fix(effect): support stable Effect 4

[NathanFlurry]

• Upgrade @rivetkit/effect and its examples to stable Effect 4 and compatible Vitest 5.
• Make Effect example tests reliable across repeated local Engine runs.

[railway-app]

🚅 Deployed to the rivet-pr-5842 environment in rivet-frontend

Service	Status	Web	Updated
frontend-inspector	✅ Success (View Logs)	Web	Oct 6, 2026 at 12:37 am UTC
mcp-hub	✅ Success (View Logs)	Web	Oct 6, 2026 at 12:36 am UTC
website	❌ Build Failed (View Logs)	Web	Oct 6, 2026 at 12:35 am UTC
ladle	🕒 Building (View Logs)	Web	Oct 6, 2026 at 12:35 am UTC
kitchen-sink	🕒 Building (View Logs)	Web	Oct 6, 2026 at 12:35 am UTC
frontend-cloud	🕒 Building (View Logs)	Web	Oct 6, 2026 at 12:35 am UTC

[claude]

Review: fix(effect): support stable Effect 4

I read the full diff except pnpm-lock.yaml. I did not run CI, the type checks, or the tests, so everything below comes from reading the code.

Things to double check

• API renames. The PR changes Schema.TaggedErrorClass to Schema.TaggedError, Config.string to Config.String, Config.redacted to Config.Redacted, transformOrFail to transformEffect, and effect/unstable/* to effect/*. It also swaps the argument order of SchemaIssue.InvalidValue. The same rename is applied across ~20 error classes, so one mismatch would show up everywhere. Please confirm check-types and the package tests pass against the published effect@4.0.0 in CI.
• Resolved versions. effect: ^4.0.0 as a peer dependency is right for stable. Check that @effect/ai-openai, @effect/platform-node and @effect/vitest all resolve to compatible 4.0.0 versions in the lockfile.
• vitest ^5.0.3 is a major bump across the package, @vitest/coverage-v8, and two examples. It deserves its own attention in CI.

Code quality

• chat-room-effect/tests/chat-room.test.ts builds the endpoint from process.env.RIVET_RUN_ENGINE_PORT ?? "6420" and provides a second Client.layer. This couples the test to the port Registry.test picks internally. If that default changes, the moderator client will silently point at the wrong engine. Prefer reading the endpoint from the test runtime, or add a comment saying where the port comes from.
• Random.nextUUIDv4 was replaced by crypto.randomUUID() in tests and examples. This works, but it moves the examples away from idiomatic Effect. If Random.nextUUIDv4 no longer exists in stable, ignore this. Otherwise a short comment would explain the choice.
• hello-world-effect tests now use random keys. This fixes state bleeding across repeated runs against a persistent local engine, which matches the PR goal.
• The import reordering and type import changes (telemetry.ts, logging.ts, Registry.ts) look like Biome output. They are harmless but unrelated churn.

Docs and tests

• The examples/ai-agent-effect/README.md import path is updated. Check whether anything under website/ or docs/ still references effect/unstable/* or TaggedErrorClass.
• No new tests, which is reasonable for a dependency upgrade. Make sure the *.test-d.ts type tests actually run in CI, since they catch schema and type changes most directly.

No security concerns. This is a mechanical upgrade that looks consistent. The main risk is the API renames, so a green CI run is the thing to wait for.

[the-company-company]

🟠 1 medium-severity finding

_{Reviewed commit ab3f087.}

[the-company-company]

🟠 Medium · Vitest 5 raises the Node floor beyond the declared runtime support

vitest@5.0.3 declares node: ^22.12.0 || ^24.0.0 || >=26.0.0, while this workspace still declares Node >=20 and the rivetkit package declares >=22.0.0. As a result, supported Node 20 and Node 22.0–22.11 environments now install an unsupported test runner for this package and the three updated examples. Since stable @effect/vitest requires Vitest 5, update the relevant engines.node declarations/toolchain floor to >=22.12.0 (and add matching engines to the examples) in the same migration.