Sienge SaaS RPA

Problem

Each robot (purchase and contract invoices, service invoices, invoice capture and finance) was its own repo, with its own database, scheduler and Docker setup, on Selenium. One change in the Sienge UI broke all of them at once.

Solution

I merged everything into one multi-tenant platform, with Playwright instead of Selenium, Prefect flows, a single PostgreSQL database, a FastAPI backend and a Next.js dashboard to follow runs.

Highlights

• 01ResilientPage, a layer over Playwright with retries, fallback selectors and iframe handling, which uses an LLM to recover broken selectors.
• 02Page objects shared by every robot.
• 03Works with the new React-based Sienge 9 UI.