Docker Deployment

The app ships with a multi-stage Dockerfile that builds the frontend and serves it with nginx.

Build & run

docker build -t mm-annotator .
docker run -p 80:80 \
  -e BACKEND_URL=wss://your-host/uima \
  -e UCE_URL=https://your-uce-host \
  mm-annotator

The app is then available at http://localhost.

How it works

1. Build stage — node:22-alpine runs npm install and npm run build.
2. Serve stage — the dist/ output is copied into nginx:alpine with a custom nginx.conf.

SPA routing

nginx.conf falls back to index.html so client-side routes resolve correctly.

Runtime configuration

docker-entrypoint.sh runs at container start and writes the environment variables into a config.js file (defining window._env_) served next to the app. This lets one image target different backends without rebuilding. See Configuration for the available variables.

Docker Compose

Build from source:

services:
  frontend:
    build: .
    ports:
      - "80:80"
    environment:
      - BACKEND_URL=wss://your-host/uima
      - UCE_URL=https://your-uce-host

Reproducible capsule

For a reproducible deployment, use the prebuilt image published to the TTLab registry instead of building locally. The frontend is one service of the wider TextAnnotator + RAG stack:

ta-frontend:
  image: docker.texttechnologylab.org/textannotator-rag-demo:latest
  ports:
    - ${PORT_FRONTEND}:80
  environment:
    BACKEND_URL: ${WS_URL}
    UCE_URL: ${UCE_URL}

PORT_FRONTEND, WS_URL and UCE_URL are supplied via the stack's .env file; WS_URL points at the TextAnnotator WebSocket (…/uima) and UCE_URL at the UCE host serving the RAG Bot.

The UCE host is provided by the Unified Corpus Explorer (UCE), which serves the RAG Bot backend; see its repository for deployment instructions.