🗃️ Considerations for future non-breaking database migrations

[falexwolf]

BranchUser BranchULabel BranchProject Branch._status_code

extra_data field on Artifact & Run

description on Run

RunArtifact link model: https://laminlabs.slack.com/archives/C04A88BKZPX/p1770751652970929?thread_ts=1770742723.370099&cid=C04A88BKZPX

a column related_name on Feature to express the name of the inverse relationship, e.g. for "plan" it could be "planned"

---

• ✨ Consider adding ProjectUser #3468

Use SET_NULL for ON_DELETE everywhere (context e.g.: #1547)

Feature names conditionally unique?

---

consider renaming run.output_artifacts to run.created_artifacts: https://github.com/laminlabs/pfizer-lamin-usage/issues/742

---

transform.key should not be nullable

---

transform.suffix for consistency with artifact.suffix - probably not a good idea

---

and consider allowing kind unions for transforms that have multiple entrypoints with different kinds (e.g. function and script).

---

• Add .synonyms field to ULabel
• Add .ontology_id field to ULabel
• Add .source field to ULabel, pointing to bionty.Source. So that ULabel can behave like a biorecord. -- then we'll need the Source table inside lamindb

Last set of migrations:

• 🗃️ Migrations for LaminDB v2 #3217