Setting up uv with Google Cloud Run Functions

An exciting aspect of kicking off a fresh new project is that you get the opportunity to try not just one, but as many new technologies as you wish.

Recently, we embarked on a new Python web project for a client, built on Google Cloud Run Functions (with HTTP triggers) — a new technology for our arsenal. This served as an opportunity to not stop there, but also give a go at uv, a novel package manager for Python, built by Astral and written in Rust, which brings tremendous performance benefits.

Turned out though that as of this time (January 2025), in order to package and deploy a project to Google Cloud Run Functions the old-school requirements.txt file with the dependency list is required. This looked like an issue initially, since uv keeps dependencies as declared by the user in pyproject.toml, while consulting the automatically generated uv.lock file, for the complete dependency tree resolved as flat dependencies.

Thankfully, the team behind uv had already done the work for such use cases. After just a bit of searching, we found out that uv comes with the built-in uv export --format requirements-txt command, allowing to export all dependencies of the project in a flat requirements.txt file. Eventually, all we had to do is:

1. Add requirements.txt in .gitignore, to keep the resolved dependency tree checked out only in uv.lock.
2. Run uv export --format requirements-txt --locked > requirements.txt in our CI/CD, before the deployment command to generate a requirements.txt file based on the dependency tree in uv.lock

Finally, both uv and Google Cloud Run Functions are beasts on their own, so we will dedicate at least one blog article to each one of these technologies soon.