feat(cockpit): aimock harness library + per-example e2e (Phase 2)#356
Merged
Conversation
…le layout Phase 1's single-globalSetup pattern doesn't scale to 15+ cockpit examples. Phase 2 introduces libs/internal/aimock-harness with a createGlobalSetup factory and migrates per-example e2e dirs to live next to each example's Angular app. Streaming gets migrated; c-tool-calls lands as the first new-pattern example. Future phases each add one example as a small additive PR.
11 tasks. Task 0 de-risks path-alias resolution at Playwright runtime (falls back to relative imports if aliases don't work). Tasks 1-5 scaffold + implement the library. Task 6 wires the alias (or skips if relative imports needed). Tasks 7-8 migrate streaming + add c-tool-calls. Tasks 9-10 delete the old layout + update CI. Task 11 verifies + ships.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
…r sequential CI runs CI's sequential per-example e2e loop hit `OSError: Port 8123 is already in use` on the second run. Three compounding causes: 1. The factory spawned langgraph and Angular non-detached, so SIGTERM to the `uv`/`npx` parent didn't propagate to the actual server children (`python langgraph dev`, `node nx serve`). Process tree survived teardown. 2. The teardown's "wait for port free" did a TCP connect-refused check, not a real bind() check. langgraph's _is_port_available does a real bind, which fails on TIME_WAIT sockets that connect refuses don't surface. 3. Even with both fixed, TIME_WAIT sockets on 8123 from the first run's client connections (Playwright + Angular both opened many) blocked langgraph's bind() on the second run for far longer than the 5s sleep between targets. Fixes: - spawn(detached: true) + process.kill(-pid, 'SIGKILL') in teardown to kill the whole process group. - waitForPortFree now does a real bind() check (mirrors langgraph's check). - Each per-example pins its OWN langgraph port: streaming keeps 8123, tool-calls offsets to 8124. Angular proxy.conf.json target updated to match. Future examples pick the next unused port. Decouples examples from each other — TIME_WAIT on one example's port no longer blocks the next example. - CI loop replaced with explicit shell loop (was nx run-many --parallel=1) for clearer per-example failure attribution and a 5s settle between targets. Verified locally: 2-run sequential loop passes consistently.
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.
Summary
Phase 2 of the cockpit aimock e2e plan. Restructures the harness so each cockpit example owns its own e2e dir next to its Angular app, backed by a shared internal library at `libs/internal/aimock-harness`.
Sits on Phase 1 (#349) + the c-* aviation refactor (#347 + #350).
Helper improvement (multi-turn safety)
Mid-implementation discovery: `sendPromptAndWait`'s per-message `data-streaming="false"` wait races on tool-call flows — the first finalized assistant bubble (tool-call chip) trips the helper before the continuation bubble streams in. Fixed by waiting for the SEND button to flip back from "Stop generating" to "Send" — the agent-level idle signal that survives multi-turn flows. Streaming spec (single-LLM-call) still passes; tool-calls spec now 3/3 stable.
Path-alias note for reviewers
Implementer found that TypeScript path aliases work in vitest/tsc but NOT at Playwright config-load time in this repo. The `@ngaf-internal/aimock-harness` alias is kept (helps `tsc` typechecking) but per-example playwright configs use relative imports (`../../../../../libs/internal/aimock-harness/src`). Investigating a Playwright-side fix (e.g., `tsconfig-paths/register`) is out of scope for this PR; relative imports work today.
Test plan
Out-of-scope follow-ups noted during implementation
Spec: `docs/superpowers/specs/2026-05-15-cockpit-aimock-harness-lib-design.md`
Plan: `docs/superpowers/plans/2026-05-15-cockpit-aimock-harness-lib.md`