erdify 0.8.0 — faster scans, deterministic output (heads-up: ERDs reorder once)

[FabianClemenz]

erdify 0.8.0 is out 🎉

This release makes scans faster and output deterministic. Install with pip install -U erdify (or bump your pre-commit rev to v0.8.0).

⚠️ Heads-up: your committed ERDs will reorder once

Model files are now discovered in sorted order, so the elements in the generated .puml / .mmd appear in a new, deterministic order.

• The diagram content is unchanged — identical entities, columns and relationships, only their ordering differs.
• Expect a one-time diff the next time erdify regenerates (pre-commit hook or CI --check). Commit it once; output is then stable across machines and CI runs.
• This is why it ships as a minor bump rather than a patch.

If you pin erdify in pre-commit, the first run after upgrading will produce that reorder diff — that's expected, just commit it.

What changed

• Faster scans — the models.py scan now prunes excluded directories (venv / site-packages / caches) during the directory walk instead of after a full rglob. Large non-project trees like .venv are never traversed: ~3.5× faster end-to-end on a real repo (~6800 → ~120 directories walked).
• Deterministic output — discovered model files are sorted, giving a machine-independent ERD element order (no more spurious diffs between developers or CI).

Maintenance

• CI now runs on macOS and Windows (min+max Python) alongside the full Linux Python sweep.

📦 Release: https://github.com/devsuit-berlin/erdify/releases/tag/v0.8.0
📝 Changelog: https://github.com/devsuit-berlin/erdify/blob/main/CHANGELOG.md