j7/StarrOS
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

decisions: ADR-0004 (multiplex = per-wolf) + Honcho decommission (Gate-0 D1/D2) #7

Merged
j7 merged 2 commits from decisions/gate-0-shakedown-resolutions into main 2026-07-09 20:05:35 -04:00
Conversation 0 Commits 2 Files changed 5 +69 -7
j7 commented 2026-07-09 19:55:17 -04:00
Owner
Copy link

Branch: decisions/gate-0-shakedown-resolutions → main
Depends on nothing — this PR stands alone against current main, does not require PR #6 to merge first.
Reviewer: Jigo (touches ADR + spec + kanban per specs/starros-stack-spec.md §4.1 authority table)

What this is

J7 answered the two Gate-0 decisions from the shakedown audit (documents/audits/claude-shakedown-2026-07-09.md, PR #6) directly, rather than via San's decision-brief flow. This PR is those two decisions landed as canon.

D1 — multiplex target: per-wolf (ADR-0004)

Per-wolf gateways are canonical, matching what's already deployed (9 wolves live on :8644-8652). Multiplexing was attempted and did not reach a working state — not adopted. This resolves the spec §2 item 3 self-contradiction (dossier issue #1) that was blocking Phase P0's definition of done.

  • New decisions/0004-multiplex-target-per-wolf.md
  • specs/starros-stack-spec.md §2 item 3 rewritten; two dangling anchors fixed to real runtime-state.md §2/§3; P0 phase-table title corrected (cross-reference said Phase P1, table said P0 — now consistent)
  • kanban/backlog.md E3-T1 retitled: scope is now just "fix Ashitaka's :8643 binding," multiplex-adoption language dropped

D2 — Honcho: decommission, no Mnemosyne migration

Honcho and Mnemosyne are both memory-layer components (ADR-0003 glossary); Mnemosyne is the surviving canonical layer, so there's nothing to combine. Honcho's podman unit is still live on :8000 despite three docs using three different status words for it (dossier issue #2) — normalized here to one status, with the actual stop tracked as a new ops card (E3-T15) since editing a doc doesn't stop a running unit.

  • decisions/0003-canonical-spec-boundaries.md glossary status normalized
  • docs/runtime-state.md §2 port row corrected + decommission note
  • specs/starros-stack-spec.md historical-only line corrected (previously claimed "removed" while the container is still running)
  • kanban/backlog.md new E3-T15, owner Toki + Ishii — this is the one actionable item in this PR that isn't a doc fix; stopping the container is outside repo/CI scope

Merge gates

  • Jigo review
  • Toki/Ishii pick up E3-T15 (Honcho container stop) once merged
  • San's execution prompt (prompts/san-execute-shakedown.md, PR #6) already updated to skip Gate 0 and reference this PR directly
**Branch:** `decisions/gate-0-shakedown-resolutions` → `main` **Depends on nothing** — this PR stands alone against current `main`, does not require PR #6 to merge first. **Reviewer:** Jigo (touches ADR + spec + kanban per `specs/starros-stack-spec.md` §4.1 authority table) ## What this is J7 answered the two Gate-0 decisions from the shakedown audit (`documents/audits/claude-shakedown-2026-07-09.md`, PR #6) directly, rather than via San's decision-brief flow. This PR is those two decisions landed as canon. ## D1 — multiplex target: per-wolf (ADR-0004) Per-wolf gateways are canonical, matching what's already deployed (9 wolves live on :8644-8652). Multiplexing was attempted and did not reach a working state — not adopted. This resolves the spec §2 item 3 self-contradiction (dossier issue #1) that was blocking Phase P0's definition of done. - New `decisions/0004-multiplex-target-per-wolf.md` - `specs/starros-stack-spec.md` §2 item 3 rewritten; two dangling anchors fixed to real `runtime-state.md` §2/§3; P0 phase-table title corrected (cross-reference said Phase P1, table said P0 — now consistent) - `kanban/backlog.md` E3-T1 retitled: scope is now just "fix Ashitaka's :8643 binding," multiplex-adoption language dropped ## D2 — Honcho: decommission, no Mnemosyne migration Honcho and Mnemosyne are both memory-layer components (ADR-0003 glossary); Mnemosyne is the surviving canonical layer, so there's nothing to combine. Honcho's podman unit is still live on `:8000` despite three docs using three different status words for it (dossier issue #2) — normalized here to one status, with the actual stop tracked as a new ops card (E3-T15) since editing a doc doesn't stop a running unit. - `decisions/0003-canonical-spec-boundaries.md` glossary status normalized - `docs/runtime-state.md` §2 port row corrected + decommission note - `specs/starros-stack-spec.md` historical-only line corrected (previously claimed "removed" while the container is still running) - `kanban/backlog.md` new **E3-T15**, owner Toki + Ishii — this is the one actionable item in this PR that isn't a doc fix; stopping the container is outside repo/CI scope ## Merge gates - [ ] Jigo review - [ ] Toki/Ishii pick up E3-T15 (Honcho container stop) once merged - [ ] San's execution prompt (`prompts/san-execute-shakedown.md`, PR #6) already updated to skip Gate 0 and reference this PR directly 
J7 answered the two shakedown Gate-0 decisions directly:

D1 — multiplex target: per-wolf gateways, matching what's already deployed
(9 wolves live on :8644-8652). Multiplexing was attempted and did not reach
a working state; not adopted. Resolves the spec §2 item 3 self-contradiction
(dossier issue #1). P0 scope narrows to fixing Ashitaka's :8643 binding only.

  - decisions/0004-multiplex-target-per-wolf.md (new ADR)
  - specs/starros-stack-spec.md: §2 item 3 rewritten, dangling anchors fixed
    to real runtime-state §2/§3, P0 phase-table title corrected (was
    mislabeled Phase P1 in the cross-reference, table itself says P0)
  - kanban/backlog.md: E3-T1 retitled to match narrowed scope

D2 — Honcho: decommission, no Mnemosyne migration needed. Honcho and
Mnemosyne are both memory-layer components per the ADR-0003 glossary;
Mnemosyne is the surviving canonical layer. Honcho container is still
live on :8000 despite being called de-scoped/removed in three different
docs with three different words (dossier issue #2) — normalized to a
single status here, and the actual stop is tracked as a new ops card
since a doc edit doesn't stop a running podman unit.

  - decisions/0003-canonical-spec-boundaries.md: glossary status normalized
  - docs/runtime-state.md: §2 port row corrected, decommission action noted
  - specs/starros-stack-spec.md: historical-only line corrected (was
    claiming 'removed' when container is still running)
  - kanban/backlog.md: new E3-T15, owner Toki + Ishii (operational, not
    doc-only — actually stopping the unit is outside repo scope)

Ref: documents/audits/claude-shakedown-2026-07-09.md (PR #6, pending)
j7 merged commit f4cef71d12 into main 2026-07-09 20:05:35 -04:00
j7 deleted branch decisions/gate-0-shakedown-resolutions 2026-07-09 20:05:35 -04:00
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
j7/StarrOS!7
No description provided.