[joseph-isaacs]

What changes

bench (database schema and docs only; the site's pages and charts don't change).

The PR benchmark comments in vortex-data/vortex compare against baselines read from this database. SQL benchmark file sizes were never stored here, so when vortex-data/vortex#10987 moved the baseline from S3 to Postgres, the comments lost their file-size section. This adds a table for those sizes.

• db/migrations/014_file_sizes.sql creates file_sizes, keyed by (commit_sha, benchmark, scale_factor, format, file) with value_bytes.
  • Its index is (benchmark, scale_factor, format, commit_sha). CI chooses the baseline commit from the query measurements, then reads that commit's files for each scope, and this index serves that read as a range scan.
  • It needs no grants: the ALTER DEFAULT PRIVILEGES rules in 004, 005 and 011 give bench_ingest, bench_read and bench_ci_read their access when migrator creates the table.
  • It needs no requires-superuser marker either, so bench-schema-deploy.yml can apply it.
• docs/bench/CONTRACT.md, docs/bench/README.md and docs/bench/architecture/data-pipeline.md add the file_size kind. CONTRACT.md also records where the new golden vectors came from, and why a new kind with its own table needs no SCHEMA_VERSION bump.
• infra/RESTORE.md adds the table to the restore data check, and db/migrations/README.md lists 014.

The writer and the reader are in vortex-data/vortex#11443. Apply this migration in production before that PR merges. Develop's ingest is required and writes each file in one transaction, so file_size records sent before the table exists would fail every develop SQL benchmark job. The PR-side reader now skips a missing table, but develop's ingest does not.

How it was checked

• Applied the migrations to a local PostgreSQL 16 and loaded 1,500 commits × 100 files. With the PR 100 commits behind develop, the vortex PR's baseline fetch took 5 ms, and the file read used idx_file_sizes_baseline (an index scan of 100 rows). Under the earlier index and commit search it took 5.3 s.

• An independent review applied 001–013 as a bootstrap role and 014 as migrator through db/migrate-schema.py. It found file_sizes owned by migrator, with SELECT/INSERT/UPDATE for bench_ingest and SELECT for both read roles.

• bun run check:ci and db/tests were not run locally. CI ran them on the first commit and passed.

• Every commit is signed off (git commit -s), for the DCO check.

• bun run check:ci passes.

• Each new image or other asset has an annotation in REUSE.toml. (No new assets.)

🤖 Generated with Claude Code

https://claude.ai/code/session_01UeniB9quRiv5fCx8AhzH9s

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
vortex.dev	Ready	Preview	Nov 13, 2026 1:56pm UTC

[github-actions]

🔦 Lighthouse audit

URL	Perf	A11y	Best practices	SEO
/	🟢 100	🟢 100	🟢 100	🟢 100
/bench	🟢 100	🟢 100	🟢 100	🟢 100
/bench/chart/ct.eyJrIjoiQ29tcHJlc3Npb25UaW1lIiwiZGF0YXNldCI6ImFyYWRlIiwiZGF0YXNldF92YXJpYW50IjpudWxsfQ	🟢 100	🟢 100	🟢 100	🟢 100
/blog	🟢 100	🟢 100	🟢 100	🟢 100
/blog/btrblocks-compressor	🟢 100	🟢 100	🟢 100	🟢 100

Run 43438100699 · 2026-11-13 14:00 UTC