacks (e.g. http://localhost:3000).
BETTER_AUTH_SECRET	Secret used to sign auth tokens. Required — generate a random value.
TRUSTED_ORIGINS	Optional comma-separated list of extra trusted origins (BETTER_AUTH_URL is always trusted).
GOOGLE_CLIENT_ID / GOOGLE_CLIENT_SECRET	Google OAuth credentials for social sign-in.
OPENAI_API_KEY	OpenAI API key for AI features.
SMTP_HOST / SMTP_PORT / SMTP_USER / SMTP_PASS	SMTP settings (default to the Mailpit service below).

3. Start local services

This brings up PostgreSQL (with pgvector) and Mailpit for catching outgoing email:

docker compose up -d

• Postgres is exposed on localhost:5432.
• Mailpit's web UI is at http://localhost:8025 (SMTP on port 1025).

4. Set up the database

Push the Drizzle schema to your database:

pnpm db:push

5. Run the dev server

pnpm dev

The app runs at http://localhost:3000.

Scripts

Script	Description
pnpm dev	Start the Next.js dev server.
pnpm build	Build for production.
pnpm start	Start the production server.
pnpm lint	Run ESLint.
pnpm format	Format .ts/.tsx files with Prettier.
pnpm typecheck	Type-check with tsc --noEmit.
pnpm ba:generate	Regenerate the better-auth Drizzle schema (lib/db/schema/auth-schema.ts).
pnpm db:generate	Generate Drizzle migration files.
pnpm db:migrate	Apply migrations.
pnpm db:push	Push the schema directly to the database (recommended for local dev).
pnpm db:studio	Open Drizzle Studio to inspect the database.

Project structure

app/
  (marketing)/        Public landing, about, privacy, terms
  (auth)/             Sign-in / sign-up / sign-out
  (app)/              Protected app (workspaces, account, chat)
  api/                API routes
    assistant/        AI: generate, ask, insert, replace
    workspaces/       Artifacts, members, folders, collections, invitations
    auth/[...all]/    better-auth endpoints
  share/[artifactId]/ Public read-only artifact view
  invite/[token]/     Invitation acceptance
components/           UI (shadcn/ui), workspace, marketing, auth, account
lib/
  auth.ts             better-auth server instance
  ai/                 Prompts, embeddings, chunking, usage limits
  db/                 Drizzle client, schema/, queries/
  mail/               Nodemailer transport and email templates
  data/               TanStack Query definitions
drizzle/              Migrations and schema snapshots
docker-compose.yml    Local Postgres (pgvector) + Mailpit

AI assistant

The assistant is backed by OpenAI through the Vercel AI SDK. Documents are chunked by heading, embedded with text-embedding-3-small, and stored as vectors so the assistant can ground its responses in relevant artifacts. Daily per-user usage limits are enforced.

Route	Purpose
POST /api/assistant	Generate a new document ({ title, content }).
POST /api/assistant/ask	Q&A about an artifact (streaming response).
POST /api/assistant/insert	Insert content between sections.
POST /api/assistant/replace	Replace a section.

Adding UI components

This project uses shadcn/ui. To add a component:

npx shadcn@latest add button

Components are placed in the components/ui directory and imported like:

import { Button } from "@/components/ui/button";