cloudemu

Imported from stackshy/cloudemu (AGENTS.md). Install upstream with npx skills add stackshy/cloudemu. Copyright stays with the author.

AGENTS.md

Guidance for AI agents working in the cloudemu repository. (Human contributors: see CONTRIBUTING.md.)

Integrating cloudemu into an existing service = an endpoint override on the already-running service (AWS_ENDPOINT_URL / o.BaseEndpoint, option.WithEndpoint, or the Azure ARM endpoint override). Server mode (run the binary/Docker image and point real code at the printed endpoints) is the default for integration/E2E; library/in-process mode (httptest.NewServer, typed mocks) is for Go unit tests written inside cloudemu-aware code. Details: docs/integration.md.

What cloudemu is

Zero-cost, in-memory emulation of AWS, Azure, and GCP cloud APIs. It runs three ways: as a standalone server (the cloudemu serve binary or the ghcr.io/stackshy/cloudemu Docker image) that any app in any language points at, and in-process from Go via either the SDK-compat HTTP server or the typed mock API. It emulates control surfaces, not a real cloud — it does not run workloads/containers, serve real traffic, authenticate requests, or enforce quotas. State is in-memory and resettable, so it is ephemeral by default (lost on process exit unless saved); persistence is opt-in — the whole emulator's state can be snapshotted to one JSON file and restored identity-preservingly across all four providers (--persist, snapshot save/load, the /_cloudemu/snapshot endpoint, the persist package). See the README for the full framing and scope.

Where the capabilities are

Do not scrape prose to answer "what can cloudemu do." Use the generated, can't-drift sources:

• docs/coverage/README.md — human index: every service, every operation, native name per provider.
• docs/coverage/coverage.json — the full capability set, machine-readable (parse this instead of scraping the docs).

These are produced from the driver interfaces in services/*/driver by go generate, so they never promise a capability the code lacks.

Architecture (one paragraph)

Three layers: a portable API (services/<svc>/) wraps a driver interface (services/<svc>/driver/), which each provider implements in providers/{aws,azure,gcp,oci}/<native>/ with memstore-backed mocks. AWS, Azure, and GCP are fully implemented. OCI (providers/oci/) is in progress — its foundation is in place and services land one at a time, so a service not yet built reads as nil; consult docs/coverage/ for which ones exist rather than any prose. Full detail: docs/architecture.md.

Build, test, lint

go build ./...
go test ./...
golangci-lint run --timeout=9m ./...

Run all three before proposing a change. Lint must be clean (0 issues).

Conventions that matter

• Mirror across providers. A behavior added to one provider should be added to AWS, Azure, and GCP unless the capability genuinely doesn't exist there.
• Regenerate coverage after interface or wiring changes. If you touch a services/*/driver interface or wire a service into a provider factory, run go generate ./... and commit the updated docs/coverage/ output.
• Per-service non-goals are hand-maintained in docs/coverage/nongoals/<service>.md and inlined by the generator; the rest of docs/coverage/ is generated — do not edit it by hand.
• Deterministic time via config.FakeClock for time-dependent tests.