P6b — Playwright + MCP integration spec

Branch: feat/p6b-playwright-e2e Status: SHIPPED (this branch) Worktree: /root/damascus-orchestrator-p6b Base: main @ acec3ea (P6a merged)

Background

PR #20 (cfcd571, "Damascus Entry Points P6: E2E verification") already shipped the P6b deliverables on main — tests/e2e/test_entry_points_e2e.py (667 lines, 4-phase Playwright + MCP test) and tests/e2e/conftest.py. The P6b kanban card was drafted before the P6 split landed, so the body overlaps with P6 instead of complementing it.

P6b's contribution on this branch is therefore a re-verification plus a few small improvements:

Re-verification against post-PR-#21 main — the test runs end-to-end against the stack as it exists after the Ask-Hermes UX PR (#21) merged, and it still passes (3 back-to-back clean runs at 29–33s each).
DAMASCUS_ROOT / DAMASCUS_EVIDENCE_NAME env vars — the test now reads these from the environment instead of hardcoding /root/damascus-orchestrator. Same file is now reusable from a worktree.
tests/e2e/requirements.txt — pinned deps for a fresh venv.

Changes on this branch vs `main`

docs/P6B.md                              | new (this file)
tests/e2e/requirements.txt                | new (pinned deps)
tests/e2e/test_entry_points_e2e.py       | 6-line patch: env-var indirection

The patched test runs identically against main (where the env vars default to the original paths). Run from the worktree with:

cd /root/damascus-orchestrator-p6b
DAMASCUS_ROOT=/root/damascus-orchestrator-p6b DAMASCUS_EVIDENCE_NAME=p6b \
    python3 -m pytest tests/e2e/test_entry_points_e2e.py -q -s

Evidence (on disk, gitignored)

.hermes/evidence/p6b/
├── README.md                              (run instructions + AC checklist)
├── pytest.log                             (3rd consecutive green run, 29.35s)
└── screenshots/
    ├── 01_dashboard.png
    ├── 01_ingest.png
    ├── 02_build.png
    ├── 03_review.png
    ├── 04_merged.png
    ├── 05_awaiting_human_drawer.png
    └── 06_answered.png

7 screenshots + pytest.log prove the test ran green against the live stack on 2026-06-26 14:29 UTC. The .hermes/evidence/ tree is gitignored (see .gitignore line 46), so evidence is intentionally not committed — it regenerates from the test.

Acceptance criteria

pytest tests/e2e/test_entry_points_e2e.py -q -s exits 0 (last run: 1 passed in 29.35s).
All 7 screenshots present in .hermes/evidence/p6b/screenshots/.
MCP stdio subprocess communicates cleanly (no init-error logs).
Spec uses live stack (api at 127.0.0.1:9110, MCP launched in stdio against the api container).
No browser console errors during Phase 2 / 3.

PR description (draft)

Damascus Entry Points — P6b: Playwright + MCP integration spec

Re-verifies the existing P6 e2e test (tests/e2e/test_entry_points_e2e.py, shipped via PR #20) against the post-PR-#21 stack and adds a tiny ergonomic improvement: DAMASCUS_ROOT and DAMASCUS_EVIDENCE_NAME are now read from the environment so the same test is reusable from a worktree without forking it. Also adds tests/e2e/requirements.txt pinning the test deps.

Three back-to-back clean runs at ~30s each against the live stack. Evidence (screenshots + pytest.log) regenerated on the worktree at .hermes/evidence/p6b/ (gitignored by design).

Complements P6a (scripts/verify.sh, bash recipe) and P6 itself (the test file already on main).

Notes

The P6b kanban task's body describes the test as "outstanding work" but the file has been on main since 2026-06-25 via PR #20. The body was drafted before the P6 split, so this branch documents the overlap and ships the small improvement.
CI is intentionally out of scope per the task body. The spec runs locally against a live docker compose up stack.

3.9 KiB Raw Permalink Blame History Unescape Escape