Replacing a brittle LaTeX pipeline with Typst — 40× faster document generation
Retiring a slow, hard-to-maintain Jinja/LaTeX document pipeline in favor of an automated Typst pipeline driven from a single structured data file — 40× faster, with full PDF/A support and tracked output history.
Customer
An operation generating high volumes of templated transactional documents — offers, contracts, contingency removals, addenda — from structured data.
Challenge
The legacy pipeline built each document from LaTeX style (.sty) files driven by a custom,
per-document data format and a Jinja templating step. It was slow, infrequently updated,
and the owner struggled to find the LaTeX expertise needed to maintain it.
Solution
We replaced the manual Jinja/LaTeX workflow with an automated Typst pipeline that runs end-to-end, from raw data to finished PDF.
- Typst templates replacing the fragile LaTeX source.
- Python preprocessing that pulls the data and generates figures automatically.
- A CI/CD pipeline that rebuilds documents on every data update.
- Database-tracked outputs — a full history of every document ever produced.
- A single JSON file driving every document for a transaction.
Documents now generate automatically on demand — 40× faster than the LaTeX pipeline, using 4× fewer files, with no Jinja templating step and materially better PDF/A support, all at lower infrastructure cost.