[claudio525]

• Ingest script is just AI generated and for a single specific use case, and so review of it should be minimal.

[claudio525]

@claude review

[claude]

Claude finished @claudio525's task in 2m 3s —— View job

---

Review complete

• Gather context (diff, CLAUDE.md, REVIEW.md, existing review comments — none found)
• Review imdb/schema.py changes
• Review scripts/alpine_vs30_ingest.py
• Review tests/test_imdb.py changes
• Post findings

Findings: 1 medium

• imdb/schema.py: is_real BOOLEAN NOT NULL DEFAULT TRUE and the SCHEMA_VERSION bump look correct; NOTES["is_real"] documents it clearly and nothing else in the library hardcodes a sites column list that would need updating.
• tests/test_imdb.py: switching the hardcoded "1" to schema.SCHEMA_VERSION is a good fix — the assertion won't go stale on the next version bump.
• scripts/alpine_vs30_ingest.py: one inline comment posted — the vs30_comparison.csv merge in load_sites() (line ~77) is an inner join that silently drops any station missing from that file, unlike the Z1.0/Z2.5 merge a few lines below which explicitly validates for missing values. A dropped station would later surface as an opaque KeyError from _id_map when add_records/add_site_event run, rather than a clear error naming the missing station. Given the PR description notes this script is single-use/AI-generated, I kept the rest of the review light — the rest of the ingestion logic (distance calcs, IM array reshaping/ordering, GMM logic-tree wiring) checked out consistent on inspection.

No security issues or regressions found elsewhere in the diff.

[joelridden]

We should add EAS as a component as well, unsure if the alpine run produced EAS outputs