n8n or plain Python on Cloud Run or sth else? Solo dev dilemma
I’m about to be the only person at a small company handling automations and LLM stuff. We’re on GCP and Google Workspace. A typical project would be something like: invoices land in a shared inbox, an LLM pulls out the key fields, the data gets pushed into our accounting tool, and the approver gets a Slack ping. Only a light touch of AI (more advanced in the future), mostly glue between systems. Some of our tools are on-prem, and more projects will follow.
I’m torn between:
- Self-hosted n8n on a VM
- Python (FastAPI) on Cloud Run, with Cloud Tasks for queues and retries, Postgres for run history, and Grafana for logs and token usage
What matters to me:
- easy to extend later,
- solid logs, retries,
- being able to check old runs and re-run failed ones,
- and low cost (open source preferred).
My worry is that n8n gives me the UI for free but gets awkward with versioning and testing as things grow, while code gives me control but I’d have to build the dashboard and retry tooling myself.
Anyone been in a similar spot? What did you pick, and what would you change? I’m especially curious how you handled run history and retries without n8n.
And if you went a completely different way (Prefect, Temporal, Windmill, something else), I’d love to hear what it looked like and whether you’d recommend it. Happy to be talked out of both of my options if there’s a better one.
Thanks!