Documentation · Releases · Triptolemus

Triptolemus — H.1.4

v.Chronos.1.Triptolemus — legacy v.H.1.4

Why “Triptolemus”

Triptolemus was the mortal protégé of Demeter — the youth she took into her confidence at Eleusis, taught the art of agriculture, and sent out across the world in a winged chariot to spread cultivation to all peoples. The codename captures the shape of this release relative to the one before it: the v.H.1.3 node (whose historical “Demeter” Name was re-drawn to the minor-pool draw Daedalus) made the disk a first-class cultivated entity — benched, scored, and stamped in its own right. Triptolemus propagates that cultivation everywhere. It makes the greenlight-gated reuse decision real at bench time — gold is reused, red is rebenched — instead of merely scored after the fact, and it adds bulk orchestration that carries freshened gold results, wave by wave, across the whole fleet. The Daedalus node cultivated the ground; Triptolemus spreads the harvest.

Headline changes

• Greenlight-gated reuse is now a real bench-time decision. When a trigger fires, a gold upstream bench that fits the consumer is reused — skipped, not re-run. A red, missing, stale-version, or non-fitting upstream is re-benched fresh, so nothing red is ever silently reused. This is the operator’s standing “red ⇒ must rebench” rule enforced at bench time, not just reflected in the score that is read back. Reuse is a skip of execution only — no value is copied; every consumer still reads the shared disk entity and ghost-CPU layer at read time, so the single-source guarantee is unchanged.
• Universal force-fresh toggle, default OFF. Every bench trigger — single-entity and bulk — now defaults reuse-aware: a gold upstream is skipped automatically. A single explicit “force fresh even if gold” toggle (default OFF) on every trigger overrides the gold-skip and re-benches regardless, so a clean re-measure is always one click away. The old 30-day recency window is now only a secondary safety; the primary gate is version-gold, not age.
• Three per-kind “Bench all” buttons plus a master. Bench all Disks, Bench all Virtual Container Entities, Bench all Real Container Entities, and a master Bench all Entities. Every bulk run is wave-ordered Disk → Virtual → Real so each downstream wave reuses the gold results just freshened upstream in the same run. Each wave skips gold entities and benches only the red or missing ones. A per-kind button auto-cascades only the reuse-skip-heavy upstream waves it needs to refresh its own dependencies; the master button is the full fleet reconcile.
• Bulk runs are admin-gated, audited, and previewed. The “Bench all” controls are Ancient-admin only and every run is audit-logged so a per-kind run and the master reconcile are distinguishable in the trail. The confirm dialog cites the per-wave plan — how many entities are enqueued versus skipped because they are already gold — plus an estimated wall-time and a downtime statement. After the trigger, a per-wave progress watch shows done out of total and how many were skipped as gold.
• Disk bench-version and Status now in the score card. The per-node score card surfaces the disk entity’s bench-version and its gold / amber / red Status, tied to the physical drive — exactly the way the container bench-version and Status already show, using the same version-mismatch classifier and colour palette. A disk with a missing or non-live stamp reads red, a gold disk reads gold, and an in-between stamp reads amber, so the reuse gate is visible right where the operator reads the scores.

For operators

The bulk “Bench all” controls are admin-only and live on the Argus tabs and header — Bench all Disks, Bench all Virtual Container Entities, Bench all Real Container Entities, and the master Bench all Entities. They reconcile a whole category, or the whole fleet, to the live greenlit version with one click, skipping anything already gold so a bulk run only re-benches what is actually red or missing.

The per-entity triggers on the Triton panel are unchanged except that the universal force-fresh toggle now defaults OFF — every trigger is reuse-aware by default and only re-runs gold upstream when you explicitly ask it to. And a red disk in the score card now means a rebench is genuinely needed: the reuse gate is visible where the scores are read, so a red disk is your signal to refresh it rather than something to interpret after the fact.

Related

• /docs/scoring — the canonical scoring documentation, now describing the now-real greenlight-gated bench-time reuse: a gold upstream that fits is reused, a red / missing / stale upstream is re-benched fresh, and the universal force-fresh toggle that overrides the gold-skip on every trigger.
• Daedalus — v.Chronos.1.Daedalus (legacy v.H.1.3) — the immediately preceding arc; Triptolemus opens the H.1.4 minor on top of the v.H.1.3 node (the historical “Demeter” Name was re-drawn off the reserved major-tier deity slot to the minor-pool draw Daedalus), carrying its first-class cultivated disk entity forward and propagating the greenlight-gated entity model into real bench-time reuse and fleet-wide bulk orchestration.