perf(artifacts): serve the library listing from UserArtifactsIndex

[philmerrell]

Completes the index work. list_for_user now queries UserArtifactsIndex (GSI2PK=USER#{uid}, GSI2SK descending) instead of the base table.

Safe to merge: the index is ACTIVE in dev and verified populated — see below. Do not merge to main until the same is confirmed there.

What it changes

HEAD and version rows share the base partition, so the old Query spanned ~3× the rows it returned and then date-sorted them in memory. Only HEAD rows carry the GSI2 keys, so the index holds one row per artifact already newest-first. The amplification and the sort both go away, and ordering comes from the store rather than being recomputed per request.

Still returns the whole library in one response, paging the index internally. Exposing pagination is a bigger change than it looks — search and the type filter live in the SPA, and a filter that sees only the loaded page is worse than no filter because it looks authoritative, so both would move server-side in the same change. The index makes that possible whenever it's wanted.

Verified against the live dev index

UserArtifactsIndex   ACTIVE   Backfilling: none
indexed rows: 22                 (= the 22 HEAD rows the backfill stamped)
one user's partition: 9 rows     (= the "9 of 9" the library page shows)
descending by updated_at: True
all rows are #HEAD: True

ItemCount still reads 0 in DescribeTable — that figure refreshes about every 6 hours, so it's stale metadata, not truth. The scan and query above are the real check.

Two bugs this turned up

The library tests were passing against the old code path. My first splice used inverted slice boundaries and left duplicate list_for_user and heads_for_session definitions — Python takes the last, which was the original base-table version. The suite went green while testing nothing new. I caught it by asking why tests that should need a GSI passed against a fixture that had none.

The fixture now declares the index, so moto raises ResourceNotFoundException if the query ever stops using it. That's what makes these tests exercise the index instead of silently falling back.

Undated rows would have vanished. A sparse index omits any HEAD row without GSI2PK, permanently and silently. Rows predating updated_at can't carry a real timestamp, so the first version of the backfill skipped them — and they'd have dropped out of their owner's library, which test_undated_legacy_rows_are_returned_and_sort_last exists to forbid.

They're now stamped with an empty timestamp segment (ARTIFACT##{aid}). That isn't a fabricated time: # sorts below every digit, so read descending the row lands last — exactly where the old in-memory sort put it. Neither dev nor prod holds such a row today; this is the defensive branch preserving a contract the tests already assert.

The pagination stub now yields the newer row on page 1, matching a descending index. With the old stub a client-side re-sort would have passed either way and hidden a broken sort key.

Tests

942 passed across app_api, architecture and the artifact writer suites.

🤖 Generated with Claude Code