Every project had the same problem: cloud identifiers like S3 URIs, IAM ARNs, Docker refs, and Azure resource IDs living as bare strings with hand-rolled validation.

pydantypes gives you Pydantic v2 types that validate and decompose these into structured attributes. S3Uri gives you .bucket and .key, DockerImageRef gives you .registry, .tag, .digest, etc.

Covers AWS, Azure, GCP, DevOps, web identifiers, and data engineering.

There’s also LabelEnum for AI/ML classification labels with built-in deprecation, retirement, and alias resolution.

If you’ve ever had to evolve a label taxonomy in production, that one might be worth a look on its own.

Complements pydantic-extra-types. 1k+ tests, fully typed, MIT. Happy to take suggestions for types to add.