New ISL/OSL

[Oseltamivir]

@frontend-claude

There is a new ISL/OSL combination soon, 8k256, make a new branch for this and PR

[Oseltamivir]

@claude

[github-actions]

Claude finished @Oseltamivir's task in 4m 40s —— View job • claude/issue-309-20260505-0409 • Create PR ➔

---

Task: Add new ISL/OSL combination 8k/256

• Read AGENTS.md and adding-entities docs
• Add 8k/256 (ISL=8192, OSL=256) to sequenceToIslOsl / islOslToSequence in packages/constants/src/models.ts
• Add Sequence.EightK_256 enum + SEQUENCE_CONFIG entry (compact: 8k256, label: 8K / 256, category: default) in packages/app/src/lib/data-mappings.ts
• Update parseIslOsl regex in packages/db/src/etl/normalizers.ts to allow optional k after the OSL (so 8k256 parses as { isl: 8192, osl: 256 })
• Refactor getModelAndSequenceFromArtifact to use a generic islOslToCompact() helper instead of the 1k/8k heuristic (was producing wrong compact keys for any OSL ≠ 1024)
• Update AI chart prompt template to advertise the new sequence
• Added tests in models.test.ts, models-mapping.test.ts, and normalizers.test.ts
• pnpm typecheck ✓
• pnpm lint ✓
• pnpm fmt ✓
• Unit tests for changed packages ✓ (the 3 pre-existing connection.test.ts failures around Neon/TLS are unrelated and present on master)
• Commit and push

Notes

• This is a preemptive add per docs/adding-entities.md — the new sequence will appear in the dropdown but charts will show "No data available" until the first benchmark run with 8k256 artifacts is ingested.
• No parseIslOsl change was strictly mandated by the docs (which note "any {n}k{m}k pattern is supported"), but the new 8k256 artifact convention does not end with k, so the regex needed to accept an optional trailing k on the OSL.
• E2E (Playwright) verification was not run because there is no benchmark data yet for this sequence — the 8K / 256 option is selectable in the sequence dropdown but the chart will (correctly) say "No data" until ingest. Will verify on the Vercel preview after merge once data lands.

· Branch