Skip to content

Add bounded parallel E2E runner for local and CI use - #362

Draft
guima-why wants to merge 39 commits into
mainfrom
feature-aone-e2e
Draft

Add bounded parallel E2E runner for local and CI use#362
guima-why wants to merge 39 commits into
mainfrom
feature-aone-e2e

Conversation

@guima-why

@guima-why guima-why commented Sep 26, 2026 •
edited
Loading

Copy link
Copy Markdown
Collaborator

Changes

  • Add one local/CI E2E runner with bounded parallelism, per-case hard timeouts, process-tree cleanup, an explicit case allowlist, and Markdown/HTML/JUnit/JSON reports.
  • Select deterministic A2A, REPL and Web API contracts plus the A2A execution-control matrix; add four opt-in live scenarios, including two ROS resource creation cases.
  • Adapt the REPL pipeline contract to accept either semantic candidate controls or terminal readiness across PTY implementations.

Verification

  • 128 targeted offline tests passed; Ruff passed.
  • Local full deterministic suite: 25/25 passed with three concurrent cases.
  • CI fast suite: 5/5 passed. CI full suite: 25/25 passed in 71 seconds of test wall time; the test report and artifact upload succeeded.
  • Live cloud suite is prepared but awaits test-only credentials.

Notes

  • a2a-recovery-contract is excluded after reproducible failure in isolation: its expected persisted aliyun_api result is missing.
  • Real cloud cases enforce a 45-minute hard timeout and allow up to 15 minutes for cleanup before killing the process group. Check cleanup status for any forced termination.

@guima-why
guima-why force-pushed the feature-aone-e2e branch 5 times, most recently from 32d13c8 to 710fba6 Compare October 5, 2026 12:28
@guima-why
Add bounded parallel E2E runner and repair pipeline recovery
0a8232b 
Add isolated local and CI E2E execution, model pools, grounded question assistance, safe diagnostics, native acceptance evidence, and ownership-verified cleanup. Preserve E2E telemetry identities and adapt model thinking contracts.

Repair execution ownership, abandoned question streams, rollback attempt isolation, candidate completion and selection validation, and REPL input readiness. Keep acceptance conditions enforceable and reject contradictory success reports or unverified cleanup.

Make offline image and backup/concurrent-cancellation tests independent of installed Chinese fonts and scheduler startup timing.
guima-why added 23 commits October 5, 2026 22:01
@guima-why
Fix AGUI lead-in accounting and capture provider failure facts
1030ccd
@guima-why
Use resolved network subnet in recovery question answers
1eac4b8
@guima-why
Expose safe pre-selector permission diagnostics
357de4c
@guima-why
Capture precise schema, adjustment preview and quote guard diagnostics
ad90764
@guima-why
Answer native REPL questions while awaiting pipeline completion
053ea5b
@guima-why
Preserve native stack polling status and REPL candidate handoff trace
7d03921
@guima-why
fix(pipeline): retain schema failure coordinates for model repair
98c4879
@guima-why
fix(e2e): check actual Step 1 materialization calls
d8c71ae
@guima-why
fix(e2e): wait for native multimodal question checkpoints
4da7995
@guima-why
fix(e2e): validate advisory question selections before use
57d524c
@guima-why
fix(e2e): fail promptly on deployment replanning and retain failure f…
9ee9cf8 
…acts
@guima-why
fix(e2e): allow scoped local materialization in AGUI selector flow
4e935a6
@guima-why
test(e2e): retain safe instance geometry and constraint evidence diag…
28da655 
…nostics
@guima-why
fix(a2a): tolerate null bytes in public path projection text
7c2dbd0
@guima-why
test(e2e): retain golden source and rollback stream failure evidence
60bf882
@guima-why
fix(e2e): answer planning handoffs before rollback Step 2 fault
de4c1d9
@guima-why
fix(e2e): handle clarification and image question input boundaries
a98a020
@guima-why
fix(e2e): retain native question boundary during image interruption
74da4c3
@guima-why
fix(e2e): submit image paths without controls in active questions
b1e3c75
@guima-why
test(e2e): retain native evidence for target, quote and confirmation …
fa90c1c 
…failures
@guima-why
fix(e2e): report native reconfirmation instead of waiting for completion
480c31e
@guima-why
fix(e2e): parse product handoff JSON before safety instructions
7e69ac8
@guima-why
test(e2e): correlate native confirmation failures with completion guards
55dc9bf
guima-why added 15 commits October 6, 2026 13:19
@guima-why
test(e2e): classify localized native confirmation guard failures
e8e7c2f
@guima-why
fix(e2e): refresh fixture ownership before publishing network facts
7f8141e
@guima-why
fix(e2e): distinguish unspecified restrictions from missing question …
ad589cb 
…facts
@guima-why
fix(e2e): select only available network fixtures predating the run
88e2ece
@guima-why
fix: read coherent OAuth refresh state and control projection fixture…
4992e73 
… timestamps
@guima-why
fix: bound Windows subnet reservation lock acquisition retries
0dd1725
@guima-why
fix(e2e): honor deferred identities in planning preference answers
1058cb0
@guima-why
test(skill): check manager reuse before enabling idle expiry
2fefe7a
@guima-why
fix(e2e): use external input for image parameter questions
c123568
@guima-why
fix(e2e): answer current cloud question and retain native failure dia…
cfce304 
…gnostics
@guima-why
test(e2e): write AGUI diagnostic fixture with explicit UTF-8
e04eae3
@guima-why
fix: observe accepted AGUI pipeline continuation and answer current p…
16575f9 
…arameter
@guima-why
test: cancel permission snapshot batch at the actual tool boundary
d4f682f
@guima-why
fix: observe resumed pipeline past stale input status
ed28fb0
@guima-why
test: synchronize cancellation with runtime factory lifecycle
0918c50

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@guima-why