chore(deps): upgrade Effect to stable 4.0.0

[juliusmarminge]

Effect 4.0.0 is out as the first stable v4 release. We were on 4.0.0-rc.115. Between the two, Effect removed the effect/unstable/* export paths and the effect/Encoding module, and changed a few APIs we depend on.

Changes

• Imports. All 675 files that imported effect/unstable/* now use the flattened effect/<area> paths. httpapi is now http-api, and arbitrary/Arbitrary is now effect/Arbitrary. effect/Encoding is replaced by effect/encoding/{Base64,Base64Url,Hex}.
• Tracing for the browser trace proxy. HttpRouter.serve now builds its app in a private layer memo map. Because of that, the server-wide TracerDisabledWhen predicate, which was merged into makeRoutesLayer, stopped applying, and the browser OTLP proxy route was traced again. server.test.ts caught this. A new withUntracedRequests helper provides the predicate to the served layer, and both the server and the test use it.
• Closing scopes. Scope.close now only accepts a closeable scope.
  • SSH tunnels take their entry scope explicitly instead of reading it from the context.
  • The Codex session runtime forks a child of the caller's scope so close can still end it.
• Smaller API changes:
  • Effect.partition now returns successes first.
  • Stream.scan takes a lazy initial state.
  • SchemaGetter.onSome is replaced by transformEffect, and .compose by SchemaGetter.compose / SchemaTransformation.composeTransformation.
  • Schema.brand requires a single concrete brand key.
  • Effect.orElseSucceed now passes the error, which needed a type annotation on the WSL test stub.
• Patches, re-ported onto 4.0.0:
  • MCP HTTP DELETE. This is now registered in the HTTP protocol layer against the new internal stateful MCP runtime.
  • RpcClient. The request/ping hooks, the three-missed-pong tolerance and the getSetCookie guard carry over. The ping timeout now uses upstream's SocketReadError reason.
  • @effect/vitest. The patch is still just the vite-plus/test import swap. pnpm rejects catalog: in the @effect/vitest packageExtensions entry once it re-resolves, so vite-plus is pinned there literally.

Known gap: relay

infra/relay still uses alchemy@2.0.0-beta.79, the latest published version. It and its @distilled.cloud/* dependencies import the removed effect/unstable/* paths, so the relay will not typecheck, test or deploy until alchemy publishes beta.80. Alchemy main already supports the 4.0.0 module layout. Its preview tarballs depend on URL-pinned distilled packages, which our blockExoticSubdeps policy rejects, so this PR does not pin them. The follow-up is to bump alchemy once beta.80 is on npm.

.repos/effect-smol will be synced in a separate refs PR, as before.

Verification

• tsc --noEmit is clean for server, web, desktop, mobile, contracts, shared, client-runtime, ssh, tailscale, effect-acp, effect-codex-app-server, scripts and oxlint-plugin-t3code.
• Focused tests pass:
  • server.test.ts, including the OTLP untraced test
  • McpHttpServer.test.ts, including MCP session DELETE
  • session.test.ts, including missed-pong tolerance
  • The Codex adapter, the SSH tunnel and replay markers
  • Schema JSON and the call sites that moved to the new encoding modules

🤖 Generated with Claude Code

---

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ The exact PR base did not have a successful artifact. Baseline uses the latest successful main measurement shown below.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.5 KiB	13.5 KiB	−1 B (−0.0%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.0 KiB	7.1 KiB	+15 B (+0.2%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−16 B (−0.2%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	56.3 KiB	56.2 KiB	−44 B (−0.1%)	66.4 KiB	✅
Codex	Live turn messages	10	9	−1 (−10.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	−1 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	+3 B (+0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−4 B (−0.1%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.0 KiB	57.0 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	9	9	0 (0.0%)	21	✅

Baseline: 408ff8a · PR result: 6512a81 · Source CI: failure

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 114.0 KiB
• Claude decoded thread snapshot: 114.7 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This is a 703-file Effect major-version migration that includes runtime tracing, scope-lifecycle, parsing, and authentication-path changes rather than only import renames. An unresolved High-severity finding also identifies unbounded ACP parser buffering that can exhaust server memory.

Not approved because:

• 1 blocking correctness issue found at or above your repo's Minimum Blocking Severity

^{Adjust the Minimum Blocking Severity for this repo — including turning it Off — in Settings. You can add or adjust custom eligibility rules. Learn more.}

[macroscopeapp]

🟠 High src/protocol.ts:91

An agent that never sends \n causes buffer to grow without limit, so the server eventually exhausts its memory instead of terminating the session. Add the 16 MiB incomplete-frame cap used by ndJsonRpc() and throw when it is exceeded.

 const makeStrictNdJsonRpcParser = () => {
+  const MAX_BUFFERED_FRAME_SIZE = 16 * 1024 * 1024;
   const codec = RpcSerialization.jsonRpc().makeUnsafe();
   const decoder = new TextDecoder();
   let buffer = "";
   return {
     decode: (bytes: Uint8Array | string): ReadonlyArray<unknown> => {
       buffer += typeof bytes === "string" ? bytes : decoder.decode(bytes, { stream: true });
+      if (buffer.length > MAX_BUFFERED_FRAME_SIZE) {
+        throw new Error("Maximum incomplete JSON-RPC frame size exceeded");
+      }
       const lines = buffer.split("\n");

🚀 Reply "fix it for me" or copy this AI Prompt for your agent:

In file @packages/effect-acp/src/protocol.ts around lines 91-100:

An agent that never sends `\n` causes `buffer` to grow without limit, so the server eventually exhausts its memory instead of terminating the session. Add the 16 MiB incomplete-frame cap used by `ndJsonRpc()` and throw when it is exceeded.

[juliusmarminge]

Thanks for working on this. We merged the orchestrator V2 rewrite in #2829, and we are closing this PR as part of that transition.

This PR upgrades Effect across 703 pre-V2 files, modifying 34 files removed by the rewrite and no orchestration-v2 files. The newly merged server must be included in the dependency migration and validation. Rebuild the upgrade on current main.

Sorry for the extra work this creates. If the change is still needed on V2, please rebuild it on current main, verify it there, and open a new PR linking back here. We're closing the current implementation without assuming the underlying request is resolved.