feat(compass-e2e): dogfood harness leg 5 — persist + resume across a container boundary (SEA-1789 H5) - #212
Open
seal-agent wants to merge 2 commits into
Open
Conversation
…container boundary (SEA-1789 H5) Adds the leg-5 persist+resume scenario to the dogfood e2e harness: run a turn, tear the agent's container down, provision a fresh container, and resume the logical session into it, asserting the durable transcript lineage carried across the boundary. - go/e2e/agent_ops.go: new Resume(ctx, containerName, resumeSessionID, initialPrompt) primitive, mirroring StartSession but setting ResumeSessionId. Returns the server-minted live session id (a new id — the durable transcript stays keyed under the original logical id). - go/e2e/legfive_test.go: TestLegFivePersistAndResume. //go:build podman, podmanUsable() skip-guarded (present-but-skipped RED like leg-2/leg-3/4; green gated on the assembled agent image + the full agent-lane). Two distinct canned replies drive the pre-teardown and resumed turns; the transcript under the ORIGINAL logical session id must carry BOTH replies (resume loaded the reconstructed body, not a fresh session), and the resumed lifetime's minted id must differ from the original (the control-plane live id is separate from the durable transcript key). The carried transcript is asserted under the original session id, not the minted resume id: entry_seq is monotonic per logical session across resumes and BindLifetime rebases the new lifetime's frames onto that stable key, so querying the minted id would be a silent false-green. Refs SEA-1789 Co-authored-by: Matt Wilkinson <matt@sealedsecurity.com>
…ency, redundant-pin note (SEA-1789 H5) Two comment-precision fixes from review (both non-gating lows, comment-only, no behavior change): - reply1 assertion: the prior comment overclaimed that a fresh-load resume would make reply1 ABSENT. reply1's durable row survives RemoveWorkspace regardless (RemoveWorkspace does not prune the transcript); its discrimination is contingent on lifetime-2's start-checkpoint carrying the reconstructed body forward. Reworded to state that contingency and to note the primary rebase-onto-logical-id proof is the reply2 assertion. - resumedSessionID != "" guard: noted it is an explicit contract statement for the reader, since AwaitSessionSettled on resumedSessionID above would already fail an empty id before this point. Refs SEA-1789 Co-authored-by: Matt Wilkinson <matt@sealedsecurity.com>
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.
This PR is part of a stack containing 3 PRs:
mainAdds the leg-5 persist+resume scenario to the dogfood e2e harness: run a
turn, tear the agent's container down, provision a fresh container, and
resume the logical session into it, asserting the durable transcript
lineage carried across the boundary.
initialPrompt) primitive, mirroring StartSession but setting
ResumeSessionId. Returns the server-minted live session id (a new id —
the durable transcript stays keyed under the original logical id).
podmanUsable() skip-guarded (present-but-skipped RED like leg-2/leg-3/4;
green gated on the assembled agent image + the full agent-lane). Two
distinct canned replies drive the pre-teardown and resumed turns; the
transcript under the ORIGINAL logical session id must carry BOTH replies
(resume loaded the reconstructed body, not a fresh session), and the
resumed lifetime's minted id must differ from the original (the
control-plane live id is separate from the durable transcript key).
The carried transcript is asserted under the original session id, not the
minted resume id: entry_seq is monotonic per logical session across resumes
and BindLifetime rebases the new lifetime's frames onto that stable key, so
querying the minted id would be a silent false-green.
Refs SEA-1789
Co-authored-by: Matt Wilkinson matt@sealedsecurity.com