Self-hosting with Docker

If you’re not on Vercel — or you want the highest-fidelity output — run the service from the Docker image. It’s built on the official texlive/texlive:latest image and compiles with full TeX Live (pdflatex / xelatex / lualatex) rather than Tectonic.

Build and run

1. Build the server bundle and image.

   npm run build:server
   docker build -f docker/Dockerfile -t platex .

2. Run it.

   docker run -p 3001:3001 platex

   Or with Docker Compose:

   docker compose -f docker/docker-compose.yml up

3. Point your app at it.

   PLATEX_SERVICE_URL=http://localhost:3001

Note

build:server (the Docker bundle → dist/server.cjs) is separate from build:vercel (which also downloads the Tectonic binary). The Docker image relies on TeX Live already present in the base image, so it doesn’t bundle Tectonic.

Why full TeX Live

With system TeX Live available, the library runs the real engines natively:

• Same engine flags as a local build: -interaction=nonstopmode -halt-on-error -file-line-error
• Real multi-pass logic: detects \citation{} in the .aux, runs bibtex/biber, re-runs LaTeX until cross-references and labels stabilize
• All three engines: pdflatex, xelatex, and lualatex, selectable per call via the engine option
• pdflatex font metrics — the small differences that XeTeX-based Tectonic can introduce are gone

This is the mode to choose when output must match a specific local TeX toolchain exactly. See How output accuracy works.

Configuration

The container is zero-config — it reads environment variables, so a deployment needs nothing but docker run:

docker run -p 3001:3001 \
  -e PLATEX_API_KEY=your-secret-key \
  -e PLATEX_MAX_CONCURRENT=8 \
  platex

See Server configuration for every setting.

Where to next

Server configurationAuth, concurrency, limits, and body-size caps.

Deploy the serviceThe Vercel and bundled-server paths.