ADR 0032: Release renumbering: 0006 is the workbench migration; stage gates move to R5 as 0008

Status

Accepted

Date

2026-09-16

Deciders

Founder

Context

Contract 11.4 reserved 0006_stage_gates (tables -> 62, routes 189 -> 196, tone 62, run kind stage_readiness, keys +4 -> 91, PERSONA_VERSION 4) for R4. Phase 4 puts workbench roles, the evidence panel, skills and memory tiers first, and stage gates depend on the lens and memory work landing.

Decision

• R4 migration is 0006_workbench (evidence_pins, document_assets, memory_records plus columns; 60 -> 63). R5: 0007_lenses (63 -> 65) then 0008_stage_gates (65 -> 67). R6: 0009_user_grants (67 -> 68).
• The 11.4 R4 lines are superseded verbatim by contract 12.7: routes 189 -> 206 in R4 (+17) and 223 in R5 (lens reviews +3, role memory +6, propose-share +1, stage gates +7); tone map 61 in R4, 62 in R5; run kinds 13 in R4, 15 in R5 (lens_review, stage_readiness); permissions 87 and PERSONA_VERSION 3 in R4, 93 and version 4 in R5.
• Files this decision touches: api/tests/test_migrations.py comment chain (60 + 3 + 2 + 2 + 1), api/scripts/designed-routes.json (the R4 rows of contract 12.7 now; the stage-gate rows of Part 7 section 7.4 R4 copied at R5, count verified on the day), contract 11.4 (superseded lines, never edited), contract 12.7.

Consequences

• Nothing about stage gates changes in substance (ADR 0024 stands); only the release and the revision number move.
• The approval matrix rows “Stage gate ready / reviewed / closed (R4, designed)” read R5 in the docs mirror when regenerated; the generator, not this record, owns the text.

Alternatives considered

• Both migrations in R4: two migrations and three writers on the same tables in one pass; rejected.
• Editing 11.4 in place: contract history is never rewritten; superseded, not edited.

Follow-ups

• Re-derive every count on the day each migration lands.