Test Ramen

Any coding agent — Claude Code, Cursor, Copilot, Gemini CLI, or a human — can run this. No agent-specific tooling required: a shell and the languages' own toolchains (uv, cargo) are enough.

Inputs

• A working tree of this repo, checked out (submodule or standalone clone).
• uv and cargo installed (rust-toolchain.toml pins the Rust version; uv resolves Python 3.14 itself).
• Optional: docker (for make demo / the compose stack), kubectl + kind (for tests/kind/), a real Postgres if testing storage/postgres.py's live-integration case.

Steps

1. Read facts/STATE.md, then docs/CONTRACTS.md for the area you're touching — the binding spec for what correct behavior looks like.
2. Run make test (equivalently make test-runtime test-node test-console one at a time). Each component enforces --cov-fail-under=90; a run below that threshold is a failure, not a warning.
3. Run make lint — ruff (Python, shared ruff.toml), cargo fmt --check + cargo clippy -- -D warnings (Rust), helm lint (both charts, both providers), terraform fmt -check (both clouds), actionlint and shellcheck when installed.
4. If your change touches tests/ (cross-component conformance/e2e), run cd tests && uv sync -q && uv run pytest -q. Suites that need infrastructure they can't find (a built node-rs binary, a live console URL, a kind cluster) skip with a stated reason — that's expected locally; CI provides what each suite needs.
5. If your change touches multi-zone behavior (autoscale, rebalance, worker routing), bring up a local kind cluster (deploy/kind/up.sh) and run deploy/kind/test.sh — this is the only suite that exercises real Kubernetes HPA/Deployment scaling, not a mock.
6. Write the report as a short paragraph, not a new file: what ran, pass/fail counts, coverage per component, anything skipped and why. If nothing changed since the last run, say so instead of re-running everything from scratch.

Validate

(hand this section, verbatim, to a second agent or a fresh review pass — it should not need the context above)

• V1 make test exits 0; no component's coverage output falls below 90%.
• V2 make lint exits 0.
• V3 If tests/ was touched: cd tests && uv run pytest -q — no unexpected failures (skips with a reason are fine; a failure or an unexplained skip is not).
• V4 If the change affects deploy config (deploy/helm/, deploy/terraform/): helm lint/helm template both providers, terraform validate on any .tf file touched.
• V5 The commit/PR message states the coverage numbers and what, if anything, was skipped and why.

Boundaries

• Never lower the 90% coverage gate or add a blanket coverage exclusion to make a run pass — fix the gap or justify a narrow, commented exclusion for genuinely untestable lines (e.g. a real-cloud-only code path already covered by a fake-backed unit test, matching the pattern in console/src/ramen_console/storage/).
• Never skip a test by commenting it out or deleting it to unblock a run; mark it xfail/skip with a reason and open a note in facts/STATE.md or the PR description instead.
• A kind-cluster test run changes real cluster state (scales deployments). Restore what you changed (kubectl scale) before finishing if a later suite in the same session depends on a clean baseline — a zone left scaled up raises the floor for every test that runs after it.