0007 — PETROVA G.1: rocky surface stance
Date: 2026-05-06 Status: closed Supersedes: none — first ratification of G.1 Superseded-by: none — current
Context
Upstream petrova-hq opened sub-project G to ratify the surfaces
that the rocky and eva integrations expose
(petrova-hq/docs/decisions/2026-05-06-sub-project-g-rocky-eva-surface-ratification.md).
G.1 is the rocky-hq portion: does ROCKY continue to be Pebble-fronted
indefinitely, or does rocky-hq stand up its own service surface?
The current integration_rocky.evidence shape in
petrova-hq/contracts/contract.schema.json directly models Pebble's
/api/knowledge/repos and /api/cairn/stones responses
(lore_collection_slug, cairnet_agent_id, cairnet_stones_endpoint,
emission_rate_30d, emission_budget_30d, decay_policy). That's
load-bearing today; the upstream rocky probe (cli/src/probes/rocky.ts)
calls those endpoints unmodified.
rocky-hq's own roadmap (README.md, phase plans
docs/plans/2026-05-02-rocky-scaffold-phase-1.md …) has HEARTH
(Go, per-workspace CAIRNET+LORE provisioner with tiers
solo / team / studio / bespoke) shipping in Phase 5.
Phases 1–4 are scaffold + console lift + RALPH lift + contracts;
Phase 5 is when rocky-hq actually exposes its own provisioner
surface. We are currently mid Phase 3a.
Decision
Pebble continues to front the rocky integration today.
Rationale:
- The current schema shape is correct for the current reality: LORE
and CAIRNET are deployed behind Pebble at
mcp.devarno.cloud; that has not changed and is not changing in Phases 3 or 4. - HEARTH (the natural future surface for "rocky-hq's own service")
is a Phase 5 deliverable. Until HEARTH ships and a consumer repo
actually wires against it, declaring a rocky-hq-owned surface
in the schema would be ungrounded — exactly what
00-bootstrap.mdwarns against. - The 2026-05-06 ownership transfer was an ownership/responsibility change, not a deployment migration; the upstream realignment decision says so explicitly. Schema continuity reinforces that.
When HEARTH ships in Phase 5:
- A successor decision (likely
0XYZ-petrova-g1-rocky-hearth-surface.mdhere, matched by a sub-project G.1' decision in petrova-hq) will re-open G.1. - Schema edits will then add a HEARTH endpoint field
(e.g.
hearth_workspace_url) tointegration_rocky.evidence. - The rocky probe will branch on whether a consumer points at Pebble or at HEARTH.
No schema or probe change today. G.1 closes confirmed.
Alternatives considered
- Pre-declare a rocky-hq surface now, ahead of HEARTH. Rejected: ungrounded; we'd be declaring a contract for a service that does not exist. petrova-hq's invariant against inventing claims applies.
- Wait for HEARTH and leave G.1 open. Rejected: G is open as a blocker for closing the sub-project F → G handoff cleanly. A positive ratification (Pebble continues) is more honest than leaving the question floating.
Consequences
For rocky-hq: Phase 5 plans now own the work to re-open G.1 once HEARTH is concrete. No code changes.
For petrova-hq: G.1 closure recorded; sub-project G.3 in petrova-hq makes no schema change for the rocky integration. eva-hq ratifies G.2 separately.
For invariants: MR-7 upheld (new doc). MR-3 spirit upheld (rocky-hq ratifies its own surface; petrova-hq accepts).
References
- Upstream G decision:
petrova-hq/docs/decisions/2026-05-06-sub-project-g-rocky-eva-surface-ratification.md - Upstream rocky/eva realignment:
petrova-hq/docs/decisions/2026-05-06-rocky-ownership-and-eva-introduction.md - Local roadmap:
README.md, Phase 5 = HEARTH. - Schema today:
petrova-hq/contracts/contract.schema.json#integration_rocky.
Sign-off
- Subagent: petrova-control-plane (2026-05-06)
- Human: pending