Home / Projects / Orchestrata Ultimate / Decisions / ADR index / ADR-0115

ADR-0115: E-1 ruled: the component DAG gains integration → knowledge_graph, governance, tools (27 → 30 edges)

← ADR-0114 · ADR-0116 →

The record

Status: Accepted (ruling, Mohamed, 2026-10-20, verbatim: "Clear E-1 per the standing recommendation") · Date: 2026-10-20

Context

The 2026-10-16 components design session escalated E-1 (AD-33, extended by AD-35): 20 frozen-bytes outbound import rows — kg 7, governance 3, tools 10 (integration boundary census §3, every row line-cited) — target components outside the drawn 27-edge component DAG. Per AD-33, "Until Mohamed rules, every line that needs them carries <PENDING Mohamed E-1>; no design routes around it." 18 of the 25 INT increments carry at least one such line and were E-1-gated, dispatching last (packages/integration/increments/README.md §Ruling gates). D-I6 (packages/integration/design/decisions.md) restated the escalation with fresh frozen-bytes evidence and one measured correction: AD-33 listed D-K8 (knowledge_graph.server) among the E-1-conditional moves, but its whole first-party surface is contracts + kernel — legal without E-1; INT-10 is gated only because it packs with kernel.pipeline_cli. The recommendation stood unchanged: +3 edges (27 → 30).

The prod orchkg DAG seed (2026-10-20, per-instance verified: 10 components, 30 DEPENDS_ON) and clean-target.md's declared-edge table ("AD-33/AD-35 + R5 (30-edge DAG) — declared edges, not violations") already carry the 30-edge shape; the design registers kept the rows pending because only Mohamed can clear them.

Decision (Mohamed, verbatim: "Clear E-1 per the standing recommendation")

1. E-1 is cleared per the standing D-I6 recommendation. integration → knowledge_graph, integration → governance, and integration → tools are DECLARED edges of the component DAG. The DAG is 30 edges.
2. The 18 E-1-gated INT increments are unblocked and dispatch in the queue's order. Their frozen-bytes consumer lines stay byte-faithful as designed — no port-invention, no kernel pass-through (the alternatives AD-33 rejected).
3. Per-line <PENDING Mohamed E-1> markers flip to a citation of this ADR at each increment's dispatch (or in a documentation pass); the design registers (decisions.md NEEDS table, D-I6, increments README) flip in the same registration commit as this ADR.
4. N4 (INT-11) and N-I1 (INT-04) remain gated — they are separate rulings and this ADR does not touch them.

Alternatives rejected

• Kernel pass-through routing (integration's governance/kg/tools reaches detour through kernel) — rejected by AD-33: invents indirection the frozen bytes do not contain and falsifies the oracle.
• Port-invention (new abstractions to avoid the edges) — same rejection: the census rows are real runtime dependencies; a port that exists only to dodge an edge is a fabricated boundary.
• Continue holding E-1 — 18 of 25 INT increments idle indefinitely behind a recommendation that the frozen-bytes census already evidences, while the graph and clean-target tables carry the 30-edge shape.

Consequences

• The component DAG is 30 edges in every home: the seeded prod graph (already true), clean-target.md (already true), and the design registers (flipped by this registration).
• INT increments past wave 1 become dispatchable as the queue reaches them; E-1 stops being a dispatch-order constraint.
• <PENDING Mohamed E-1> markers remaining in per-increment files are historical until each increment's dispatch flips them — a marker found after an increment dispatched is a defect.
• RUL-023 pressure: the edges are declared, so the DAG-BOUNDARY checker (build order 5) reconciles import edges × boundaries.json × graph rows against the 30-edge set, not the 27.

Provenance

Mohamed's in-session ruling, 2026-10-20, on the coordinator's decision prompt (handoff §4 item 2 — "E-1 ruling clear (unblocks 18 INT increments)?"), verbatim answer: "Clear E-1 per the standing recommendation". Escalation of record: AD-33/AD-35 (2026-10-16 design session). Restatement + evidence: packages/integration/design/decisions.md D-I6 and the NEEDS table E-1 row; packages/integration/design/boundary-census.md §3 (row-level frozen-bytes citations); packages/integration/design/clean-target.md E-1 row. Graph state: prod orchkg seed 2026-10-20 (10 components, 30 DEPENDS_ON, per-instance verified; wiki projects/orchestrata-ultimate/component-dag.md). Handoff of record: plans/handoff-2026-10-20-supervisor.md §4.

Related ADRs

No other ADR is named in this record.

Rulings cited

• RUL-023: RUL-023 — A feature request means the thing is WIRED, not that code exists

Specs cited

None (or no spec page exists for the ones named).