docs(platform): e2e observation seam for the mention steer/deliver split (RIG-1788) - #581
Open
rigel-mintaka wants to merge 1 commit into
Open
docs(platform): e2e observation seam for the mention steer/deliver split (RIG-1788)#581rigel-mintaka wants to merge 1 commit into
rigel-mintaka wants to merge 1 commit into
Conversation
…lit (RIG-1788) Design record for RIG-1788 / dogfood-e2e H4: proving the mention **steer/deliver split** over the real stack. H4's other assertions (spawn F2 ownership, second container, leg-4 bus-fan) already pass on main; the split itself is an explicit deferred `TODO(SEA-1788)` in `go/e2e/legthreefour_test.go`. This record settles only the observation MECHANISM the frozen [`compass-dogfood-e2e`](../compass-dogfood-e2e/design.md) record deferred — it does not reopen that record's tier or its assertion. ### What it decides - **Seam:** an env-gated (`COMPASS_TEST_DISPATCH_TAP`) test-only tap wrapping the delivery `ControlDispatcher` in compass-server records per-recipient op-kind over a unix socket; production wiring is byte-identical when unset. Composed with `AwaitTurnSettled` on both recipients' existing `SubscribeAgentSession` tails — the tap gives KIND, the tail gives RECEIPT. Store-cursor observation (no op-kind stored) is foreclosed; a public trace-event variant is rejected as surface creep. - **Exclusion is window-scoped** to the peer's turn settle, not an absolute negative — the frozen design permits a steered-but-unacked message to be sweep-redelivered as a plain deliver later. - **Second recipient reuses the leg-3 spawner** (subscribed-but-unmentioned) — no third container. ### Production bug found during recon (T1) The runner gateway's `representable()` still lists `AgentControl_Steer` in its reject set, so every channel `@`-mention steer is silently dropped before reaching a live agent — even though `SteerControl.message` is populated and the agent already consumes it. Verified end-to-end (dispatch → `host.Deliver` → `SendControl` → `control.Send` → reject). It is a stale guard; `TestControlRejectsEmptyVariants` locks it in. T1 un-parks it as this stack's first, independent PR. The three load-bearing forks were put to Matt and ruled; the record ships Approved, freezing on merge as the contract T1-T6 execute against. Spec-impact: none. Refs RIG-1788 Co-authored-by: Matt Wilkinson <matt@rigel.build>
|
Compass engineering docs preview: https://compass-rig-1788-e2e-split-o.compass-eng-docs.pages.dev Deployed from Changed pages: |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Design record for RIG-1788 / dogfood-e2e H4: proving the mention steer/deliver split over the real stack. H4's other assertions (spawn F2 ownership, second container, leg-4 bus-fan) already pass on main; the split itself is an explicit deferred
TODO(SEA-1788)ingo/e2e/legthreefour_test.go. This record settles only the observation MECHANISM the frozencompass-dogfood-e2erecord deferred — it does not reopen that record's tier or its assertion.What it decides
COMPASS_TEST_DISPATCH_TAP) test-only tap wrapping the deliveryControlDispatcherin compass-server records per-recipient op-kind over a unix socket; production wiring is byte-identical when unset. Composed withAwaitTurnSettledon both recipients' existingSubscribeAgentSessiontails — the tap gives KIND, the tail gives RECEIPT. Store-cursor observation (no op-kind stored) is foreclosed; a public trace-event variant is rejected as surface creep.Production bug found during recon (T1)
The runner gateway's
representable()still listsAgentControl_Steerin its reject set, so every channel@-mention steer is silently dropped before reaching a live agent — even thoughSteerControl.messageis populated and the agent already consumes it. Verified end-to-end (dispatch →host.Deliver→SendControl→control.Send→ reject). It is a stale guard;TestControlRejectsEmptyVariantslocks it in. T1 un-parks it as this stack's first, independent PR.The three load-bearing forks were put to Matt and ruled; the record ships Approved, freezing on merge as the contract T1-T6 execute against.
Spec-impact: none. Refs RIG-1788
Co-authored-by: Matt Wilkinson matt@rigel.build