Reference Implementation Architecture

Created openspec/changes/accelerate-reference-spine-overview-lists/specs/reference-implementation-architecture/spec.mdView on GitHub →

MODIFIED Requirements

Requirement: Reference spine overview lists are bounded on Postgres

The reference implementation SHALL serve unfiltered first-page reference spine correlation lists (/_ref/traces, /_ref/runs, and /_ref/grants) from a bounded recent-correlation read path on Postgres. Filtered and cursor-paginated reads SHALL preserve the existing aggregate semantics.

Scenario: recent first page avoids whole-table grouping

WHEN the owner reads the first page of traces, runs, or grants with no filters and no cursor THEN the Postgres implementation uses a recent-event scan to select the page of correlation IDs AND it hydrates only the selected page.

Scenario: filtered reads keep aggregate semantics

WHEN the owner reads a spine correlation list with a status, time, source, grant, client, query, or cursor filter THEN the Postgres implementation uses the aggregate path that applies those filters to the correlation semantics AND it does not silently substitute the unfiltered recent path.

Requirement: Dashboard home does not block on unused operational lists

The console SHALL NOT fetch failed-run or failed-trace lists on the dashboard home when those lists do not drive any rendered hero, card, or action.

Scenario: dashboard attention truth comes from connector verdicts

WHEN the dashboard home renders the "needs you" hero state THEN it derives that state from connector rendered verdict attention rows AND it does not fetch failed-run or failed-trace lists for that decision.