Skip to content

docs(stage5a): live demo first, everything else one command#47

Closed
hanwencheng wants to merge 1 commit intomainfrom
docs/stage5a-collapse-into-done-script
Closed

docs(stage5a): live demo first, everything else one command#47
hanwencheng wants to merge 1 commit intomainfrom
docs/stage5a-collapse-into-done-script

Conversation

@hanwencheng
Copy link
Copy Markdown
Member

Summary

Restructures the Stage 5a manual-test guide around its two things that actually matter:

  1. The live demo (real OpenRouter signup) — moved up top, expanded, clearly marked as the centerpiece. Still blocked on the ToS-check TODO.
  2. Everything else — collapsed into a single bash harness/stage-5a-done.sh invocation. No more per-section prose for hermetic tests, CLI, MCP, and metrics.

The doc drops 494 lines and gains 186, net -308. The guide now reads: scope → demo → one-command gate → troubleshooting → 5b forecast → 2-item checklist.

Changes

  • harness/stage-5a-done.sh — grew from 21 → 93 lines to cover every non-live check:

    1. Rust unit tests (4 crates)
    2. TS install + npm test
    3. Phantom-key chaos in isolation
    4. Pattern grep guard
    5. TS typecheck (new)
    6. Clippy on Stage 5a crates with -D warnings (new; --no-deps so pre-existing lints in agentkeys-core/agentkeys-mock-server don't fail the gate)
    7. MCP tools/list asserts agentkeys.provision registered (new)
    8. Observability — asserts the three core provision_metric names are emitted (new)

    Colored banners + ✓/✗ per step; final STAGE 5a PASSED on green, first failure aborts with a clear red marker.

  • docs/manual-test-stage5.md — rewritten:

    • § 1 — The demo. Lead section. Prereqs, expected output, failure modes. DO NOT RUN YET banner unchanged.
    • § 2 — Everything else — one command. Single bash harness/stage-5a-done.sh invocation, with a table explaining what each of the 8 steps asserts.
    • § 3 — Troubleshooting, trimmed to the items that still apply in the script-driven flow.
    • § 4 — Stage 5b forecast — unchanged.
    • § 5 — Summary checklist — 2 items: script green, demo end-to-end (once ToS clears).

Test plan

  • bash harness/stage-5a-done.shSTAGE 5a PASSED (exit 0, all 8 steps green). Smoke-ran twice.
  • bash -n harness/stage-5a-done.sh clean (syntax).
  • Clippy with --no-deps confirmed to pass on Stage 5a crates even though agentkeys-core has a pre-existing ptr_arg warning (out of scope).
  • MCP tools/list subprocess returns a response containing "name":"agentkeys.provision".
  • Observability grep finds tier_used, duration_seconds, verification_result in the stores_credential test's --nocapture output.

No code changes outside the test harness + the doc. The live demo path is unchanged and remains the follow-up once the OpenRouter ToS check clears.

🤖 Generated with Claude Code

@hanwencheng
Copy link
Copy Markdown
Member Author

Superseded by the new PR on docs/stage6-aws-setup. The harness/stage-5a-done.sh refactor from this PR is byte-identical to what landed on main via #48 and is already on the new branch. The docs/manual-test-stage5.md scope has been absorbed and further evolved on the new branch (plus-addressing demo path, Stage 5b + Stage 6 migration notes, backend selector docs).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant