Skip to content

🤖 refactor: Effect Phase 11 PR 5 — DesktopLive group layers + DesktopWiringLive; thin ServiceContainer; StreamManager runner param - #4061

Merged
ThomasK33 merged 1 commit into
mainfrom
effect-phase11-pr5-desktop-layers
Sep 2, 2026
Merged

🤖 refactor: Effect Phase 11 PR 5 — DesktopLive group layers + DesktopWiringLive; thin ServiceContainer; StreamManager runner param#4061
ThomasK33 merged 1 commit into
mainfrom
effect-phase11-pr5-desktop-layers

Conversation

@ThomasK33

@ThomasK33 ThomasK33 commented Sep 2, 2026

Copy link
Copy Markdown
Member

Summary

Effect migration Phase 11, PR 5 of 6. The desktop-only tail of the service graph now builds as Effect Layers too: six Layer.effectContext group layers (BrowserLive · DesktopBridgeLive · TerminalEditorLive · MiscDesktopLiveOauthLive · WorkersLive) construct the ~40 remaining desktop services with their existing argument lists, DesktopWiringLive replays the former ServiceContainer constructor's setter / listener / global-registration statements verbatim and in order after CoreLive, and the ServiceContainer constructor shrinks to makeAppRuntime(AppLive(stores)) plus field assignment from the built context (toORPCContext() unchanged in shape; initialize()/dispose()/shutdown() untouched, §5 order fixed). StreamManager gains an optional trailing runner: EffectRunner: its partial-write debounce forks through it and AgentSession hands the same runner to its RetryManager, so both sleep on the app runtime's Clock (a TestClock in tests).

Stacked on PR 1 #4049, PR 2 #4050, PR 3 #4051, PR 4a #4054, PR 4b #4057. Plan: <details> at the bottom (§2.1/§2.2, §2.3 invariants, §3 "PR 5", §5, §7).

Implementation

  • di/tags.ts — 47 new Context.Service tags (type-only imports; §2.1 naming; Tag suffix where the bare name is not a *Service class or would shadow — WindowTag, QuickJSRuntimeFactoryTag, DesktopSessionManagerTag, …), grouped as BrowserTags | DesktopBridgeTags | TerminalEditorTags | MiscDesktopTags | OauthTags | WorkerTags = DesktopTags; AppTags = CoreRootTags | CrossCuttingTags | DesktopTags.
  • di/layers/desktop.ts — six Layer.effectContext group layers (each yields its inputs, constructs several services in the constructor's original order, returns a Context; the R annotation on each group is its declared dependency set); DesktopWiringLive (Layer.effectDiscard, yields every collaborator first, then the wiring statements); composition below.
  • di/layers/app.tsAppLive = DesktopLive ▹ CoreLive ▹ CoreOptionsFromDesktopLive ▹ CrossCuttingLive ▹ MemoryMetaLive ▹ runtimeSeams. core.tsStreamManagerLive passes yield* EffectRunnerTag (5th ctor arg); CoreInputTags gains EffectRunnerTag.
  • serviceContainer.ts — constructor = build + 66 this.x = get(Tag) lines; service imports type-only; the never-read private ptyService field is gone (the PTY lives in the graph under PTY for TerminalService). initialize(), toORPCContext(), shutdown(), dispose() byte-identical.
  • streamManager.tsconstructor(…, eventSink = () => undefined, runner: EffectRunner = defaultEffectRunner), public readonly effectRunner; schedulePartialWrite's runSync(forkIn)/runFork and interruptPartialWriteFiber's runFork(Fiber.interrupt) go through it; both Scope.close sites stay Effect.runFork (async-close precedent). agentSession.ts/retryManager.tsAgentSessionStreamManager gains readonly effectRunner?: EffectRunner; new RetryManager(…, this.streamManager.effectRunner) (undefined → RetryManager's default, so doubles and the aiService fallback are unchanged).
  • TestsserviceContainer.test.ts: exhaustive Record<keyof Omit<ORPCContext, "headers" | "effect/context" | "effect/wrap">, Tag> identity test (a new ORPC field without a tag fails to compile), a desktop-wiring behavioral test (bindings, every set* collaborator, idle-compaction outcome forwarding, SSH-prompt global registration observed via isInteractiveHostKeyApprovalAvailable(), a timing listener), and a dispose()/shutdown() order test via spies on the public methods already spied today; the original assertions are unchanged. streamManager.test.ts: the debounce fires on the injected runner's TestClock (red-checked against the global-runtime fork).

PR 5 notes

Group DAG (re-derived from the constructor argument lists)

Group Services, in the former constructor's order Declared requirements (R)
BrowserLive BrowserBridgeTokenManager → AgentBrowserSessionDiscovery → BrowserControl → BrowserSessionStateHub → BrowserBridgeServer Config
DesktopBridgeLive DesktopSessionManager → DesktopTokenManager → DesktopBridgeServer Config, Experiments, Workspace
TerminalEditorLive PTY → Terminal → Editor → Tokenizer → Instructions Config, SecretsStore, Workspace, SessionUsage, AI, Provider
MiscDesktopLive QuickJSRuntimeFactory, SshPrompt, Window, Backup, AgentPluginInstall, Project (needs SshPrompt), Update, Server, MenuEvent, Voice, Coder (singleton), ServerAuth, WorkspaceLifecycleHooks, WorktreeArchiveSnapshot Config, SecretsStore, ProvidersConfigStore, Experiments, Policy, Provider, MCPServerManager, WorkspaceMcpOverrides
OauthLive McpOauth → MuxGatewayOauth → MuxGovernorOauth → CodexOauth → CoderOauth → CopilotOauth Config, ProvidersConfigStore, FileLeaseManager, MCPConfig, Provider, Policy, Telemetry, Window
WorkersLive IdleCompaction → Heartbeat → Timeline → Refine (needs Timeline) → AgentStatus (needs Tokenizer, Window) Config, EffectRunner, Experiments, History, ExtensionMetadata, Workspace, Task, IdleDispatcher, Memory, MemoryMeta, AI, SessionUsage, Tokenizer, Window
DesktopWiringLive the former constructor's 12 wiring blocks, verbatim, in order — incl. #4043's backupService.setProjectService/setMemoryNotifier after the projectService.set* lines (rebased) — runs after CoreLive, so core listeners still precede desktop ones Config, CrossCuttingTags, CoreTags, DesktopTags
DesktopBase  = Layer.mergeAll(MiscDesktopLive, BrowserLive, DesktopBridgeLive, TerminalEditorLive)   // true siblings
DesktopUpper = Layer.mergeAll(OauthLive, WorkersLive).pipe(Layer.provideMerge(DesktopBase))          // both need Base
DesktopLive  = DesktopWiringLive.pipe(Layer.provideMerge(DesktopUpper))
AppLive      = DesktopLive ▹ CoreLive ▹ CoreOptionsFromDesktopLive ▹ CrossCuttingLive ▹ MemoryMetaLive ▹ (AppFiberScope ▹ EffectRunner ▹ Stores)

"True siblings" was checked constructor by constructor (I6 table): no base-group constructor takes or calls another desktop service, so their relative build order is a don't-care. The two upper groups' edges (WindowService, TokenizerService) are the only cross-group dependencies and are expressed with provideMerge, never with mergeAll argument order.

I6 constructor side-effect audit (38 constructions moved here; full 38-row table with file:line cites in the first PR comment)

Group Constructors (args exactly as the former constructor passed them) Beyond capturing args Order that matters
Browser BrowserBridgeTokenManager() · AgentBrowserSessionDiscoveryService({resolveWorkspaceCandidatePathsFn}) · BrowserControlService({discovery, resolveSessionEnvFn}) · BrowserSessionStateHub({control}) · BrowserBridgeServer({discovery, tokenManager, stateHub}) token manager: own unref'd cleanup setInterval (as before); bridge server: unattached WebSocketServer({noServer}) intra-group arg order only
DesktopBridge DesktopSessionManager({config, experimentsService, workspaceService}) · DesktopTokenManager() · DesktopBridgeServer({sessionManager, tokenManager}) token manager: own unref'd cleanup setInterval (as before) intra-group arg order; core Workspace (staging)
TerminalEditor PTYService() · TerminalService(config, pty, secretsStore) · EditorService(config, workspaceService) · TokenizerService(sessionUsage, ai, provider) · InstructionsService(config, ai, tokenizer) none (the tokenizer worker is created at workerPool.ts import time, not by the ctor — see Observations) intra-group arg order
Misc QuickJSRuntimeFactory() · SshPromptService() · WindowService() · BackupService(config, {gitRepo, payload}) · AgentPluginInstallService(config, {isEnabled, mcpServerManager, workspaceMcpOverridesService}) · ProjectService(config, sshPrompt, secretsStore) · UpdateService(config) · ServerService() · MenuEventService() · VoiceService(config, provider, policy, providersStore) · coderService · ServerAuthService(config) · WorkspaceLifecycleHooks() · WorktreeArchiveSnapshotService(config) AgentPluginInstallService: un-awaited startup journal reconcile + module-level discovery gate (as before, from its own args); UpdateService: config.getUpdateChannel() + un-awaited initialize() (no-op outside Electron) Project after SshPrompt (same group, in order); AgentPluginInstall after core (staging)
OAuth McpOauthService(config, mcpConfig, window, telemetry) · MuxGatewayOauthService(providersStore, provider, window) · MuxGovernorOauthService(config, window, policy) · CodexOauthService(providersStore, provider, window) · CoderOauthService(providersStore, fileLeaseManager, provider, window, policy) · CopilotOauthService(provider, window) CoderOauthService subscribes providerService.onConfigChanged (coderOauthService.ts:373) — a declared core dependency; AIService's own subscription (S3) still precedes it because every desktop group builds above CoreLive; no other desktop ctor subscribes to providerService after Misc (WindowService) via OauthLive ▹ DesktopBase
Workers IdleCompactionService(config, history, extensionMetadata, executeIdleCompaction, runner) · HeartbeatService(config, extensionMetadata, workspace, task, idleDispatcher, runner) · TimelineService(config, history, experiments) · RefineService(config, memory, memoryMeta, history, ai, experiments, {timeline, sessionUsage, emitChatMessage, acquireTurnExclusion}) · AgentStatusService(config, history, tokenizer, extensionMetadata, workspace, window, ai, {sessionUsage, requestAnalyticsIngest}) none (scopes/fibers/intervals start in start(); subscribeToWorkspace is a wiring statement) Refine after Timeline (same group); after Misc (Window) + TerminalEditor (Tokenizer) via WorkersLive ▹ DesktopBase

No moved constructor reads a setter-provided collaborator or registers listeners on workspaceService/aiService/taskService/memoryConsolidationService/mcpServerManager (only CoderOauthServiceproviderService, above). Wiring statements that used to sit between constructions now run after all of them; none of the constructors that followed them read the wired state, so the observable order of effects is unchanged.

Split decision

Raw product diff is +1149 / −517 (8 files), above the plan's ~600-line heuristic; I evaluated a 5a/5b split along group boundaries and kept one PR: the surface is mechanical relocation (~330 wiring + ~250 constructor-call lines moved verbatim, 184 lines of tags — git diff --color-moved=dimmed-zebra origin/main -- src/node/services/serviceContainer.ts src/node/services/di/layers/desktop.ts dims them), a mergeable 5a would need a throwaway hybrid constructor, and the wiring move would still land whole in one half.

Deviations from the plan / observations

  • RetryManager site. The plan places the runner hand-off in streamManager.ts; the constructor is in agentSession.ts — reached via the optional AgentSessionStreamManager.effectRunner field.
  • Tokenizer worker starts ~1 s later in xum server/ACP (not desktop). workerPool.ts creates the tokenizer Worker at import time; main reached it via serviceContainer.ts's early TokenizerService import (~0.9 s after spawn, before effect/di/* loaded), now via core.tsaiServicehistoryServicetokenizer (~1.9 s). Total startup is unchanged (spawn → initialize completed ≈ 2.55 s on both) and the desktop is unaffected (desktop/main.ts imports tokenizer first), but a SIGTERM during the worker's ≈19 s encoding load waits for its current module evaluation on both trees (so a "1 s after init" probe read 0.6 s vs 1.6 s — strace: the gap sits between exit(0) and the worker thread's exit, not in dispose(), which is 70–120 ms on both). Steady-state shutdown is unchanged (table). Left as is: pre-existing import-time worker creation, unrelated to the composition root; an explicit warm-up call site is a follow-up candidate, not a bug.
  • DesktopWiringLive is a Layer.effectDiscard over an Effect.gen body whose only yields are service tags (synchronous); no finalizers, no forks (I5) — same shape as CoreWiringLive.

Pre-review audits (plan §3)

  1. Interruption posture — moved forks: StreamManager.schedulePartialWrite (runner.runSync(forkIn(…, resourceScope)) / whitebox runner.runFork) and interruptPartialWriteFiber (runner.runFork(Fiber.interrupt)) — unsupervised through the runner exactly as through the global runtime; interrupted by the stream's resource-scope close and by re-arm, unchanged. RetryManager forks through the injected runner; cancelled by cancel()/dispose() as before. No forks in di/layers/.
  2. Uninterruptible teardowndispose()/shutdown() bodies byte-identical; the §5 order is now asserted.
  3. No defect escapes — no new Promise facades; makeAppRuntime stays the one throw site (throwing-layer test passes through the deeper graph).
  4. Spy-seam checkrg 'spyOn\(' src/node/services/serviceContainer.test.ts tests/ipc tests/ui: every target is a public method on an instance the container still exposes → intercepted, since each field is the context instance (identity test). Arity: only StreamManager gained a trailing optional param; AgentSessionStreamManager gained an optional readonly field. Typecheck of every test proves it.
  5. Sync-start — the new debounce test pins that partialWriteFiber is armed synchronously and fires on TestClock.adjust.
  6. I6 — table above. 7. I3memoryConsolidationService.ts not in the diff.

Re-recorded gate numbers (R7/R8)

Sibling worktrees under one scratch dir with shared node_modules, interleaved runs; origin/main = 1c81235c1 (4b) vs this branch. Host: 96 cores, CPU PSI some avg60 ≈ 39–41 %.

metric origin/main (4b) branch (PR 5) note
tsgo --noEmit wall, 3 interleaved pairs (median, min–max) 12.01 s (11.47–14.04) 11.94 s (11.82–14.71) flat
tsgo --extendedDiagnostics (renderer) types 1 930 432 · check 10.50 s types 1 934 836 (+0.23 %) · check 9.43 s noise
new ServiceContainer(stores) in-process, 3 runs × 15: cold (first) median 27 ms (24–29) 35 ms (35–44) +≈8 ms cold — Layer first-use for 7 more layers + 3 composition nodes (trend 12 → 18 → 23 → 27 → 35 ms across PR 3/4a/4b/main/PR 5; main's 27 includes the imperative desktop constructor)
warm median (min) 1.87 ms (1.11) 2.18 ms (1.65) +≈0.3 ms
[startup] AppRuntime built in xum server (10 runs) 11 ms (core only) 16 ms (whole graph) not comparable (main excludes the desktop ctor)
spawn → AppRuntime built / → initialize completed (3 pairs) 2.35 s / 2.55 s 2.32 s / 2.55 s unchanged
ServiceContainer.initialize completed { totalMs } (10 runs) 245 (215–336) 243 (215–279) unchanged code
SIGTERM → exit, steady state (25 s after init; 5 pairs) 151 ms (134–185), exit 0 ×5 169 ms (149–179), exit 0 ×5 noise; [shutdown] AppFiberScope closed → explicit steps → [shutdown] AppRuntime disposed in every transcript

A chained (provideMerge-only) composition of the same six groups costs the same (30–42 / 2.1–3.0 ms): the +8 ms is Effect first-use, not sibling concurrency.

Lessons for PR 6 (TestClock sweep + shutdown hardening + DI contract docs)

  • StreamManager takes a runner now: the partial-write debounce cases in streamManager.test.ts can move to makeTestEffectRunner() (5th ctor arg; the new test is the template); RetryManager gets its runner from streamManager.effectRunner, so agentSession harness tests can inject a TestClock through a stream-manager double.
  • The tokenizer worker is created at import time (workerPool.ts); [shutdown] timing probes must wait for its ≈19 s load or they measure its module-evaluation tail — PR 6's per-step [shutdown] lines will expose the AppRuntime disposedprocess.exit gap.
  • Record<keyof Omit<ORPCContext, …>, Tag> is the ORPC exhaustiveness guard (exclude effect/wrap with headers/effect/context). Six group layers + three composition nodes cost +8 ms cold / +0.3 ms warm — record the per-layer first-use cost in the contract doc.

Validation

  • make static-check green. bun test gate (streamManager*, aiService, serviceContainer, coreServicesRoot, di/*, retryManager, heartbeat, idleCompaction, cli/server, cli/cli) 375/375; all 26 agentSession* suites 279/279; bun test src/node/services src/cli src/node/orpc src/node/acp 7057 pass / 15 fail — the known host baselines (taskGitPatchEngine ×2, WorkspaceTurnManager ×2, agent_skill_delete, BackupRepoCache ×9) + one attachmentService.completedReports flake that passes in isolation on both trees.
  • TEST_INTEGRATION=1 bun x jest tests (tests/ipc + tests/ui): 669 pass / 77 fail / 49 skipped — all environment baselines: provider-backed suites (403 Forbidden from the AI bridge / missing xAI key), SSH/Docker rows, four src/**/__tests__ bun:test files jest picks up, and terminal.test.ts (1) · sendModeDropdown.test.ts (1) · reportRelocation.test.ts (1), which fail identically on pristine origin/main (re-run in the foreground in the sibling worktree). CI is the lane for the provider suites.
  • Dogfooding (headless Coder host, XUM_LOG_LEVEL=debug DEV_SERVER_SANDBOX_ARGS=--clean-projects make dev-server-sandbox): log order AppRuntime built { ms: 24 }initialize starting → six step durations → initialize completed { totalMs: 233 }; no ManagedRuntime disposed/defect lines. agent-browser: loaded the app (v0.28.3-nightly.148-28-g909dadd90), added a scratch git repo as a project, sent "Reply with exactly the single word: pong" → worktree workspace created, model replied pong, Stats tab populated (screenshot). oRPC Effect path: memory experiment enabled, memory.savesetPinned truelist (pinned: true) → setPinned false over /orpc (all handlerGen + runtime effect/context), then pinned/unpinned from the Memory tab; memory-meta.json flipped pinned true → false (screenshot). Graceful quit: SIGTERM → [shutdown] AppFiberScope closed { ms: 1 }AgentStatusService stoppedterminateAll()[analytics-worker] Shutting down, closing DuckDB[shutdown] AppRuntime disposed { ms: 7 } → nodemon clean exit, 191 ms, exit 0; plus the 5 steady-state pairs in the table (exit 0 ×10). Not exercised headless: Electron before-quit (same dispose(); tests/e2e in CI).

pong reply

memory pinned

pr5-smoke.webm

Risks

  • Low–medium. The one behavioral surface is the wiring relocation: every statement is verbatim and in order, the constructors that used to run between wiring lines are audited as not observing them (I6), the tests/ipc behavioral gate matches pristine main, and the desktop wiring test pins each collaborator edge. dispose()/shutdown() are unchanged and their order is asserted.
  • Startup: +≈8 ms cold construction; initialize() unchanged; tokenizer-worker import-order shift in xum server/ACP (observation above) — no functional change.

📋 Implementation Plan

Effect migration — Wave 3 / Phase 11: ManagedRuntime + Layer dependency injection

0. Summary

Replace the two hand-written composition roots (createCoreServices + the ServiceContainer constructor) with an Effect Layer graph built once per process by a ManagedRuntime ("AppRuntime"), while keeping every service class, constructor signature, Promise facade, private method, and test seam compatible. The runtime becomes (a) the owner of the app-lifetime Scope, (b) the provider of "effect/context" for oRPC Effect-native handlers, and (c) the source of two runtime seams: an EffectRunner (context-bound, unsupervised runner that lets clock-driven workers run on a TestClock) and an AppFiberScope (a runtime-owned, supervised scope whose close is awaited by dispose() — the slot the streamManager engine core will occupy later).

Six stacked, independently mergeable PRs. Product PRs keep existing tests unchanged; only the final test-modernization PR edits tests. Net product LoC ≈ +420 (per-PR estimates below). Service classes are not rewritten — Layers are thin adapters around existing constructors; cycle-breaking setter wiring moves into explicit "wiring layers" that replay today's order.

Unlocks (not done here): streamManager ENGINE CORE conversion, TestClock for timing suites, app-lifetime scopes.

1. Verified current state (evidence)

  • Roots. src/node/services/coreServices.ts:103-389 (createCoreServices: 25 constructions, 12 turnRequestBuilderBindings writes, ~14 setters) and src/node/services/serviceContainer.ts:161-575 (45 more constructions; aiService.on(...)/workspaceService.on(...) analytics wiring at 474-574; global registrations setGlobalCoderService/setSshPromptService at 469-471). new ServiceContainer(stores) is called by headlessEnvironment.ts:111, tests/ipc/setup.ts, src/cli/server.ts:132, src/node/acp/serverConnection.ts:155, src/desktop/main.ts:653; src/cli/run.ts:661 and src/cli/workflow.ts:376 call createCoreServices directly. ⇒ two graph roots (App vs Core), five process entry points, all constructing synchronously.
  • Startup. ServiceContainer.initialize() (577-642) awaits six initialize()s (no try/catch; failure propagates to main.ts:1255-1265 "Startup Failed" dialog + quit; server.ts/ACP log and exit), then sync start()s idleCompaction/heartbeat/agentStatus, then two fire-and-forget sweeps. All constructors are synchronous; two have side effects on declared constructor dependencies only (AIServicestreamManager.setEventSink, WorkspaceServicebackgroundProcessManager.on/aiService.on).
  • Teardown. dispose() (746-779) is explicit and hand-ordered (backgroundProcessManager.beginShutdown() MUST be first — it is a latch protecting persisted monitor records; bridges stop before sessions close; terminateAll late; timelineService.flush() last). shutdown() (718-732) is a second sequence fired concurrently by a second before-quit listener (main.ts:1321). main.ts:1296-1304 races dispose() against 5 s then app.quit(); cli/server.ts:227-268 has a 5 s process.exit(1) force timer; tests/ipc cleanup calls dispose() then shutdown(); headlessEnvironment.dispose never calls services.dispose().
  • Existing Effect surface. 25 files import effect. Only Context.Service tag: MemoryMeta (src/node/orpc/effectContext.ts:21). handlerGen (@orpc/experimental-effect) runs Effect.runPromiseExit per request and Effect.provides opts.context["effect/context"]. streamBridge.ts runs streams on the global runtime. Scope-owning workers: heartbeatService.ts:134-243, idleCompactionService.ts:86-122 (Scope.makeUnsafe + Effect.runSync(Scope.close(..)), valid only because their fibers suspend solely on the clock), oauthFlowManager.ts:164, streamManager.ts:4767/4054 (already Effect.runFork(Scope.close(..)) — the async-close precedent). memoryConsolidationService.ts:667-703, 837-860: check-and-reserve funnels with zero suspensions before inFlight.set/harvestInFlight.set.
  • effect@4.0.0-rc.112 API (verified in node_modules/effect/dist). Context.Service<Self, Shape>()("id") (module Context, not ServiceMap); Layer.{succeed,sync,effect,effectContext,effectDiscard,provide,provideMerge,mergeAll,build,buildWithScope} (no Layer.scoped; Layer.effect strips Scope from R); ManagedRuntime.make(layer){ runSync, runSyncExit, runFork, runPromise, runPromiseExit, contextEffect, cachedContext, scope, dispose(), disposeEffect }; Effect.{runSyncWith,runForkWith,runPromiseWith,runPromiseExitWith}(context); Effect.context<R>(); Effect.serviceOption; Scope.{fork,forkUnsafe,close,provide}; TestClock from effect/testing (layer, adjust, setTime, withLive); Clock.Clock is a Context.Reference (defaulted; TestClock.layer() overrides it).
  • ManagedRuntime internals the design relies on (ManagedRuntime.js): make creates scope = Scope.makeUnsafe("parallel") and layerScope = Scope.forkUnsafe(scope, "sequential"); the first runX forks a build fiber over Layer.buildWithMemoMap — a fully synchronous layer graph builds synchronously, so runtime.runSync(Effect.context()) succeeds and sets cachedContext; afterwards every runX is Effect.run…With(cachedContext) (no extra async boundary). Fibers started through runtime.runX are registered in scope (onFiberStart: Fiber.runIn(scope)). dispose() = Scope.close(scope) (interrupt registered fibers in parallel → layer finalizers sequentially in reverse), after which any runtime.runX dies with "ManagedRuntime disposed".
  • Layer composition semantics. Layer.mergeAll(A, B) is not a dependency resolver: B's requirements are not satisfied by A's outputs; requirements bubble up. Dependencies are satisfied only via Layer.provide/provideMerge chains. Siblings in mergeAll may build concurrently.
  • Test seams that pin signatures (Explore report): private-method spies (Config.saveConfig, WorkspaceService.retireKernelWorkflowRunReferences/startStartupRecovery/createSession/updateAgentStatus, MCPServerManager.startServers, AgentPluginInstallService.reconcileJournals, …); module-level export spies (agentStatusService.generateWorkspaceStatus, sshConnectionPool.verifyHostKeyAgainstPolicyEffect, …); direct construction in tests (Config 44 files, HistoryService 22, MemoryMetaService 11, WorkspaceService 7, IdleDispatcher 6, StreamManager 4, ServiceContainer 3); partial-mock casts (InitStateManager 193, AIService 158, TaskService 149, ORPCContext 62). effectBridge.test.ts:24-30 builds a partial ORPCContext via buildOrpcEffectContext + as unknown as ORPCContext.
  • Timing probes (TestClock candidates): heartbeatService.test.ts 6 real sleeps, idleCompactionService.test.ts 2, retryManager.test.ts 3 setSystemTime, streamManager.test.ts 7 (partial-write debounce), streamBridge.test.ts 11 (heartbeat ticker), OAuth device-flow suites 14 (non-goal).

2. Target architecture

2.1 Building blocks (all under src/node/services/di/; the only directory allowed to import Layer/Context/ManagedRuntime/TestClock)

Module Contents
tags.ts One Context.Service tag per service class provided by the graph. Type-only imports of service classes ⇒ no runtime import cycles. Ids "xum/<Name>". Naming: class name minus trailing Service (MemoryMeta, Workspace, History); classes without that suffix or colliding with an exported name get a Tag suffix (ConfigTag, StreamManagerTag, IdleDispatcherTag). Exports the unions CoreTags and AppTags.
effectRunner.ts interface EffectRunner { runSync<A,E>(e: Effect<A,E,never>): A; runSyncExit; runFork; runPromise; runPromiseExit } — a context-bound, unsupervised runner whose methods accept only effects with no service requirements (R = never; defaulted references like Clock do not appear in R). That makes "not a service locator" type-enforced: a fiber that needs services must take them as explicit constructor dependencies and, if it must be awaited on shutdown, fork into AppFiberScope. defaultEffectRunner = the global Effect.runX (today's exact behavior). effectRunnerFromContext(ctx) = Effect.run…With(ctx). EffectRunnerTag + EffectRunnerLive = Layer.effect(EffectRunnerTag, Effect.map(Effect.context<never>(), effectRunnerFromContext)), placed at the base of the graph so the captured context contains only refs (Clock, later Logger/Random) plus stores. Fibers forked through it are owned by the worker's own Scope (explicit start/stop), not by the ManagedRuntime; runtime.dispose() does not interrupt them. Services import only this file from di/.
appFiberScope.ts AppFiberScopeTag: Scope.Closeable. AppFiberScopeLive = Layer.effect(AppFiberScopeTag, Effect.gen(function*(){ const parent = yield* Effect.scope; return yield* Scope.fork(parent, "parallel"); })) — a child of the runtime's layer scope. Fibers forked into it via Effect.forkIn(_, appFiberScope) are interrupted and awaited when the scope closes. This is the supervised seam for I/O-suspended fibers (engine core, later). ServiceContainer.dispose() closes it explicitly and early (§5) so interrupted fibers can still use their dependencies during finalization; runtime.dispose() later re-closes it idempotently as a backstop. No production occupant in Phase 11; the seam exists with tests.
appRuntime.ts makeAppRuntime(layer): ManagedRuntime.make(layer) + eager synchronous build (runtime.runSync(Effect.context<R>()); assert(runtime.cachedContext !== undefined)); a layer body that suspends is a programming error and throws here — exactly where a throwing constructor throws today, so every entry point's existing catch/dialog/log path is preserved. disposeAppRuntime(runtime, timeoutMs) and closeScopeBounded(scope, timeoutMs) share one shape: Effect.uninterruptible teardown shell around Effect.interruptible(target.pipe(Effect.timeout(timeoutMs))) where target is runtime.disposeEffect resp. Scope.close(scope, Exit.void) (never a non-cancellable JS Promise wrapper); Effect.catchTag("TimeoutError", …) + Effect.catchDefectlog.warn; run via Effect.runPromise; never rejects; idempotent (Scope.close is idempotent; disposeEffect is guarded by a latch). Verify the exact rc Effect.timeout error type at implementation time (rc.112: fails with Cause.TimeoutError, _tag: "TimeoutError"). Module doc comment = the DI contract (§2.3, §5).
layers/stores.ts StoresLive(stores: ConfigStores) = Layer.mergeAll of Layer.succeed for ConfigTag, SessionLocatorTag, ProvidersConfigStoreTag, SecretsStoreTag, FileLeaseManagerTag (true siblings — no inter-dependencies). StoresFromCoreOptionsLive reproduces the opts.x ?? new X(config.rootDir) defaults of coreServices.ts:106-112 for the CLI root.
layers/core.ts CoreOptionsTag (today's CoreServicesOptions minus stores — carries the optional cross-cutting services exactly as today). PR 3: CoreProjectionLive = Layer.effectContext(...) wrapping the existing createCoreServices body and returning a Context<CoreTags> (coarse projection, zero behavior change). PR 4: peel into per-service Layer.effect(Tag, Effect.gen(...)) layers composed in explicit dependency stages (Layer.provideMerge between stages; Layer.mergeAll only for true siblings within a stage — every sibling claim below was checked against the constructor argument lists in coreServices.ts and must be re-checked in the PR): S1 History · InitState · Provider · BackgroundProcess · ExtensionMetadata · MemoryMeta · TerminalAttention · IdleDispatcher · WorkspaceMcpOverrides(default) · TurnRequestBuilderBindingsTag (Layer.succeed(_, {})) → S2a SessionUsage · Goal · Memory → S2b StreamManager (needs SessionUsage) → S3 AIService → S4 Consolidation · MCPConfig → S5 MCPServerManager → S6 Workspace → S7 Task → S8 TurnManager → CoreWiringLive (Layer.effectDiscard, Effect.sync only — no acquireRelease, replays coreServices.ts:137-166, 209-210, 258-270, 288-325, 349-352, 360-367 in order).
layers/desktop.ts CrossCuttingLive (policy, telemetry, experiments, backup, sessionTiming, analytics, devTools, workspaceMcpOverrides, browserBridgeTokenManager), CoreOptionsFromDesktopLive (derives CoreOptionsTag from those tags + extensionMetadataPath), then group layers (Layer.effectContext returning a Context of several tags, constructed in today's order): BrowserLive, DesktopBridgeLive, OauthLive, WorkersLive (idleCompaction, heartbeat, agentStatus, timeline, refine), TerminalEditorLive, MiscDesktopLive; staged with provideMerge where one group needs another. DesktopWiringLive (Effect.sync only) = setters + aiService.on/workspaceService.on/memoryConsolidationService.on wiring + global registrations.
layers/app.ts AppLive(stores) = DesktopLive ▹ CoreLive ▹ CoreOptionsFromDesktopLive ▹ CrossCuttingLive ▹ AppFiberScopeLive ▹ EffectRunnerLive ▹ StoresLive(stores) — read X ▹ Y as "X is provided with Y, and both stay exposed", i.e. X.pipe(Layer.provideMerge(Y)) (rc.112 signature: provideMerge(that: provider)(self: consumer); the right-hand operand is the dependency). Every keeps all tags visible in the final Context<AppTags>.
testEffectRunner.ts (test helper, sibling of testHistoryService.ts) makeTestEffectRunner(){ runner, adjust(duration), setTime(ms), dispose } over one memoised ManagedRuntime.make(EffectRunnerLive.pipe(Layer.provideMerge(TestClock.layer()))) (the TestClock is the provider; the runner captures it), so the worker under test and TestClock.adjust share one TestClock.

2.2 Composition roots after Phase 11

flowchart TB
  Stores["StoresLive(stores)<br/>Config · SessionLocator · ProvidersConfigStore · SecretsStore · FileLeaseManager"]
  Runner["EffectRunnerLive (unsupervised, ref-bound)<br/>+ AppFiberScopeLive (supervised, closed on dispose)"]
  Cross["CrossCuttingLive (desktop only)<br/>Policy · Telemetry · Experiments · Analytics · SessionTiming · DevTools · WorkspaceMcpOverrides · Backup"]
  Opts["CoreOptionsTag<br/>desktop: derived from CrossCutting · CLI: Layer.succeed(opts)"]
  Core["CoreLive<br/>PR 3: coarse CoreProjectionLive → PR 4: stages S1…S8 + CoreWiringLive"]
  Desk["DesktopLive — group Layers<br/>Browser · DesktopBridge · OAuth · Workers · TerminalEditor · Misc → DesktopWiringLive"]
  RT["AppRuntime = ManagedRuntime.make(AppLive)<br/>eager sync build · Context<AppTags> = oRPC effect/context · dispose() last"]
  Stores --> Runner --> Cross --> Opts --> Core --> Desk --> RT
  CLI["CLI root (xum run / xum workflow)<br/>createCoreServices(opts) = makeAppRuntime(CoreLive ▹ StoresFromCoreOptionsLive ▹ AppFiberScopeLive ▹ EffectRunnerLive ▹ succeed(CoreOptionsTag, opts))"]
  Core -.same Layer definitions.-> CLI
Loading

ServiceContainer keeps its public fields and the synchronous new ServiceContainer(stores): the constructor calls makeAppRuntime(AppLive(stores)), stores this.serviceContext = runtime.runSync(Effect.context<AppTags>()), and assigns fields via Context.get(this.serviceContext, Tag). toORPCContext() returns the same plain fields plus "effect/context": this.serviceContext. initialize() is untouched. dispose() follows §5.

createCoreServices(opts) keeps its signature and return shape plus runtime and appFiberScope fields; cli/run.ts:1574-1580 and cli/workflow.ts:275-320 cleanup lists gain closeScopeBounded(appFiberScope) before session.dispose() and disposeAppRuntime(runtime) as the final step (PR 3).

Staged composition skeleton (PR 4 shape; direction matters):

// Each stage depends only on stages defined above it. `provideMerge` keeps both sides exposed.
const S1 = Layer.mergeAll(HistoryLive, InitStateLive, ProviderLive, /* … true siblings only */);
const S2a = Layer.mergeAll(SessionUsageLive, GoalLive, MemoryLive).pipe(Layer.provideMerge(S1));
const S2b = StreamManagerLive.pipe(Layer.provideMerge(S2a));          // StreamManager needs SessionUsage
const S3 = AIServiceLive.pipe(Layer.provideMerge(S2b));
// … S4 … S8 likewise …
export const CoreLive = CoreWiringLive.pipe(Layer.provideMerge(S8));  // wiring runs after every service exists

oRPC typing. OrpcEffectServices (in effectContext.ts) becomes AppTags, so ORPCContext["effect/context"]: Context<AppTags> is satisfied by the runtime context in production. buildOrpcEffectContext stays as the narrow test helper it already is (its only caller, effectBridge.test.ts:24-30, deliberately builds a partial context and casts it via unknown); no production caller remains after PR 1.

2.3 Invariants (the "DI contract"; enforced by tests and the appRuntime.ts doc comment)

# Invariant Constraint served
I1 Phase 11 compatibility contract, not permanent law: layer bodies are synchronous (Layer.succeed/Layer.sync/Layer.effect over sync effects; acquireRelease with a sync acquire is fine). makeAppRuntime asserts the eager build completed. Future async resource acquisition belongs in initialize()/startup effects or an explicit async factory root (ServiceContainer.create()), never silently inside a layer. #2 sync-start, #5 startup parity
I2 Services never hold the ManagedRuntime. Workers hold an EffectRunner (default defaultEffectRunner); EffectRunner.runXEffect.run…With(ctx) — same sync-start semantics as Effect.runX, and still valid after runtime.dispose(), so late callbacks cannot hit "ManagedRuntime disposed". Supervision, when needed, is explicit via AppFiberScope. #2, #3
I3 Per-call pipelines (Effect.runPromise(this.effects…) facades) and the memoryConsolidationService funnels are untouched. Audit item: no DI lookup, runner call, or await may be inserted before inFlight.set / harvestInFlight.set. Only lifecycle forks in workers move to this.runner.runX. #1, #2
I4 Constructors, facades, private methods, module exports unchanged; new constructor parameters are optional, trailing, defaulting to defaultEffectRunner. #1, #6
I5 Teardown order stays explicit in dispose()/shutdown(). Layer bodies and wiring layers register no finalizers in Phase 11 (Effect.sync only), so runtime.dispose() reorders nothing. The one supervised resource (AppFiberScope) is closed explicitly at a fixed position in dispose() (§5). #3
I6 Wiring layers replay today's setter/listener order; a constructor may touch only its declared dependencies (built earlier by staging). Per-PR audit: grep each moved constructor for calls on setter-provided collaborators → forbidden. Dependency order is expressed only with provide/provideMerge stages; never rely on mergeAll sibling order. #6
I7 No persisted-data changes; DI is in-process only. #4
I8 Every process root builds from the same Layer definitions (CoreLive shared by App and CLI). Unit harnesses (createTestHistoryService, createTestToolConfig, createAgentSessionHarness, …) intentionally bypass Layers. #7

2.4 Decisions and alternatives (product-LoC deltas)

D1 — Granularity: coarse core first (PR 3), per-service core stages behind a decision gate (PR 4), group layers for the desktop tail (PR 5)

Honest framing: the three unlocks (engine-core async scope, TestClock, app-lifetime scope) are delivered by AppRuntime + EffectRunner + AppFiberScope and do not require per-service layers. Per-service core layers are migration leverage: typed requirement sets for the engine-core work, per-service swap in integration tests, explicit dependency stages instead of implicit ordering.

  • (A) Per-service everywhere (~70 layers): +900/−700. Desktop tail has hand-tuned teardown that must not become finalizers, so per-service there buys uniformity only. Rejected.
  • (B) Recommended: PR 3 coarse CoreProjectionLive (+120/−10) delivers the shared root and runtime ownership; PR 4 peels the core into staged per-service layers (+330/−290) only if PR 3's typecheck/startup budgets hold (gate in §3); desktop tail as ~6 group layers (+170/−150). Tags for all services either way (~3 LoC each).
  • (C) Coarse only: stop after PR 3 + desktop projection (~+200 total). Cheapest; the engine-core phase would then redo dependency declarations. Remains the fallback if PR 4's gate fails.
D2 — Async init stays an explicit `initialize()`; Layers construct only

Folding initialize() into layer construction would make the build asynchronous (breaks I1), change failure semantics (today: fail-fast → dialog/log), and move the six-step order into memoised builds. Deferred; a later phase can turn initialize() into runtime.runPromise(startupEffect) with per-step Effect.timeout.

D3 — Optional cross-cutting services stay optional via `CoreOptionsTag`, not `Effect.serviceOption`

Core layer bodies read opts.policyService etc. exactly as today, so CLI (absent) vs desktop (present) behavior is unchanged and no service gains a new undefined branch.

D4 — Two seams instead of one: `EffectRunner` (unsupervised, clock-bound) + `AppFiberScope` (supervised)

A single "runtime handle" conflates two needs. Workers need which clock (TestClock) and must keep sync stop(); the engine core needs who awaits me on shutdown. Explicit Clock injection per worker was rejected (a provideService(Clock.Clock, …) at every fork site, and it does not extend to other refs).

D5 — oRPC: `effect/context` = the runtime's `Context`; `handlerGen` unchanged

handlerGen already Effect.provides the context per request; providing ~70 entries instead of one is one Map merge per request. The existing echoAsync/echoEffect probes record the delta as a diagnostic in the PR body (no stable benchmark harness exists to make it a hard gate). effect/wrap not needed.

3. Phasing — six stacked PRs

Every PR: make static-check; gate suites below; existing tests unchanged (PR 6 is the only PR that edits tests, and only to replace real-timer probes). Before @codex review, run the house pre-review audits:

  1. Interruption posture — list every new/moved fiber fork; state what interrupts it and when (unsupervised via EffectRunner + worker scope, or supervised via AppFiberScope).
  2. Uninterruptible teardown — teardown effects are Effect.uninterruptible end-to-end; bounded waits inside use Effect.interruptible(Effect.timeout(...)) (house shape from 🤖 refactor: convert memoryConsolidationService and workspaceStatusGenerator internals to Effect; make in-flight run-lock reservation deterministic #4038).
  3. No defect escapesdisposeAppRuntime/closeScopeBounded and every Promise facade fold defects; makeAppRuntime is the one place allowed to throw (constructor semantics).
  4. Spy-seam checkrg 'spyOn\(' src/node/services/<touched>.test.ts tests/ per touched class; constructor arity and private-method Promise signatures unchanged (typecheck of tests proves it).
  5. Sync-start check — a fork through EffectRunner runs to its first sleep before runFork returns (mirrors heartbeatService.ts:199-202).
  6. Constructor side-effect audit (I6) for every constructor moved into a Layer in that PR.
  7. Zero-suspension audit (I3) whenever memoryConsolidationService is in the diff.

PR 1 — Skeleton: AppRuntime + Stores/MemoryMeta layers + runtime-backed effect/context + dispose hook (+~150 LoC)

Scope

  • di/tags.ts (ConfigTag, SessionLocatorTag, ProvidersConfigStoreTag, SecretsStoreTag, FileLeaseManagerTag, MemoryMeta moved from orpc/effectContext.ts, which re-exports it; AppTags union).
  • di/layers/stores.ts (StoresLive), di/layers/core.ts with MemoryMetaLive = Layer.effect(MemoryMeta, Effect.map(ConfigTag, c => new MemoryMetaService(c.rootDir))), di/layers/app.ts (AppLive(stores) = MemoryMetaLive ▹ StoresLive).
  • di/appRuntime.ts (makeAppRuntime, disposeAppRuntime); APP_RUNTIME_DISPOSE_TIMEOUT_MS in src/constants/.
  • coreServices.ts: CoreServicesOptions.memoryMetaService? (precedent: workspaceMcpOverridesService?).
  • serviceContainer.ts: build runtime first, pass Context.get(ctx, MemoryMeta) to createCoreServices, public readonly runtime, toORPCContext()["effect/context"] = this.serviceContext, dispose() appends disposeAppRuntime behind a disposed latch; new log.debug("[startup] AppRuntime built", { ms }).
  • orpc/effectContext.ts: OrpcEffectServices = AppTags; buildOrpcEffectContext retyped/test-helper doc.
  • headlessEnvironment.dispose calls await services.dispose() before removing the temp dir (the bench harness currently leaks the container; runtime ownership starts here).

Acceptance

  • di/appRuntime.test.ts: (a) sync build sets cachedContext; (b) a layer with an async body makes makeAppRuntime throw synchronously (I1 enforced); (c) probe layers' finalizers run in reverse order on dispose; (d) dispose is idempotent and bounded (hung finalizer → warn, resolves at the timeout); (e) runtime.runFork after the eager build starts synchronously.
  • serviceContainer.test.ts: Context.get(toORPCContext()["effect/context"], MemoryMeta) === services.memoryMetaService; dispose() closes the runtime; dispose(); shutdown() (tests/ipc order) is clean; a throwing layer surfaces as a synchronous throw from new ServiceContainer(stores) (same shape as today's constructor throw → existing entry-point catch paths).
  • effectBridge.test.ts, memoryMeta*.test.ts unchanged and green; echo-probe overhead recorded in the PR body.
  • Gate: bun test src/node/services/di src/node/services/serviceContainer.test.ts src/node/orpc src/node/services/memoryMeta* · make test-integration · make static-check.

Rollback: git revert; classes untouched.

PR 2 — Runtime seams: EffectRunner + AppFiberScope; TestClock on idleCompaction/heartbeat/retryManager (+~140 LoC)

Scope

  • di/effectRunner.ts, di/appFiberScope.ts; AppLive gains AppFiberScopeLive ▹ EffectRunnerLive at the base; ServiceContainer exposes appFiberScope (used only by dispose() in Phase 11) and closes it per §5.
  • IdleCompactionService, HeartbeatService, RetryManager: trailing optional runner: EffectRunner = defaultEffectRunner; every lifecycle Effect.runSync/runFork in start/stop/schedule/cancel becomes this.runner.runX. Deadline math (Date.now()/injected now) unchanged. ServiceContainer passes Context.get(ctx, EffectRunnerTag) to the two workers; RetryManager keeps the default until PR 5 (so streamManager.ts is untouched here).
  • di/testEffectRunner.ts helper.

Acceptance

  • New TestClock tests (existing real-timer tests untouched — they exercise the defaultEffectRunner path, which is production behavior wherever no runner is injected): heartbeat STARTUP_DELAY_MS → first tick after adjust, one tick per CHECK_INTERVAL_MS, no ticks after stop(); idleCompaction initial delay + cadence; retryManager fires exactly at delayMs, cancel() before adjust never fires.
  • Pin runtime facts: runner.runSync(Scope.close(scope, Exit.void)) completes synchronously for a fiber suspended on a TestClock sleep; runFork through the runner reaches its first sleep synchronously; Effect.context<never>() inside EffectRunnerLive sees the upstream TestClock (else the helper provides Clock.Clock explicitly — same seam, one line).
  • AppFiberScope contract tests: (i) an I/O-suspended fiber (interruptible Effect.async that never resolves, with a cancel path) forked with Effect.forkIn(_, appFiberScope) is interrupted and awaited by closeScopeBounded(appFiberScope) — and this happens before the explicit teardown steps in dispose() (assert ordering against a spy on desktopBridgeServer.stop); (ii) a fiber forked via EffectRunner is not interrupted by either close (documents the asymmetry); (iii) disposeAppRuntime afterwards idempotently re-closes the already-closed child scope (no error, no second finalizer run).
  • If TestClock.adjust leaves continuations pending, the helper adds Effect.yieldNow/Fiber.await — decided by tests.
  • Gate: heartbeatService.test.ts, idleCompactionService.test.ts, retryManager.test.ts, serviceContainer.test.ts, di/*, tests/ipc.

Rollback: revert restores defaults; no call site depends on the new params.

PR 3 — Shared core root: coarse CoreProjectionLive + createCoreServices facade + CLI runtime disposal (+120 / −10)

Scope

  • Tags for the remaining 19 core services; CoreOptionsTag; StoresFromCoreOptionsLive.
  • CoreProjectionLive = Layer.effectContext(Effect.gen(function*(){ const opts = yield* CoreOptionsTag; const stores = yield* …; const core = buildCoreGraph({ ...opts, ...stores }); return Context.make(History, core.historyService).pipe(Context.add(...)) })) where buildCoreGraph is today's createCoreServices body, unchanged, renamed.
  • createCoreServices(opts) = makeAppRuntime(CoreProjectionLive ▹ StoresFromCoreOptionsLive ▹ AppFiberScopeLive ▹ EffectRunnerLive ▹ Layer.succeed(CoreOptionsTag, opts)), returns today's CoreServices object read from the context plus runtime and appFiberScope. cli/run.ts and cli/workflow.ts cleanup lists append closeScopeBounded(appFiberScope) before session.dispose() and disposeAppRuntime(runtime) after backgroundProcessManager.terminateAll().
  • ServiceContainer stops calling createCoreServices; AppLive = CoreProjectionLive ▹ CoreOptionsFromDesktopLive ▹ CrossCuttingLive ▹ … (cross-cutting services move into CrossCuttingLive now because core options derive from them). Desktop constructions otherwise stay in the constructor.

Acceptance

  • Identity test: every CoreServices field === Context.get(ctx, Tag); serviceContainer.test.ts unchanged and green.
  • Decision gate for PR 4 recorded in the PR body: make typecheck wall time, [startup] AppRuntime built ms and initialize totals vs origin/main baseline from the sandbox (§7). Proceed to PR 4 only if typecheck regresses < 10 % and startup within noise; otherwise stop at (C).
  • Gate: bun test src/node/services, src/cli/*.test.ts (run/workflow/server/cli), tests/ipc, make static-check.

Rollback: revert restores the imperative call; PR 1/2 unaffected.

PR 4 — Peel the core into staged per-service Layers + CoreWiringLive (+330 / −290 ⇒ net ≈ +40; split 4a/4b if > ~600 diff lines)

Scope

  • Stages S1, S2a, S2b, S3…S8 (§2.1 + skeleton in §2.2) as Layer.effect adapters with today's argument lists; CoreWiringLive (Effect.sync only) replays the wiring lines in order; CoreLive = CoreWiringLive.pipe(Layer.provideMerge(S8)) replaces CoreProjectionLive; buildCoreGraph deleted.
  • Before writing any stage: re-derive the DAG from the constructor argument lists (the plan's stage table was checked once; StreamManager → SessionUsage is the kind of edge that turns "siblings" into a stage split) and record it in the PR body.
  • 4a (S1–S3: leaves through AIService) / 4b (S4–S8 + wiring) if needed — 4a alone is mergeable because the remaining services are built by a shrunken projection layer that reads S1–S3 from the context.

Acceptance

  • Wiring assertions that are behavioral (a missing wiring line fails them): turnRequestBuilderBindings fully populated; goal continuation consumer registered on idleDispatcher; streamManager MCP manager set; registration probe installed on extensionMetadata.
  • I6 audit table for all 19 constructors in the PR body; missing-provider = compile error (R must be never at makeAppRuntime) demonstrated by a type-level test (// @ts-expect-error).
  • Gate: as PR 3 plus streamManager*.test.ts, aiService.test.ts, workspaceService*.test.ts.

Rollback: revert to PR 3's projection.

PR 5 — DesktopLive group layers + DesktopWiringLive; thin ServiceContainer; StreamManager runner param (+170 / −150 ⇒ net ≈ +20)

Scope

  • Tags for the 45 desktop services; six group layers (Layer.effectContext, today's construction order inside each; provideMerge between groups that depend on each other); DesktopWiringLive (Effect.sync only) = serviceContainer.ts:209, 263-265, 271, 288-290, 334-340, 348, 365, 375, 381-382, 434, 438-471, 474-574 in order.
  • ServiceContainer constructor = makeAppRuntime(AppLive(stores)) + field assignment from the context. toORPCContext() unchanged in shape.
  • StreamManager: optional trailing runner: EffectRunner; schedulePartialWrite fork (streamManager.ts:1141) and RetryManager construction use it; Scope.close stays Effect.runFork (existing async-close precedent). WorkersLive receives EffectRunnerTag.

Acceptance

  • All four existing serviceContainer.test.ts assertions unchanged; new identity test over toORPCContext() fields vs tags; dispose()/shutdown() call order asserted via spies on the public methods already spied today.
  • I6 audit for the 45 constructors.
  • Gate: tests/ipc + tests/ui (make test-integration), src/cli/server.test.ts, src/cli/cli.test.ts, streamManager*.test.ts, aiService.test.ts.

PR 6 — TestClock adoption sweep + shutdown hardening + contract docs (+~20 LoC product; tests edited)

Scope

  • Replace real-sleep cadence probes with makeTestEffectRunner() in heartbeatService.test.ts, idleCompactionService.test.ts, retryManager.test.ts, and the partial-write debounce cases of streamManager.test.ts; keep one real-timer smoke test per worker (guards the defaultEffectRunner path).
  • cli/server.ts: [shutdown] log lines per step incl. AppRuntime disposed {ms}; confirm the whole dispose() fits the existing 5 s force-exit budget.
  • Finalize the contract doc comment in di/appRuntime.ts (I1–I8, §5).

Acceptance: converted suites have zero setTimeout-based cadence waits (grep in PR body), same assertions; make test-integration green; sandbox startup/shutdown evidence (§7).

4. TestClock story

  • Mechanism. Effect.sleep, Schedule.fixed, Effect.timeout, Clock.currentTimeMillis read the Clock reference from the running fiber's context. Workers that fork through an EffectRunner built under TestClock.layer() run on the test clock; await testRunner.adjust("2 minutes") advances it. Date.now(), setTimeout, setInterval are unaffected — heartbeat deadline math via injected now, AgentStatusService's ref'd setInterval, and backgroundProcessManager stay on real timers/injected timestamps.
  • Benefit now: heartbeatService.test.ts (6), idleCompactionService.test.ts (2), retryManager.test.ts (3 setSystemTimeadjust; Date.now-based retryAt may move to Clock.currentTimeMillis only if a test needs both clocks aligned), streamManager.test.ts debounce cases (7).
  • Deferred: streamBridge.test.ts ticker (11) — needs a context/runner parameter on subscriptionIterable; OAuth device-flow polling and oauthFlowManager.test.ts (25) — non-goal.
  • Stays real: child-process/PTY/WASM/fs-lock waits (backgroundProcessManager 72, quickjsRuntime 26, lock sleeps in workspaceService/taskService), end-to-end suites (tests/ipc, e2e).
  • Pinned in PR 2, not assumed: adjust runs due sleeps and their synchronous continuations before resolving (or the helper yields until they do); Schedule.fixed anchoring under TestClock matches the wall-clock expectations in heartbeatService.ts:149-155; sync Scope.close of a TestClock-suspended fiber completes synchronously.

5. Shutdown protocol

  1. Trigger points unchanged: main.ts before-quit (preventDefault → dispose() raced with 5 s → app.quit(); update-install path fire-and-forget), the second before-quit listener's shutdown() (unchanged, concurrent), cli/server.ts SIGINT/SIGTERM (5 s force exit), ACP close(), tests/ipc (dispose() then shutdown()), headless bench (dispose() from PR 1).
  2. ServiceContainer.dispose() order:
    1. backgroundProcessManager.beginShutdown() — unchanged, first (latch protecting persisted monitor records).
    2. closeScopeBounded(appFiberScope, APP_FIBER_SCOPE_CLOSE_TIMEOUT_MS) — interrupts and awaits supervised fibers while every dependency they might touch during finalization is still alive. No occupants in Phase 11; the position is fixed now so the engine-core phase does not have to re-derive it.
    3. The existing explicit sequence verbatim (desktopBridgeServer.stop()terminateAll()timelineService.flush()).
    4. disposeAppRuntime(runtime, APP_RUNTIME_DISPOSE_TIMEOUT_MS) — closes the runtime scope (interrupts any fiber started via runtime.runX — none long-lived in Phase 11; runs layer finalizers — none in Phase 11 by I5). Hung → warn at the timeout; never rejects.
      Budget: 2 s + 2 s inner bounds inside the callers' 5 s outer budgets; the outer race in main.ts remains the last line of defense.
      Rule for future occupants: anything forked into AppFiberScope must tolerate interruption at any suspension point and must not depend on resources torn down in step 1; anything that needs a Layer finalizer must first prove reverse-construction order is compatible with steps 2–3 (I5).
  3. Latches: disposed makes dispose() idempotent (two before-quit listeners, tests/ipc dispose+shutdown). shutdown() never touches the runtime or AppFiberScope.
  4. Late callers: EffectRunner handles keep working after runtime dispose (I2), so a stray tick()/scheduleRetry() after quit cannot defect. The ManagedRuntime is referenced only by ServiceContainer and the createCoreServices return value.
  5. Worker stop() stays synchronous (runner.runSync(Scope.close)) because their fibers suspend only on the clock. The engine core will fork into AppFiberScope (step 2.2 awaits it) — the reason both seams exist now.
  6. Crash paths: unchanged — uncaughtException/SIGKILL run no finalizers. Finalizers are best-effort; durable state must remain crash-safe without them (AGENTS.md self-healing rule). Nothing in Phase 11 makes a finalizer the sole guardian of durable state.

6. Risk register

# Risk L/I Mitigation
R1 A layer body suspends → runSync throws at startup M/H I1 assert + PR 1 test (b); doc comment; review checklist; entry-point catch paths verified in PR 1
R2 Construction-order side effects differ under staged builds L/H I6 audit per moved constructor; explicit provideMerge stages; wiring layers replay today's order; tests/ipc as behavioral gate
R3 Double teardown (shutdown()dispose(); dispose+shutdown in tests) M/M disposed latch; runtime/AppFiberScope closed only in dispose(); PR 1 test
R4 Late runtime.runX after dispose → defect M/M I2: services hold EffectRunner, never the ManagedRuntime
R5 TestClock semantics differ from assumptions M/L PR 2 pins them before any suite converts; per-suite fallback to real timers
R6 effect v4 RC churn (ContextServiceMap, Layer renames) M/M All Layer/Context/ManagedRuntime/TestClock imports confined to di/; exact pin
R7 Startup latency regression (splash) L/M AppRuntime built ms + initialize totals vs baseline in sandbox; PR 3 gate
R8 Typecheck slowdown from large requirement unions L/L PR 3 gate records make typecheck wall time; fallback (C)
R9 Per-request Effect.provide of a ~70-entry Context L/L echo-probe diagnostic in PR 1/5 bodies
R10 Spy seams / direct-construction tests break L/H I4; optional trailing params; audit 4; typecheck of tests
R11 CLI roots forget to dispose runtime/scope M/L PR 3 wires both cleanups; src/cli/*.test.ts assert the cleanup steps exist
R12 Someone forks long-lived I/O work via EffectRunner expecting dispose to await it M/M Doc on EffectRunner ("unsupervised"); PR 2 asymmetry test; review audit 1

Rollback: PRs are stacked; revert in reverse order (6→1). Service classes are never modified except for optional trailing params, so any revert restores the previous composition root wholesale with no data or API implications.

7. Dogfooding (per PR; evidence attached to the PR body)

Environment (headless Coder host, no DISPLAY):

XUM_LOG_LEVEL=debug DEV_SERVER_SANDBOX_ARGS="--clean-projects" make dev-server-sandbox   # background bash task; prints URL + XUM_ROOT
  • Startup correctness: <XUM_ROOT>/logs/*.log shows, in order: Loading services..., [startup] AppRuntime built {ms}, [startup] ServiceContainer.initialize starting, six step durations, [startup] ServiceContainer.initialize completed {totalMs, stepDurationsMs}. Paste baseline (origin/main) vs branch numbers.
  • Startup-never-crash parity (once, locally, not committed): inject a throwing scratch layer → xum server exits non-zero with the existing logged error and no unhandled-rejection trace; for desktop, confirm by code path (loadServices() rejects → main.ts:1255 dialog) and via src/cli/server.test.ts/ACP tests.
  • UI smoke (agent-browser): open <url>snapshot -i → add a scratch git repo as a project → create a workspace → send one message → screenshot the loaded app and the response; attach_file both. Video: start agent-browser record before the flow and stop it with a hard timeout (timeout 30 agent-browser record stop); if stopping hangs (known), attach the truncated WebM plus the screenshots and say so.
  • oRPC Effect path: pin/unpin a memory entry (rides handlerGen + runtime effect/context); screenshot before/after; grep logs for ManagedRuntime disposed/defect lines (expect none).
  • Graceful quit: record the terminal with script -q /tmp/<workspace>-shutdown.log (or agent-tty if present), kill -TERM <pid> → expect [shutdown] lines, AppRuntime disposed {ms}, exit 0, no force-exit message; attach the typescript. Exercise the timeout branch once with a scratch hung finalizer → warn + timely exit.
  • Electron (best effort): with Xvfb, make dev + agent-browser via CDP (electron skill): screenshot splash → main window, quit via menu, confirm exit < 5 s; otherwise state that the Electron path is covered by tests/e2e in CI and the shared dispose() path exercised by server.ts.

Gate suites per PR (plus make static-check always):

PR Must pass
1 src/node/services/di/*, serviceContainer.test.ts, src/node/orpc/*, memoryMeta*, make test-integration
2 + heartbeatService.test.ts, idleCompactionService.test.ts, retryManager.test.ts
3 + bun test src/node/services, src/cli/*.test.ts; record PR 4 gate numbers
4 + streamManager*.test.ts, aiService.test.ts, workspaceService*.test.ts
5 + tests/ui via make test-integration, src/cli/server.test.ts, src/cli/cli.test.ts
6 converted suites + full make test-integration + sandbox startup/shutdown evidence

8. Non-goals (explicit)

  • streamManager ENGINE CORE conversion (first AppFiberScope occupant; separate phase).
  • Schema at persistence boundaries; OAuth refresh/device-flow workers; AgentStatusService setInterval → Effect.
  • initialize() as a Layer/startup effect (D2); per-service optional tags (D3); streamBridge on the runtime; layer finalizers for existing dispose() steps.
  • Any change to persisted data, IPC wire shapes, or oRPC handler bodies beyond the effect/context source.

9. Assumptions stated

  • Effect.context<never>() inside EffectRunnerLive returns the enclosing build context including an upstream TestClock entry (PR 2 test; fallback: provide Clock.Clock explicitly in the helper).
  • Scope.fork(parent) inside a Layer.effect body yields a child closed by the runtime's layer scope on dispose() (PR 2 AppFiberScope test).
  • Layer bodies never need to observe sibling construction order; all ordering that matters is expressed as provide/provideMerge stages or wiring-layer statement order.
  • EffectRunner's R = never constraint is sufficient for every lifecycle fork in the three Phase 11 workers and StreamManager.schedulePartialWrite (they only use Effect.sleep/Schedule/Effect.sync/Effect.tryPromise — no service tags). Verified by typecheck in PR 2/5.
  • The desktop tail's teardown remains explicit unless a later RFC proves reverse-construction order compatible; this plan does not attempt it.

Generated with xum • Model: anthropic:claude-fable-5-1 • Thinking: xhigh • Cost: $39.39

@ThomasK33

Copy link
Copy Markdown
Member Author

PR 5 notes — full I6 audit table (38 constructions; condensed version in the body)

Legend: args = only captures/derives from its constructor arguments; order = relative construction order that matters and how the layers preserve it.

# Group Constructor (args as passed) Side effects beyond capturing args Setter-provided collaborators read in ctor? Order-sensitivity
1 Browser BrowserBridgeTokenManager() starts an unref'd setInterval cleanup timer (BrowserBridgeTokenManager.ts:32) — same as before no none
2 Browser AgentBrowserSessionDiscoveryService({ resolveWorkspaceCandidatePathsFn }) none (closure over config, lazy) no none
3 Browser BrowserControlService({ browserSessionDiscoveryService, resolveSessionEnvFn }) none no after #2 (arg; same group, in order)
4 Browser BrowserSessionStateHub({ browserControlService }) none (poll timer starts in subscribe()) no after #3 (arg)
5 Browser BrowserBridgeServer({ discovery, tokenManager, stateHub }) new WebSocketServer({ noServer: true }) (unattached) no after #1/#2/#4 (args)
6 DesktopBridge DesktopSessionManager({ config, experimentsService, workspaceService }) none no after core Workspace (staging)
7 DesktopBridge DesktopTokenManager() unref'd setInterval cleanup timer (DesktopTokenManager.ts:20) — same as before no none
8 DesktopBridge DesktopBridgeServer({ desktopSessionManager, desktopTokenManager }) new WebSocketServer({ noServer: true }) no after #6/#7 (args)
9 TerminalEditor PTYService() none no none
10 TerminalEditor TerminalService(config, ptyService, secretsStore) none no after #9 (arg)
11 TerminalEditor EditorService(config, workspaceService) none no after core Workspace
12 TerminalEditor TokenizerService(sessionUsageService, aiService, providerService) none (module workerPool.ts creates the tokenizer worker at import, not in the ctor — see "Observations") no after core S1–S3
13 TerminalEditor InstructionsService(config, aiService, tokenizerService) none no after #12 (arg)
14 Misc QuickJSRuntimeFactory() none no none
15 Misc SshPromptService() none (EventEmitter subclass) no none
16 Misc WindowService() none no none
17 Misc BackupService(config, { gitRepo: createBackupGitRepo(…), payload: createBackupPayloadStore({ config }) }) none (pure adapters) no none
18 Misc AgentPluginInstallService(config, { isEnabled, mcpServerManager, workspaceMcpOverridesService }) kicks off the un-awaited startup journal reconcile (installService.ts:471) and sets the module-level discovery gate (installService.ts:481) — both as before; both derive only from its args no after core MCPServerManager/WorkspaceMcpOverrides (staging)
19 Misc ProjectService(config, sshPromptService, secretsStore) none no (workspace/mcp collaborators arrive via setters in the wiring) after #15 (arg; same group, in order)
20 Misc UpdateService(config) reads config.getUpdateChannel(); un-awaited initialize() that is a no-op outside Electron (Electron: autoUpdater listeners) — as before no none
21 Misc ServerService() none no none
22 Misc MenuEventService() none no none
23 Misc VoiceService(config, providerService, policyService, providersConfigStore) none no none
24 Misc coderService (module singleton, not constructed) none at construction; global registration happens in the wiring n/a none
25 Misc ServerAuthService(config) none no none
26 Misc WorkspaceLifecycleHooks() none no none (hooks registered in the wiring)
27 Misc WorktreeArchiveSnapshotService(config) none no none
28 OAuth McpOauthService(config, mcpConfigService, windowService, telemetryService) none no after Misc (WindowService) — expressed by OauthLive ▹ DesktopBase
29 OAuth MuxGatewayOauthService(providersConfigStore, providerService, windowService) none no after Misc
30 OAuth MuxGovernorOauthService(config, windowService, policyService) none no after Misc
31 OAuth CodexOauthService(providersConfigStore, providerService, windowService) none no after Misc
32 OAuth CoderOauthService(providersConfigStore, fileLeaseManager, providerService, windowService, policyService) subscribes providerService.onConfigChanged(…) (coderOauthService.ts:373) — a declared core dependency; AIService's own onConfigChanged subscription (core S3) still precedes it because every desktop group builds above CoreLive; no other desktop constructor subscribes to providerService no after Misc; after core (staging) — order vs. AIService preserved
33 OAuth CopilotOauthService(providerService, windowService) none no after Misc
34 Workers IdleCompactionService(config, historyService, extensionMetadata, executeIdleCompaction, effectRunner) none (scope/fiber created in start()) no after core
35 Workers HeartbeatService(config, extensionMetadata, workspaceService, taskService, idleDispatcher, effectRunner) none (listeners attached in start()) no after core
36 Workers TimelineService(config, historyService, experimentsService) none (subscribeToWorkspace is a wiring statement) no after core
37 Workers RefineService(config, memoryService, memoryMetaService, historyService, aiService, experimentsService, { timelineService, sessionUsageService, emitChatMessage, acquireTurnExclusion }) none no after #36 (arg; same group, in order)
38 Workers AgentStatusService(config, historyService, tokenizerService, extensionMetadata, workspaceService, windowService, aiService, { sessionUsageService, requestAnalyticsIngest }) none (setInterval starts in start()) no after TerminalEditor (Tokenizer) and Misc (WindowService) — expressed by WorkersLive ▹ DesktopBase

Listener-order check. No moved constructor registers listeners on workspaceService, aiService, taskService, memoryConsolidationService or mcpServerManager; the only constructor-time subscription is #32 on providerService. Every on(...)/subscribeToWorkspace/setter/global registration is a DesktopWiringLive statement, replayed verbatim in the constructor's original order after CoreLive (so the core listeners registered by WorkspaceService/TaskService constructors and CoreWiringLive keep preceding the desktop analytics/timing listeners, as before). Wiring statements that used to sit between constructions (projectService.set* before DesktopSessionManager, setIdleCompactionOutcomeListener before HeartbeatService, timelineService.subscribeToWorkspace before WindowService, mcpServerManager.setMcpOauthService before the other OAuth services, setTerminalService before EditorService) now run after all constructions; none of the constructors that followed them read the wired state (table above), so the observable order of effects is unchanged.

@ThomasK33

Copy link
Copy Markdown
Member Author

@codex review

@ThomasK33

Copy link
Copy Markdown
Member Author

@codex security review

@chatgpt-codex-connector

This comment has been minimized.

@chatgpt-codex-connector

This comment has been minimized.

@chatgpt-codex-connector

Copy link
Copy Markdown

Codex Review: Didn't find any major issues. Keep them coming!

Reviewed commit: 1fa718e20b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

@chatgpt-codex-connector

This comment has been minimized.

…WiringLive; thin ServiceContainer; StreamManager runner param

Six `Layer.effectContext` group layers build the ~40 desktop-only services
(Browser · DesktopBridge · TerminalEditor · MiscDesktop → OAuth · Workers),
`DesktopWiringLive` replays the former ServiceContainer constructor's
setter/listener/global-registration statements verbatim, and the
ServiceContainer constructor becomes `makeAppRuntime(AppLive(stores))` plus
field assignment from the built context. StreamManager gains an optional
trailing `EffectRunner` (partial-write debounce + AgentSession's RetryManager
share the app runtime's clock).
@ThomasK33
ThomasK33 force-pushed the effect-phase11-pr5-desktop-layers branch from 1fa718e to 3543b7b Compare September 2, 2026 13:30
@ThomasK33

Copy link
Copy Markdown
Member Author

@codex review

Rebased onto d813fe124 (#4043): DesktopWiringLive now also replays the two new backupService.setProjectService/setMemoryNotifier lines right after the projectService.set* block (same position as in the former constructor); the wiring test asserts both. No other changes since 1fa718e.

@ThomasK33

Copy link
Copy Markdown
Member Author

@codex security review

@chatgpt-codex-connector

Copy link
Copy Markdown

Codex Review: Didn't find any major issues. Hooray!

Reviewed commit: 3543b7be5d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

@chatgpt-codex-connector

This comment has been minimized.

@ThomasK33
ThomasK33 added this pull request to the merge queue Sep 2, 2026
Merged via the queue into main with commit cd1b11c Sep 2, 2026
20 checks passed
@ThomasK33
ThomasK33 deleted the effect-phase11-pr5-desktop-layers branch September 2, 2026 14:02
asm pushed a commit to asm/mux that referenced this pull request Sep 2, 2026
…n] timing, DI contract docs (coder#4062)

## Summary

Effect migration Phase 11, PR 6 of 6 — the phase's closing PR, and the
only one allowed to edit existing tests. **(1) TestClock sweep:** the
real-timer probes that actually waited on a worker's clock now run on a
`TestClock` through the workers' injected `EffectRunner`
(`makeTestEffectRunner()`), with exactly one default-runner smoke per
worker guarding the production (real-clock) path. **(2) Shutdown
hardening:** every step of `ServiceContainer.dispose()`, the `xum
server` signal handler and the CLI roots' cleanup lists now writes a
`[shutdown] <step> {ms}` debug line (new `shutdownStep` helper, no
suspension added between synchronous steps), which made the
long-standing "SIGTERM ~6 s after startup exits ~11 s later" gap
measurable — and root-caused it to a spot **outside** `dispose()`
(below). **(3) DI contract:** the `di/appRuntime.ts` module comment is
now the durable contract — invariants I1–I8, the two-seam asymmetry, the
§5 shutdown order, the rule for future `AppFiberScope` occupants, the
recorded Layer-machinery costs, and the R6 firewall.

Stacked on PR 1 #4049, PR 2 #4050, PR 3 #4051, PR 4a #4054, PR 4b #4057,
PR 5 #4061 (all on `main`). Plan: `<details>` at the bottom (§3 "PR 6",
§4 TestClock story, §5 shutdown protocol, §2.3 invariants, §7
dogfooding).

## Implementation

- **`retryManager.test.ts`** — the hand-rolled
`setTimeout`/`clearTimeout` spy harness (`runNextTimer()`,
`scheduledTimers`) is gone; every timing case drives the backoff with
`clock.adjust(...)` on a `makeTestEffectRunner()` passed as the 4th ctor
arg. "Timer pending / no timer pending" assertions became
`isRetryPending` plus a negative `adjust` far past any backoff (a
still-armed retry would fire). PR 2's separate
`retryManager.testClock.test.ts` is folded in (its three cases are now
the main suite's "exactly at the backoff delay", "cancel clears pending
retry timer" and "reschedules…" cases) and deleted. One default-runner
smoke remains: it intercepts the real `setTimeout` registration once to
prove the default runner's sleep lands on Effect's default clock with
the backoff delay and fires `onRetry` — without a 2 s wall-clock wait.
`setSystemTime` stays: it only pins `Date.now()` for the `scheduledAt`
equality and never drove timing.
- **`streamManager.test.ts`** — "interrupts a pending debounced partial
write when the stream ends" (the one real cadence wait:
`sleep(throttleMs + 200)` ≈ 720 ms) now injects a TestClock runner (5th
ctor arg, PR 5's template) and proves the negative with `adjust(2 ×
throttle)` — 47 ms. A new default-runner smoke ("a debounced partial
write arms a real setTimeout through the default runner", ≈ 7 ms;
intercepts the real timer registration like the RetryManager smoke, so
there is no wall-clock window to overrun) replaces the real-clock
coverage it took away.
- **`idleCompactionService.test.ts`** — gains the missing default-runner
smoke (`start()` → nothing sweeps within 20 ms of a 60 s initial delay →
synchronous `stop()`); its `testClock` sibling's doc comment claimed the
real-timer suite covered the default runner, but that suite never called
`start()`. **`heartbeatService.test.ts`** — unchanged except a comment
marking "startup does not fire heartbeats immediately" as the retained
smoke (see "What the plan's counts meant").
- **`shutdownStep.ts`** (new) — `shutdownStep(name, run)`: writes
`[shutdown] <name> starting` before `run()`, times it, and writes
`[shutdown] <name> {ms}` on completion, all at debug level — so a hung
step (an awaited disposer that never settles *or* a blocking synchronous
call) is named by the last line before silence. Overloads: a
thenable-returning step (thenable check, not `instanceof Promise`, so a
cross-realm promise is still awaited) is awaited and logged via
`.finally`; a synchronous step is logged before returning with **no
Promise created**, so wrapping one adds no suspension point and adjacent
synchronous teardown statements still run on the same tick (audit 2).
Errors propagate unchanged. The `Promise` overload is declared first
because `Promise<void>` is assignable to `void`;
`@typescript-eslint/no-misused-promises` guards the other direction.
`shutdownStep.test.ts` pins the sync-no-Promise / thenable-awaited /
error-propagation contract.
- **`serviceContainer.ts`** — `disposeOnce()` wraps each of its 21
explicit steps; `closeScopeBounded`/`disposeAppRuntime` keep their own
lines; `[shutdown] ServiceContainer.dispose starting/completed
{totalMs}` bracket the sequence. Order byte-identical (asserted by the
PR 5 order test). **`cli/server.ts`** —
`terminalService.closeAllSessions` and `serverService.stopServer` are
timed and a final `[shutdown] exiting {totalMs}` is the last JS-side
line before `process.exit(0)`. **`cli/runCleanup.ts`** — the loop times
each step; **`cli/workflow.ts`** — `disposeWorkflowResources` now builds
the same kind of step list and runs it through `runBestEffortCleanup`
(same containment as its former eight `try/catch` blocks; warn wording
is now `xum workflow: cleanup step failed: <step>`).
- **`di/appRuntime.ts`** — module doc comment rewritten as the contract
(details in "PR 6 notes").

## PR 6 notes

### Suite wall time, before → after (bun test, 3 runs each, same loaded
host: 96 cores, load ≈ 145, CPU PSI some avg60 ≈ 33 %)

| suite | before (wall, min–max) | after | tests | note |
|---|---|---|---|---|
| `streamManager.test.ts` | 2.47–2.59 s | **1.69–1.78 s** | 122 → 123 |
the 720 ms real wait ("interrupts a pending debounced partial write…")
is now 47 ms on virtual time; +7 ms default-runner smoke |
| `retryManager.test.ts` (+ deleted `retryManager.testClock.test.ts`) |
0.31–0.43 s (+ ≈0.3 s for the separate file) | 0.35–0.42 s | 14 + 3 → 15
| already fake-timer; the value is one harness (`TestClock`) instead of
two, and one file instead of two |
| `idleCompactionService.test.ts` | 0.69–0.73 s | 0.72–0.78 s | 19 → 20
| +20 ms default-runner smoke that did not exist |
| `heartbeatService.test.ts` | 1.78–1.93 s | 1.76–1.77 s | 74 |
untouched (comment only) |

Grep in the converted files (acceptance): `streamManager.test.ts` no
longer waits `setTimeout(…, throttleMs + …)`; `retryManager.test.ts` has
no `spyOn(globalThis, "setTimeout")` outside the single smoke and no
`runNextTimer`.

### What the plan's probe counts meant on inspection (deviations)

The plan's §1 counts (heartbeat 6 · idleCompaction 2 · retryManager 3 ·
streamManager 7) came from a `setTimeout`/`setSystemTime`/fixture grep.
Reading each site:

- **heartbeat "6"** — one is the clock-cadence probe ("startup does not
fire heartbeats immediately", 100 ms) and is exactly the default-runner
smoke to keep; the other five are `waitForCondition` polls and 20 ms
settles on **Promise chains** (`tick()`→`resyncFromConfig`→queue) plus
one 300 ms mock dispatch delay whose assertion is `Date.now()` deadline
math — plan §4 "stays real: injected timestamps". None waits on the
scheduler fiber's clock, so a TestClock changes nothing there. Left as
is.
- **idleCompaction "2"** — both are Promise-settlement waits; the
real-timer suite never called `start()`, so there was **no**
default-runner smoke to keep — added one.
- **retryManager "3"** — the `setSystemTime` calls pin `Date.now()` for
`scheduledAt`; the actual timer surrogate was the global `setTimeout`
spy harness, which is what the TestClock replaces. `setSystemTime` stays
(no test needs the two clocks aligned, so `scheduledAt` stays on
`Date.now()`).
- **streamManager "7"** — seven fixtures set `lastPartialWriteTime`
inside the throttle window, but they call
`attachWorkflowRunToToolCall`/`appendPartAndEmit(…, false)`, which flush
**immediately** and never wait on the debounce. The single real cadence
wait was the 720 ms scope-interrupt probe — converted.
- `agentSession*` harness tests (optional in the plan): their sleeps
wait on `waitForStartupAutoRetryRerunWindow` (plain `setTimeout` inside
`agentSession.ts`) and Promise settlement, not on `RetryManager`'s clock
— not convertible through a stream-manager double; skipped.
- **Red check** on the converted scope-interrupt probe: with the
`forkIn(resourceScope)` branch replaced by a plain `runFork`, both the
original 720 ms version and the TestClock version still pass — the
stream-end path also calls `interruptPartialWriteFiber` directly, so the
scope close is a second guard. Same discriminating power before and
after (recorded, not changed: the test's purpose is "no late write",
which it does prove).

### The "SIGTERM ~6 s after startup exits ~11 s later" gap — root cause
and disposition

Per-step lines make it unambiguous. `xum server` (`node
dist/cli/index.js server`, temp `XUM_ROOT`, `XUM_LOG_LEVEL=debug`, under
`script -q -e -f`), 5 runs each:

| SIGTERM sent | JS teardown (`Shutting down server...` → `[shutdown]
exiting`) | SIGTERM → process gone | exit code |
|---|---|---|---|
| 6 s after `initialize completed` (≈ 8.4 s after spawn) | 64–67 ms
(`dispose` 61–64 ms) | **10.57 / 10.89 / 10.83 / 10.64 / 10.75 s** | 0
×5 |
| 30 s after `initialize completed` | 79–85 ms (`dispose` 76–82 ms) |
**171 / 161 / 183 / 150 / 168 ms** | 0 ×5 |

In every 6 s run the last JS-side line (`[shutdown] exiting { totalMs:
67 }`) is printed ≈ 70 ms after SIGTERM; the process then lingers ≈ 10.7
s **after `process.exit(0)`**. Nothing inside `dispose()` (every step
0–16 ms; `AppRuntime disposed` 6–16 ms), nothing in
`serverService.stopServer()` (1 ms — PR 2's guess was wrong; PR 5's
strace was right).

Cause: `workerPool.ts` creates the tokenizer `Worker` at import time (≈
1.9 s after spawn in `xum server`); the worker evaluates
`ai-tokenizer/encoding` (31 MB of encodings), which takes **≈ 18 s** on
this host (`require("ai-tokenizer/encoding")`: 17.9 / 18.1 / 18.1 s
standalone). `process.exit()` terminates worker threads via V8
`TerminateExecution`, which cannot interrupt a parse/compile in
progress, so the main thread joins the worker until its current module
finishes. A controlled repro (`new Worker(tokenizer.worker.js)` +
`process.exit` at *t*; unref'd, no other work):

| `process.exit` at | process gone at | wait |
|---|---|---|
| 1 s | 3.1 s | 2.1 s |
| 4 s | 5.0 s | 1.0 s |
| **7 s** | **17.3 s** | **10.3 s** |
| 10 s | 17.5 s | 7.4 s |
| 13 s | 17.6 s | 4.5 s |
| 16 s | 17.1 s | 1.0 s |

i.e. one ≈ 10 s uninterruptible window from ≈ 7 s to ≈ 17 s into the
worker's load (a single huge encoding module). SIGTERM 6 s after init
lands ≈ 6.5 s into that load → ≈ 10.7 s wait; at 30 s the worker is long
done → 150–180 ms. **Disposition:** not a leak inside `dispose()`'s
scope (the plan's fix criterion), pre-existing on `main` (PR 2/5 saw the
same numbers), and not DI-related → recorded as a follow-up, not fixed
here. Follow-up options: create the tokenizer worker lazily on first
`run()` (an idle `xum server`/ACP never pays), or split the worker's
encoding import per model (`encoding[model.encoding]` is already
selected per call) so the uninterruptible window is one encoding, not
all of them. The desktop is unaffected in practice (it imports the
tokenizer first and shuts down long after startup).

Two smaller observations from the transcripts (both pre-existing, both
left alone): (a) a one-time 25–45 ms gap right after `[shutdown]
AppFiberScope closed` is `source-map-support` (registered by
`cli/server.ts`) mapping Effect-internal frames the first time the log
helper captures a stack **inside a fiber** — verified standalone: first
in-fiber `new Error().stack` 47.8 ms with source maps vs 0.3 ms after /
0.3 ms without; not teardown work. (b) In the CLI roots' lists
`appFiberScope.close`/`appRuntime.dispose` are timed by
`runBestEffortCleanup` *and* log their own `… closed`/`… disposed` line
(`boundedTeardown`); kept uniform rather than special-casing two steps —
the outer line adds the Promise-settle time.

Whole-`dispose()` latch and both bounded teardowns: unchanged and
re-asserted (`serviceContainer.test.ts` "shares one teardown across
concurrent dispose() calls", `appRuntime.test.ts` timeout/never-rejects
cases, the PR 5 order test).

### Pre-review audits (plan §3 preamble)

1. **Interruption posture** — unchanged: no new fibers or forks in
product code; `shutdownStep` creates none. The TestClock suites fork the
same effects through a `TestClock`-bound runner.
2. **Uninterruptible teardown** — `boundedTeardown` untouched. In
`disposeOnce()` synchronous steps are timed without a Promise (no new
suspension point); async steps get one `.finally` microtask after an
await that already existed. Order asserted unchanged.
3. **No defect escapes** — `shutdownStep` rethrows after logging
(containment unchanged: `disposeOnce` propagates as before,
`runBestEffortCleanup` contains as before); `log.debug` cannot throw
(`safePipeLog` catches). Both bounded teardowns still never reject.
4. **Spy-seam check** — `rg 'spyOn\('
src/node/services/serviceContainer.test.ts tests/`: the same public
methods are spied (`desktopBridgeServer.stop`,
`desktopSessionManager.closeAll`, `browserBridgeServer.stop`,
`analyticsService.dispose`, `timelineService.flush`,
`telemetryService.shutdown`) and `shutdownStep` calls them on the
instance, so every spy intercepts (56/56 in the container + CLI suites).
No constructor arity changed; the converted tests use the existing
optional trailing runner params.
5. **Sync-start** — still pinned by the converted suites
(`isRetryPending` true and `partialWriteFiber` defined synchronously
after the scheduling call, before any `adjust`) and directly by
`di/effectRunner.test.ts`.
6. / 7. N/A (no constructor moved; `memoryConsolidationService` not in
the diff).

### Phase 11 completion state

- **The DI graph now owns:** every service in the process (5 stores →
`EffectRunner`/`AppFiberScope` → `MemoryMeta` → 8 cross-cutting → 19
core layers in 8 stages + `CoreWiringLive` → 6 desktop group layers +
`DesktopWiringLive`), built once per process by one `ManagedRuntime`
(`AppLive` for desktop/`xum server`/ACP/tests-ipc; `CoreRootLive` for
`xum run`/`xum workflow`); the oRPC `effect/context`; the two runtime
seams (`EffectRunner` in the three clock-driven workers and
`StreamManager`/`RetryManager`; `AppFiberScope` with its fixed dispose
slot and bounded close); startup/shutdown observability
(`[startup]`/`[shutdown]` lines). Product LoC for the whole phase
(pre-PR 1 → this branch, non-test): `di/` +2327 (tags 389, layers 1571,
runtime/seams/helper ≈ 370), composition roots +360/−851,
workers/stream/CLI/misc +237/−89 — net ≈ +1.98k, well above the plan's ≈
+420 estimate (the per-service tags and the layer adapters that restate
every constructor call are the bulk; each PR body recorded its actual
diff). Service classes untouched except optional trailing runner params.
- **Still imperative (explicit non-goals, D2/§8):**
`ServiceContainer.initialize()` (six awaited `initialize()`s + three
`start()`s — a future `runtime.runPromise(startupEffect)` with per-step
`Effect.timeout`); `streamBridge.ts` streams on the global runtime
(needs a runner/context parameter on `subscriptionIterable`, which would
also let `streamBridge.test.ts`'s 11 ticker waits move to a TestClock);
the hand-ordered `dispose()`/`shutdown()` steps (layer finalizers would
require proving reverse-construction order compatible — I5);
`AgentStatusService`'s ref'd `setInterval`; OAuth device-flow polling.
- **First `AppFiberScope` occupant (next phase):** the streamManager
engine core — fork the per-stream engine fiber into `AppFiberScope` so
`dispose()` step 2 interrupts and awaits in-flight streams while
`historyService`/`sessionUsage` are still alive; the position, bound
(`APP_FIBER_SCOPE_CLOSE_TIMEOUT_MS`), asymmetry tests and occupant rules
are in place, so that phase does not have to re-derive shutdown.

## Validation

- `make static-check` green (typecheck both projects, prettier, eslint,
docs).
- Converted/touched suites: `retryManager` 15/15, `streamManager`
123/123, `idleCompactionService` 20/20 + `testClock` 2/2,
`heartbeatService` 74/74 + `testClock` 2/2, `serviceContainer` +
`runCleanup` + `workflow` + `server` + `cli` 56/56, `di/*`.
- Full local gate on this host: `bun test src` **13 900 pass / 12 fail /
8 skip** (834 files, 1045 s) — the 12 are exactly the known host
baselines (taskGitPatchEngine ×2, gitNoHooksEnv ×3, WorkspaceTurnManager
×2, agent_skill_delete, BackupRepoCache, WorkspaceFooterBar load flake
×3), none in files this PR touches. `TEST_INTEGRATION=1 bun x jest
tests`: numbers appended when it finishes (provider-backed suites 403
from the AI bridge here; CI is authoritative). CI round 1: `Test /
Integration` failed only on
`tests/ipc/providers/anthropicCacheStrategy.test.ts` ("Expected cache
creation but got 0 tokens" — a live-provider cache-token assertion
unrelated to this diff).
- **Dogfooding** (headless Coder host): the two `xum server` SIGTERM
matrices above (exit 0 ×10, every `[shutdown]` line present in every
transcript); `xum workflow` echo run from the branch — `AppRuntime
built` → `ok from pr6` → `[shutdown]
backgroundProcessManager.beginShutdown` → `AppFiberScope closed` →
`session.dispose` → … → `terminateAll` → `AppRuntime disposed` (the
order `workflow.test.ts` pins), exit 0; **dev-server sandbox**
(`XUM_LOG_LEVEL=debug DEV_SERVER_SANDBOX_ARGS=--clean-projects make
dev-server-sandbox`): `AppRuntime built { ms: 21 }` → `initialize
completed`; via agent-browser loaded the app
(`v0.28.3-nightly.148-29-gac4fc78c2`), added a scratch git repo as a
project, created a worktree workspace, sent "Reply with exactly the
single word: pong" → `pong`, Stats tab populated (screenshot below);
then SIGTERM to the sandbox backend with the live workspace → full
`[shutdown]` sequence incl. `[analytics-worker] Shutting down, closing
DuckDB`, `ServiceContainer.dispose completed { totalMs: 82 }`, process
gone in 173 ms. Not exercisable here: Electron quit (no `DISPLAY`;
covered by `tests/e2e` in CI and the shared `dispose()` path above),
provider-backed integration suites (AI bridge 403s — CI lane).

## Risks

Low. Product behavior changes are limited to debug-level log lines and
the `xum workflow` cleanup list going through the same best-effort
runner as `xum run` (same containment; warn wording generalized). The
teardown order is unchanged and asserted; the timing helper adds no
suspension between synchronous steps. Test changes replace timer
surrogates with the runner seam that PRs 2/5 already made production
behavior, and each worker keeps one real-clock smoke.

![Sandbox smoke on this branch: pong reply in a worktree workspace,
Stats tab
populated](https://github.com/user-attachments/assets/2591febd-0ffc-45e6-b331-8ec7cec3a3da)

---

<details>
<summary>📋 Implementation Plan</summary>

# Effect migration — Wave 3 / Phase 11: ManagedRuntime + Layer
dependency injection

## 0. Summary

Replace the two hand-written composition roots (`createCoreServices` +
the `ServiceContainer` constructor) with an **Effect `Layer` graph**
built once per process by a **`ManagedRuntime`** ("AppRuntime"), while
keeping every service class, constructor signature, Promise facade,
private method, and test seam compatible. The runtime becomes (a) the
owner of the app-lifetime `Scope`, (b) the provider of
`"effect/context"` for oRPC Effect-native handlers, and (c) the source
of two runtime seams: an **`EffectRunner`** (context-bound,
*unsupervised* runner that lets clock-driven workers run on a
`TestClock`) and an **`AppFiberScope`** (a runtime-owned, *supervised*
scope whose close is awaited by `dispose()` — the slot the streamManager
engine core will occupy later).

Six stacked, independently mergeable PRs. Product PRs keep existing
tests unchanged; only the final test-modernization PR edits tests. Net
product LoC ≈ **+420** (per-PR estimates below). Service classes are
*not* rewritten — Layers are thin adapters around existing constructors;
cycle-breaking setter wiring moves into explicit "wiring layers" that
replay today's order.

Unlocks (not done here): streamManager ENGINE CORE conversion,
`TestClock` for timing suites, app-lifetime scopes.

## 1. Verified current state (evidence)

- **Roots.** `src/node/services/coreServices.ts:103-389`
(`createCoreServices`: 25 constructions, 12 `turnRequestBuilderBindings`
writes, ~14 setters) and `src/node/services/serviceContainer.ts:161-575`
(45 more constructions; `aiService.on(...)`/`workspaceService.on(...)`
analytics wiring at 474-574; global registrations
`setGlobalCoderService/setSshPromptService` at 469-471). `new
ServiceContainer(stores)` is called by `headlessEnvironment.ts:111`,
`tests/ipc/setup.ts`, `src/cli/server.ts:132`,
`src/node/acp/serverConnection.ts:155`, `src/desktop/main.ts:653`;
`src/cli/run.ts:661` and `src/cli/workflow.ts:376` call
`createCoreServices` directly. ⇒ two graph roots (App vs Core), five
process entry points, all constructing **synchronously**.
- **Startup.** `ServiceContainer.initialize()` (577-642) awaits six
`initialize()`s (no try/catch; failure propagates to `main.ts:1255-1265`
"Startup Failed" dialog + quit; `server.ts`/ACP log and exit), then sync
`start()`s idleCompaction/heartbeat/agentStatus, then two
fire-and-forget sweeps. All constructors are synchronous; two have side
effects on **declared constructor dependencies** only (`AIService` →
`streamManager.setEventSink`, `WorkspaceService` →
`backgroundProcessManager.on/aiService.on`).
- **Teardown.** `dispose()` (746-779) is explicit and hand-ordered
(`backgroundProcessManager.beginShutdown()` MUST be first — it is a
latch protecting persisted monitor records; bridges stop before sessions
close; `terminateAll` late; `timelineService.flush()` last).
`shutdown()` (718-732) is a *second* sequence fired concurrently by a
second `before-quit` listener (`main.ts:1321`). `main.ts:1296-1304`
races `dispose()` against 5 s then `app.quit()`; `cli/server.ts:227-268`
has a 5 s `process.exit(1)` force timer; `tests/ipc` cleanup calls
`dispose()` then `shutdown()`; `headlessEnvironment.dispose` never calls
`services.dispose()`.
- **Existing Effect surface.** 25 files import `effect`. Only
`Context.Service` tag: `MemoryMeta`
(`src/node/orpc/effectContext.ts:21`). `handlerGen`
(`@orpc/experimental-effect`) runs `Effect.runPromiseExit` per request
and `Effect.provide`s `opts.context["effect/context"]`.
`streamBridge.ts` runs streams on the global runtime. Scope-owning
workers: `heartbeatService.ts:134-243`,
`idleCompactionService.ts:86-122` (`Scope.makeUnsafe` +
`Effect.runSync(Scope.close(..))`, valid only because their fibers
suspend solely on the clock), `oauthFlowManager.ts:164`,
`streamManager.ts:4767/4054` (already `Effect.runFork(Scope.close(..))`
— the async-close precedent). `memoryConsolidationService.ts:667-703,
837-860`: check-and-reserve funnels with zero suspensions before
`inFlight.set`/`harvestInFlight.set`.
- **effect@4.0.0-rc.112 API (verified in `node_modules/effect/dist`).**
`Context.Service<Self, Shape>()("id")` (module `Context`, not
`ServiceMap`);
`Layer.{succeed,sync,effect,effectContext,effectDiscard,provide,provideMerge,mergeAll,build,buildWithScope}`
(no `Layer.scoped`; `Layer.effect` strips `Scope` from R);
`ManagedRuntime.make(layer)` → `{ runSync, runSyncExit, runFork,
runPromise, runPromiseExit, contextEffect, cachedContext, scope,
dispose(), disposeEffect }`;
`Effect.{runSyncWith,runForkWith,runPromiseWith,runPromiseExitWith}(context)`;
`Effect.context<R>()`; `Effect.serviceOption`;
`Scope.{fork,forkUnsafe,close,provide}`; `TestClock` from
`effect/testing` (`layer, adjust, setTime, withLive`); `Clock.Clock` is
a `Context.Reference` (defaulted; `TestClock.layer()` overrides it).
- **ManagedRuntime internals the design relies on**
(`ManagedRuntime.js`): `make` creates `scope =
Scope.makeUnsafe("parallel")` and `layerScope = Scope.forkUnsafe(scope,
"sequential")`; the first `runX` forks a build fiber over
`Layer.buildWithMemoMap` — a **fully synchronous layer graph builds
synchronously**, so `runtime.runSync(Effect.context())` succeeds and
sets `cachedContext`; afterwards every `runX` is
`Effect.run…With(cachedContext)` (no extra async boundary). Fibers
started through `runtime.runX` are registered in `scope` (`onFiberStart:
Fiber.runIn(scope)`). `dispose()` = `Scope.close(scope)` (interrupt
registered fibers in parallel → layer finalizers sequentially in
reverse), after which any `runtime.runX` dies with `"ManagedRuntime
disposed"`.
- **Layer composition semantics.** `Layer.mergeAll(A, B)` is *not* a
dependency resolver: B's requirements are not satisfied by A's outputs;
requirements bubble up. Dependencies are satisfied only via
`Layer.provide`/`provideMerge` chains. Siblings in `mergeAll` may build
concurrently.
- **Test seams that pin signatures** (Explore report): private-method
spies (`Config.saveConfig`,
`WorkspaceService.retireKernelWorkflowRunReferences/startStartupRecovery/createSession/updateAgentStatus`,
`MCPServerManager.startServers`,
`AgentPluginInstallService.reconcileJournals`, …); module-level export
spies (`agentStatusService.generateWorkspaceStatus`,
`sshConnectionPool.verifyHostKeyAgainstPolicyEffect`, …); direct
construction in tests (`Config` 44 files, `HistoryService` 22,
`MemoryMetaService` 11, `WorkspaceService` 7, `IdleDispatcher` 6,
`StreamManager` 4, `ServiceContainer` 3); partial-mock casts
(`InitStateManager` 193, `AIService` 158, `TaskService` 149,
`ORPCContext` 62). `effectBridge.test.ts:24-30` builds a partial
`ORPCContext` via `buildOrpcEffectContext` + `as unknown as
ORPCContext`.
- **Timing probes** (TestClock candidates): `heartbeatService.test.ts` 6
real sleeps, `idleCompactionService.test.ts` 2, `retryManager.test.ts` 3
`setSystemTime`, `streamManager.test.ts` 7 (partial-write debounce),
`streamBridge.test.ts` 11 (heartbeat ticker), OAuth device-flow suites
14 (non-goal).

## 2. Target architecture

### 2.1 Building blocks (all under `src/node/services/di/`; the *only*
directory allowed to import
`Layer`/`Context`/`ManagedRuntime`/`TestClock`)

| Module | Contents |
|---|---|
| `tags.ts` | One `Context.Service` tag per service class provided by
the graph. Type-only imports of service classes ⇒ no runtime import
cycles. Ids `"xum/<Name>"`. Naming: class name minus trailing `Service`
(`MemoryMeta`, `Workspace`, `History`); classes without that suffix or
colliding with an exported name get a `Tag` suffix (`ConfigTag`,
`StreamManagerTag`, `IdleDispatcherTag`). Exports the unions `CoreTags`
and `AppTags`. |
| `effectRunner.ts` | `interface EffectRunner { runSync<A,E>(e:
Effect<A,E,never>): A; runSyncExit; runFork; runPromise; runPromiseExit
}` — a **context-bound, unsupervised** runner whose methods accept only
effects with **no service requirements** (`R = never`; defaulted
references like `Clock` do not appear in `R`). That makes "not a service
locator" type-enforced: a fiber that needs services must take them as
explicit constructor dependencies and, if it must be awaited on
shutdown, fork into `AppFiberScope`. `defaultEffectRunner` = the global
`Effect.runX` (today's exact behavior). `effectRunnerFromContext(ctx)` =
`Effect.run…With(ctx)`. `EffectRunnerTag` + `EffectRunnerLive =
Layer.effect(EffectRunnerTag, Effect.map(Effect.context<never>(),
effectRunnerFromContext))`, placed at the **base** of the graph so the
captured context contains only refs (`Clock`, later `Logger`/`Random`)
plus stores. Fibers forked through it are owned by the worker's own
`Scope` (explicit `start/stop`), **not** by the ManagedRuntime;
`runtime.dispose()` does not interrupt them. Services import only this
file from `di/`. |
| `appFiberScope.ts` | `AppFiberScopeTag: Scope.Closeable`.
`AppFiberScopeLive = Layer.effect(AppFiberScopeTag,
Effect.gen(function*(){ const parent = yield* Effect.scope; return
yield* Scope.fork(parent, "parallel"); }))` — a child of the runtime's
layer scope. Fibers forked into it via `Effect.forkIn(_, appFiberScope)`
are interrupted **and awaited** when the scope closes. This is the
**supervised** seam for I/O-suspended fibers (engine core, later).
`ServiceContainer.dispose()` closes it explicitly and early (§5) so
interrupted fibers can still use their dependencies during finalization;
`runtime.dispose()` later re-closes it idempotently as a backstop. No
production occupant in Phase 11; the seam exists with tests. |
| `appRuntime.ts` | `makeAppRuntime(layer)`:
`ManagedRuntime.make(layer)` + **eager synchronous build**
(`runtime.runSync(Effect.context<R>())`; `assert(runtime.cachedContext
!== undefined)`); a layer body that suspends is a programming error and
throws here — exactly where a throwing constructor throws today, so
every entry point's existing catch/dialog/log path is preserved.
`disposeAppRuntime(runtime, timeoutMs)` and `closeScopeBounded(scope,
timeoutMs)` share one shape: `Effect.uninterruptible` teardown shell
around `Effect.interruptible(target.pipe(Effect.timeout(timeoutMs)))`
where `target` is `runtime.disposeEffect` resp. `Scope.close(scope,
Exit.void)` (never a non-cancellable JS Promise wrapper);
`Effect.catchTag("TimeoutError", …)` + `Effect.catchDefect` →
`log.warn`; run via `Effect.runPromise`; **never rejects**; idempotent
(`Scope.close` is idempotent; `disposeEffect` is guarded by a latch).
Verify the exact rc `Effect.timeout` error type at implementation time
(rc.112: fails with `Cause.TimeoutError`, `_tag: "TimeoutError"`).
Module doc comment = the DI contract (§2.3, §5). |
| `layers/stores.ts` | `StoresLive(stores: ConfigStores)` =
`Layer.mergeAll` of `Layer.succeed` for `ConfigTag`,
`SessionLocatorTag`, `ProvidersConfigStoreTag`, `SecretsStoreTag`,
`FileLeaseManagerTag` (true siblings — no inter-dependencies).
`StoresFromCoreOptionsLive` reproduces the `opts.x ?? new
X(config.rootDir)` defaults of `coreServices.ts:106-112` for the CLI
root. |
| `layers/core.ts` | `CoreOptionsTag` (today's `CoreServicesOptions`
minus stores — carries the *optional* cross-cutting services exactly as
today). **PR 3:** `CoreProjectionLive = Layer.effectContext(...)`
wrapping the existing `createCoreServices` body and returning a
`Context<CoreTags>` (coarse projection, zero behavior change). **PR 4:**
peel into per-service `Layer.effect(Tag, Effect.gen(...))` layers
composed in **explicit dependency stages** (`Layer.provideMerge` between
stages; `Layer.mergeAll` only for true siblings within a stage — every
sibling claim below was checked against the constructor argument lists
in `coreServices.ts` and must be re-checked in the PR): S1 History ·
InitState · Provider · BackgroundProcess · ExtensionMetadata ·
MemoryMeta · TerminalAttention · IdleDispatcher ·
WorkspaceMcpOverrides(default) · `TurnRequestBuilderBindingsTag`
(`Layer.succeed(_, {})`) → S2a SessionUsage · Goal · Memory → S2b
StreamManager (needs SessionUsage) → S3 AIService → S4 Consolidation ·
MCPConfig → S5 MCPServerManager → S6 Workspace → S7 Task → S8
TurnManager → `CoreWiringLive` (`Layer.effectDiscard`, **`Effect.sync`
only — no `acquireRelease`**, replays `coreServices.ts:137-166, 209-210,
258-270, 288-325, 349-352, 360-367` in order). |
| `layers/desktop.ts` | `CrossCuttingLive` (policy, telemetry,
experiments, backup, sessionTiming, analytics, devTools,
workspaceMcpOverrides, browserBridgeTokenManager),
`CoreOptionsFromDesktopLive` (derives `CoreOptionsTag` from those tags +
`extensionMetadataPath`), then **group layers** (`Layer.effectContext`
returning a `Context` of several tags, constructed in today's order):
`BrowserLive`, `DesktopBridgeLive`, `OauthLive`, `WorkersLive`
(idleCompaction, heartbeat, agentStatus, timeline, refine),
`TerminalEditorLive`, `MiscDesktopLive`; staged with `provideMerge`
where one group needs another. `DesktopWiringLive` (`Effect.sync` only)
= setters +
`aiService.on/workspaceService.on/memoryConsolidationService.on` wiring
+ global registrations. |
| `layers/app.ts` | `AppLive(stores) = DesktopLive ▹ CoreLive ▹
CoreOptionsFromDesktopLive ▹ CrossCuttingLive ▹ AppFiberScopeLive ▹
EffectRunnerLive ▹ StoresLive(stores)` — read `X ▹ Y` as "X is *provided
with* Y, and both stay exposed", i.e.
**`X.pipe(Layer.provideMerge(Y))`** (rc.112 signature:
`provideMerge(that: provider)(self: consumer)`; the *right-hand* operand
is the dependency). Every `▹` keeps all tags visible in the final
`Context<AppTags>`. |
| `testEffectRunner.ts` (test helper, sibling of
`testHistoryService.ts`) | `makeTestEffectRunner()` → `{ runner,
adjust(duration), setTime(ms), dispose }` over one memoised
`ManagedRuntime.make(EffectRunnerLive.pipe(Layer.provideMerge(TestClock.layer())))`
(the TestClock is the *provider*; the runner captures it), so the worker
under test and `TestClock.adjust` share one `TestClock`. |

### 2.2 Composition roots after Phase 11

```mermaid
flowchart TB
  Stores["StoresLive(stores)<br/>Config · SessionLocator · ProvidersConfigStore · SecretsStore · FileLeaseManager"]
  Runner["EffectRunnerLive (unsupervised, ref-bound)<br/>+ AppFiberScopeLive (supervised, closed on dispose)"]
  Cross["CrossCuttingLive (desktop only)<br/>Policy · Telemetry · Experiments · Analytics · SessionTiming · DevTools · WorkspaceMcpOverrides · Backup"]
  Opts["CoreOptionsTag<br/>desktop: derived from CrossCutting · CLI: Layer.succeed(opts)"]
  Core["CoreLive<br/>PR 3: coarse CoreProjectionLive → PR 4: stages S1…S8 + CoreWiringLive"]
  Desk["DesktopLive — group Layers<br/>Browser · DesktopBridge · OAuth · Workers · TerminalEditor · Misc → DesktopWiringLive"]
  RT["AppRuntime = ManagedRuntime.make(AppLive)<br/>eager sync build · Context<AppTags> = oRPC effect/context · dispose() last"]
  Stores --> Runner --> Cross --> Opts --> Core --> Desk --> RT
  CLI["CLI root (xum run / xum workflow)<br/>createCoreServices(opts) = makeAppRuntime(CoreLive ▹ StoresFromCoreOptionsLive ▹ AppFiberScopeLive ▹ EffectRunnerLive ▹ succeed(CoreOptionsTag, opts))"]
  Core -.same Layer definitions.-> CLI
```

`ServiceContainer` keeps its public fields and the synchronous `new
ServiceContainer(stores)`: the constructor calls
`makeAppRuntime(AppLive(stores))`, stores `this.serviceContext =
runtime.runSync(Effect.context<AppTags>())`, and assigns fields via
`Context.get(this.serviceContext, Tag)`. `toORPCContext()` returns the
same plain fields plus `"effect/context": this.serviceContext`.
`initialize()` is untouched. `dispose()` follows §5.

`createCoreServices(opts)` keeps its signature and return shape plus
`runtime` and `appFiberScope` fields; `cli/run.ts:1574-1580` and
`cli/workflow.ts:275-320` cleanup lists gain
`closeScopeBounded(appFiberScope)` before `session.dispose()` and
`disposeAppRuntime(runtime)` as the final step (PR 3).

**Staged composition skeleton (PR 4 shape; direction matters):**

```ts
// Each stage depends only on stages defined above it. `provideMerge` keeps both sides exposed.
const S1 = Layer.mergeAll(HistoryLive, InitStateLive, ProviderLive, /* … true siblings only */);
const S2a = Layer.mergeAll(SessionUsageLive, GoalLive, MemoryLive).pipe(Layer.provideMerge(S1));
const S2b = StreamManagerLive.pipe(Layer.provideMerge(S2a));          // StreamManager needs SessionUsage
const S3 = AIServiceLive.pipe(Layer.provideMerge(S2b));
// … S4 … S8 likewise …
export const CoreLive = CoreWiringLive.pipe(Layer.provideMerge(S8));  // wiring runs after every service exists
```

**oRPC typing.** `OrpcEffectServices` (in `effectContext.ts`) becomes
`AppTags`, so `ORPCContext["effect/context"]: Context<AppTags>` is
satisfied by the runtime context in production. `buildOrpcEffectContext`
stays as the narrow test helper it already is (its only caller,
`effectBridge.test.ts:24-30`, deliberately builds a partial context and
casts it via `unknown`); no production caller remains after PR 1.

### 2.3 Invariants (the "DI contract"; enforced by tests and the
`appRuntime.ts` doc comment)

| # | Invariant | Constraint served |
|---|---|---|
| I1 | **Phase 11 compatibility contract, not permanent law:** layer
bodies are synchronous (`Layer.succeed`/`Layer.sync`/`Layer.effect` over
sync effects; `acquireRelease` with a sync acquire is fine).
`makeAppRuntime` asserts the eager build completed. Future async
resource acquisition belongs in `initialize()`/startup effects or an
explicit async factory root (`ServiceContainer.create()`), never
silently inside a layer. | #2 sync-start, #5 startup parity |
| I2 | Services never hold the `ManagedRuntime`. Workers hold an
`EffectRunner` (default `defaultEffectRunner`); `EffectRunner.runX` ≡
`Effect.run…With(ctx)` — same sync-start semantics as `Effect.runX`, and
still valid after `runtime.dispose()`, so late callbacks cannot hit
"ManagedRuntime disposed". Supervision, when needed, is explicit via
`AppFiberScope`. | #2, #3 |
| I3 | Per-call pipelines (`Effect.runPromise(this.effects…)` facades)
and the `memoryConsolidationService` funnels are untouched. **Audit
item:** no DI lookup, runner call, or `await` may be inserted before
`inFlight.set` / `harvestInFlight.set`. Only lifecycle forks in workers
move to `this.runner.runX`. | #1, #2 |
| I4 | Constructors, facades, private methods, module exports unchanged;
new constructor parameters are optional, trailing, defaulting to
`defaultEffectRunner`. | #1, #6 |
| I5 | Teardown order stays explicit in `dispose()`/`shutdown()`. Layer
bodies and wiring layers register **no finalizers** in Phase 11
(`Effect.sync` only), so `runtime.dispose()` reorders nothing. The one
supervised resource (`AppFiberScope`) is closed explicitly at a fixed
position in `dispose()` (§5). | #3 |
| I6 | Wiring layers replay today's setter/listener order; a constructor
may touch only its *declared* dependencies (built earlier by staging).
Per-PR audit: grep each moved constructor for calls on setter-provided
collaborators → forbidden. Dependency order is expressed only with
`provide`/`provideMerge` stages; never rely on `mergeAll` sibling order.
| #6 |
| I7 | No persisted-data changes; DI is in-process only. | #4 |
| I8 | Every process root builds from the same Layer definitions
(`CoreLive` shared by App and CLI). Unit harnesses
(`createTestHistoryService`, `createTestToolConfig`,
`createAgentSessionHarness`, …) intentionally bypass Layers. | #7 |

### 2.4 Decisions and alternatives (product-LoC deltas)

<details>
<summary>D1 — Granularity: coarse core first (PR 3), per-service core
stages behind a decision gate (PR 4), group layers for the desktop tail
(PR 5)</summary>

Honest framing: the three unlocks (engine-core async scope, TestClock,
app-lifetime scope) are delivered by `AppRuntime` + `EffectRunner` +
`AppFiberScope` and **do not require per-service layers**. Per-service
core layers are *migration leverage*: typed requirement sets for the
engine-core work, per-service swap in integration tests, explicit
dependency stages instead of implicit ordering.

- **(A) Per-service everywhere** (~70 layers): +~900/−~700. Desktop tail
has hand-tuned teardown that must not become finalizers, so per-service
there buys uniformity only. Rejected.
- **(B) Recommended:** PR 3 coarse `CoreProjectionLive` (+~120/−~10)
delivers the shared root and runtime ownership; PR 4 peels the core into
staged per-service layers (+~330/−~290) **only if** PR 3's
typecheck/startup budgets hold (gate in §3); desktop tail as ~6 group
layers (+~170/−~150). Tags for all services either way (~3 LoC each).
- **(C) Coarse only:** stop after PR 3 + desktop projection (~+200
total). Cheapest; the engine-core phase would then redo dependency
declarations. Remains the fallback if PR 4's gate fails.
</details>

<details>
<summary>D2 — Async init stays an explicit `initialize()`; Layers
construct only</summary>

Folding `initialize()` into layer construction would make the build
asynchronous (breaks I1), change failure semantics (today: fail-fast →
dialog/log), and move the six-step order into memoised builds. Deferred;
a later phase can turn `initialize()` into
`runtime.runPromise(startupEffect)` with per-step `Effect.timeout`.
</details>

<details>
<summary>D3 — Optional cross-cutting services stay optional via
`CoreOptionsTag`, not `Effect.serviceOption`</summary>

Core layer bodies read `opts.policyService` etc. exactly as today, so
CLI (absent) vs desktop (present) behavior is unchanged and no service
gains a new `undefined` branch.
</details>

<details>
<summary>D4 — Two seams instead of one: `EffectRunner` (unsupervised,
clock-bound) + `AppFiberScope` (supervised)</summary>

A single "runtime handle" conflates two needs. Workers need *which
clock* (TestClock) and must keep sync `stop()`; the engine core needs
*who awaits me on shutdown*. Explicit `Clock` injection per worker was
rejected (a `provideService(Clock.Clock, …)` at every fork site, and it
does not extend to other refs).
</details>

<details>
<summary>D5 — oRPC: `effect/context` = the runtime's `Context`;
`handlerGen` unchanged</summary>

`handlerGen` already `Effect.provide`s the context per request;
providing ~70 entries instead of one is one Map merge per request. The
existing `echoAsync`/`echoEffect` probes record the delta as a
**diagnostic** in the PR body (no stable benchmark harness exists to
make it a hard gate). `effect/wrap` not needed.
</details>

## 3. Phasing — six stacked PRs

Every PR: `make static-check`; gate suites below; existing tests
unchanged (PR 6 is the only PR that edits tests, and only to replace
real-timer probes). Before `@codex review`, run the **house pre-review
audits**:

1. **Interruption posture** — list every new/moved fiber fork; state
what interrupts it and when (unsupervised via `EffectRunner` + worker
scope, or supervised via `AppFiberScope`).
2. **Uninterruptible teardown** — teardown effects are
`Effect.uninterruptible` end-to-end; bounded waits inside use
`Effect.interruptible(Effect.timeout(...))` (house shape from #4038).
3. **No defect escapes** — `disposeAppRuntime`/`closeScopeBounded` and
every Promise facade fold defects; `makeAppRuntime` is the one place
allowed to throw (constructor semantics).
4. **Spy-seam check** — `rg 'spyOn\('
src/node/services/<touched>.test.ts tests/` per touched class;
constructor arity and private-method Promise signatures unchanged
(typecheck of tests proves it).
5. **Sync-start check** — a fork through `EffectRunner` runs to its
first `sleep` before `runFork` returns (mirrors
`heartbeatService.ts:199-202`).
6. **Constructor side-effect audit (I6)** for every constructor moved
into a Layer in that PR.
7. **Zero-suspension audit (I3)** whenever `memoryConsolidationService`
is in the diff.

### PR 1 — Skeleton: AppRuntime + Stores/MemoryMeta layers +
runtime-backed `effect/context` + dispose hook (+~150 LoC)

**Scope**
- `di/tags.ts` (`ConfigTag`, `SessionLocatorTag`,
`ProvidersConfigStoreTag`, `SecretsStoreTag`, `FileLeaseManagerTag`,
`MemoryMeta` moved from `orpc/effectContext.ts`, which re-exports it;
`AppTags` union).
- `di/layers/stores.ts` (`StoresLive`), `di/layers/core.ts` with
`MemoryMetaLive = Layer.effect(MemoryMeta, Effect.map(ConfigTag, c =>
new MemoryMetaService(c.rootDir)))`, `di/layers/app.ts`
(`AppLive(stores) = MemoryMetaLive ▹ StoresLive`).
- `di/appRuntime.ts` (`makeAppRuntime`, `disposeAppRuntime`);
`APP_RUNTIME_DISPOSE_TIMEOUT_MS` in `src/constants/`.
- `coreServices.ts`: `CoreServicesOptions.memoryMetaService?`
(precedent: `workspaceMcpOverridesService?`).
- `serviceContainer.ts`: build runtime first, pass `Context.get(ctx,
MemoryMeta)` to `createCoreServices`, `public readonly runtime`,
`toORPCContext()["effect/context"] = this.serviceContext`, `dispose()`
appends `disposeAppRuntime` behind a `disposed` latch; new
`log.debug("[startup] AppRuntime built", { ms })`.
- `orpc/effectContext.ts`: `OrpcEffectServices = AppTags`;
`buildOrpcEffectContext` retyped/test-helper doc.
- `headlessEnvironment.dispose` calls `await services.dispose()` before
removing the temp dir (the bench harness currently leaks the container;
runtime ownership starts here).

**Acceptance**
- `di/appRuntime.test.ts`: (a) sync build sets `cachedContext`; (b) a
layer with an async body makes `makeAppRuntime` **throw synchronously**
(I1 enforced); (c) probe layers' finalizers run in reverse order on
dispose; (d) dispose is idempotent and bounded (hung finalizer → `warn`,
resolves at the timeout); (e) `runtime.runFork` after the eager build
starts synchronously.
- `serviceContainer.test.ts`:
`Context.get(toORPCContext()["effect/context"], MemoryMeta) ===
services.memoryMetaService`; `dispose()` closes the runtime; `dispose();
shutdown()` (tests/ipc order) is clean; a throwing layer surfaces as a
synchronous throw from `new ServiceContainer(stores)` (same shape as
today's constructor throw → existing entry-point catch paths).
- `effectBridge.test.ts`, `memoryMeta*.test.ts` unchanged and green;
echo-probe overhead recorded in the PR body.
- Gate: `bun test src/node/services/di
src/node/services/serviceContainer.test.ts src/node/orpc
src/node/services/memoryMeta*` · `make test-integration` · `make
static-check`.

**Rollback:** `git revert`; classes untouched.

### PR 2 — Runtime seams: `EffectRunner` + `AppFiberScope`; TestClock on
idleCompaction/heartbeat/retryManager (+~140 LoC)

**Scope**
- `di/effectRunner.ts`, `di/appFiberScope.ts`; `AppLive` gains
`AppFiberScopeLive ▹ EffectRunnerLive` at the base; `ServiceContainer`
exposes `appFiberScope` (used only by `dispose()` in Phase 11) and
closes it per §5.
- `IdleCompactionService`, `HeartbeatService`, `RetryManager`: trailing
optional `runner: EffectRunner = defaultEffectRunner`; every lifecycle
`Effect.runSync/runFork` in `start/stop/schedule/cancel` becomes
`this.runner.runX`. Deadline math (`Date.now()`/injected `now`)
unchanged. `ServiceContainer` passes `Context.get(ctx, EffectRunnerTag)`
to the two workers; `RetryManager` keeps the default until PR 5 (so
`streamManager.ts` is untouched here).
- `di/testEffectRunner.ts` helper.

**Acceptance**
- New TestClock tests (existing real-timer tests untouched — they
exercise the `defaultEffectRunner` path, which is production behavior
wherever no runner is injected): heartbeat `STARTUP_DELAY_MS` → first
tick after `adjust`, one tick per `CHECK_INTERVAL_MS`, no ticks after
`stop()`; idleCompaction initial delay + cadence; retryManager fires
exactly at `delayMs`, `cancel()` before `adjust` never fires.
- Pin runtime facts: `runner.runSync(Scope.close(scope, Exit.void))`
completes synchronously for a fiber suspended on a TestClock sleep;
`runFork` through the runner reaches its first sleep synchronously;
`Effect.context<never>()` inside `EffectRunnerLive` sees the upstream
`TestClock` (else the helper provides `Clock.Clock` explicitly — same
seam, one line).
- `AppFiberScope` contract tests: (i) an **I/O-suspended** fiber
(interruptible `Effect.async` that never resolves, with a cancel path)
forked with `Effect.forkIn(_, appFiberScope)` is interrupted **and
awaited** by `closeScopeBounded(appFiberScope)` — and this happens
*before* the explicit teardown steps in `dispose()` (assert ordering
against a spy on `desktopBridgeServer.stop`); (ii) a fiber forked via
`EffectRunner` is *not* interrupted by either close (documents the
asymmetry); (iii) `disposeAppRuntime` afterwards idempotently re-closes
the already-closed child scope (no error, no second finalizer run).
- If `TestClock.adjust` leaves continuations pending, the helper adds
`Effect.yieldNow`/`Fiber.await` — decided by tests.
- Gate: `heartbeatService.test.ts`, `idleCompactionService.test.ts`,
`retryManager.test.ts`, `serviceContainer.test.ts`, `di/*`, tests/ipc.

**Rollback:** revert restores defaults; no call site depends on the new
params.

### PR 3 — Shared core root: coarse `CoreProjectionLive` +
`createCoreServices` facade + CLI runtime disposal (+~120 / −~10)

**Scope**
- Tags for the remaining 19 core services; `CoreOptionsTag`;
`StoresFromCoreOptionsLive`.
- `CoreProjectionLive = Layer.effectContext(Effect.gen(function*(){
const opts = yield* CoreOptionsTag; const stores = yield* …; const core
= buildCoreGraph({ ...opts, ...stores }); return Context.make(History,
core.historyService).pipe(Context.add(...)) }))` where `buildCoreGraph`
is today's `createCoreServices` body, unchanged, renamed.
- `createCoreServices(opts)` = `makeAppRuntime(CoreProjectionLive ▹
StoresFromCoreOptionsLive ▹ AppFiberScopeLive ▹ EffectRunnerLive ▹
Layer.succeed(CoreOptionsTag, opts))`, returns today's `CoreServices`
object read from the context plus `runtime` and `appFiberScope`.
`cli/run.ts` and `cli/workflow.ts` cleanup lists append
`closeScopeBounded(appFiberScope)` **before** `session.dispose()` and
`disposeAppRuntime(runtime)` **after**
`backgroundProcessManager.terminateAll()`.
- `ServiceContainer` stops calling `createCoreServices`; `AppLive =
CoreProjectionLive ▹ CoreOptionsFromDesktopLive ▹ CrossCuttingLive ▹ …`
(cross-cutting services move into `CrossCuttingLive` now because core
options derive from them). Desktop constructions otherwise stay in the
constructor.

**Acceptance**
- Identity test: every `CoreServices` field `===` `Context.get(ctx,
Tag)`; `serviceContainer.test.ts` unchanged and green.
- **Decision gate for PR 4** recorded in the PR body: `make typecheck`
wall time, `[startup] AppRuntime built` ms and `initialize` totals vs
`origin/main` baseline from the sandbox (§7). Proceed to PR 4 only if
typecheck regresses < 10 % and startup within noise; otherwise stop at
(C).
- Gate: `bun test src/node/services`, `src/cli/*.test.ts`
(run/workflow/server/cli), tests/ipc, `make static-check`.

**Rollback:** revert restores the imperative call; PR 1/2 unaffected.

### PR 4 — Peel the core into staged per-service Layers +
`CoreWiringLive` (+~330 / −~290 ⇒ net ≈ +40; split 4a/4b if > ~600 diff
lines)

**Scope**
- Stages S1, S2a, S2b, S3…S8 (§2.1 + skeleton in §2.2) as `Layer.effect`
adapters with today's argument lists; `CoreWiringLive` (`Effect.sync`
only) replays the wiring lines in order; `CoreLive =
CoreWiringLive.pipe(Layer.provideMerge(S8))` replaces
`CoreProjectionLive`; `buildCoreGraph` deleted.
- Before writing any stage: re-derive the DAG from the constructor
argument lists (the plan's stage table was checked once; `StreamManager
→ SessionUsage` is the kind of edge that turns "siblings" into a stage
split) and record it in the PR body.
- 4a (S1–S3: leaves through `AIService`) / 4b (S4–S8 + wiring) if needed
— 4a alone is mergeable because the remaining services are built by a
shrunken projection layer that reads S1–S3 from the context.

**Acceptance**
- Wiring assertions that are behavioral (a missing wiring line fails
them): `turnRequestBuilderBindings` fully populated; goal continuation
consumer registered on `idleDispatcher`; `streamManager` MCP manager
set; registration probe installed on `extensionMetadata`.
- I6 audit table for all 19 constructors in the PR body;
missing-provider = compile error (R must be `never` at `makeAppRuntime`)
demonstrated by a type-level test (`// @ts-expect-error`).
- Gate: as PR 3 plus `streamManager*.test.ts`, `aiService.test.ts`,
`workspaceService*.test.ts`.

**Rollback:** revert to PR 3's projection.

### PR 5 — `DesktopLive` group layers + `DesktopWiringLive`; thin
`ServiceContainer`; `StreamManager` runner param (+~170 / −~150 ⇒ net ≈
+20)

**Scope**
- Tags for the 45 desktop services; six group layers
(`Layer.effectContext`, today's construction order inside each;
`provideMerge` between groups that depend on each other);
`DesktopWiringLive` (`Effect.sync` only) = `serviceContainer.ts:209,
263-265, 271, 288-290, 334-340, 348, 365, 375, 381-382, 434, 438-471,
474-574` in order.
- `ServiceContainer` constructor = `makeAppRuntime(AppLive(stores))` +
field assignment from the context. `toORPCContext()` unchanged in shape.
- `StreamManager`: optional trailing `runner: EffectRunner`;
`schedulePartialWrite` fork (`streamManager.ts:1141`) and `RetryManager`
construction use it; `Scope.close` stays `Effect.runFork` (existing
async-close precedent). `WorkersLive` receives `EffectRunnerTag`.

**Acceptance**
- All four existing `serviceContainer.test.ts` assertions unchanged; new
identity test over `toORPCContext()` fields vs tags;
`dispose()`/`shutdown()` call order asserted via spies on the *public*
methods already spied today.
- I6 audit for the 45 constructors.
- Gate: tests/ipc + tests/ui (`make test-integration`),
`src/cli/server.test.ts`, `src/cli/cli.test.ts`,
`streamManager*.test.ts`, `aiService.test.ts`.

### PR 6 — TestClock adoption sweep + shutdown hardening + contract docs
(+~20 LoC product; tests edited)

**Scope**
- Replace real-sleep cadence probes with `makeTestEffectRunner()` in
`heartbeatService.test.ts`, `idleCompactionService.test.ts`,
`retryManager.test.ts`, and the partial-write debounce cases of
`streamManager.test.ts`; keep **one real-timer smoke test per worker**
(guards the `defaultEffectRunner` path).
- `cli/server.ts`: `[shutdown]` log lines per step incl. `AppRuntime
disposed {ms}`; confirm the whole `dispose()` fits the existing 5 s
force-exit budget.
- Finalize the contract doc comment in `di/appRuntime.ts` (I1–I8, §5).

**Acceptance:** converted suites have zero `setTimeout`-based cadence
waits (grep in PR body), same assertions; `make test-integration` green;
sandbox startup/shutdown evidence (§7).

## 4. TestClock story

- **Mechanism.** `Effect.sleep`, `Schedule.fixed`, `Effect.timeout`,
`Clock.currentTimeMillis` read the `Clock` reference from the running
fiber's context. Workers that fork through an `EffectRunner` built under
`TestClock.layer()` run on the test clock; `await testRunner.adjust("2
minutes")` advances it. `Date.now()`, `setTimeout`, `setInterval` are
unaffected — heartbeat deadline math via injected `now`,
`AgentStatusService`'s ref'd `setInterval`, and
`backgroundProcessManager` stay on real timers/injected timestamps.
- **Benefit now:** `heartbeatService.test.ts` (6),
`idleCompactionService.test.ts` (2), `retryManager.test.ts` (3
`setSystemTime` → `adjust`; `Date.now`-based `retryAt` may move to
`Clock.currentTimeMillis` only if a test needs both clocks aligned),
`streamManager.test.ts` debounce cases (7).
- **Deferred:** `streamBridge.test.ts` ticker (11) — needs a
context/runner parameter on `subscriptionIterable`; OAuth device-flow
polling and `oauthFlowManager.test.ts` (25) — non-goal.
- **Stays real:** child-process/PTY/WASM/fs-lock waits
(`backgroundProcessManager` 72, `quickjsRuntime` 26, lock sleeps in
`workspaceService`/`taskService`), end-to-end suites (tests/ipc, e2e).
- **Pinned in PR 2, not assumed:** `adjust` runs due sleeps and their
synchronous continuations before resolving (or the helper yields until
they do); `Schedule.fixed` anchoring under `TestClock` matches the
wall-clock expectations in `heartbeatService.ts:149-155`; sync
`Scope.close` of a TestClock-suspended fiber completes synchronously.

## 5. Shutdown protocol

1. **Trigger points unchanged:** `main.ts` `before-quit` (preventDefault
→ `dispose()` raced with 5 s → `app.quit()`; update-install path
fire-and-forget), the second `before-quit` listener's `shutdown()`
(unchanged, concurrent), `cli/server.ts` SIGINT/SIGTERM (5 s force
exit), ACP `close()`, tests/ipc (`dispose()` then `shutdown()`),
headless bench (`dispose()` from PR 1).
2. **`ServiceContainer.dispose()` order:**
1. `backgroundProcessManager.beginShutdown()` — unchanged, first (latch
protecting persisted monitor records).
2. **`closeScopeBounded(appFiberScope,
APP_FIBER_SCOPE_CLOSE_TIMEOUT_MS)`** — interrupts and awaits supervised
fibers *while every dependency they might touch during finalization is
still alive*. No occupants in Phase 11; the position is fixed now so the
engine-core phase does not have to re-derive it.
3. The existing explicit sequence verbatim (`desktopBridgeServer.stop()`
… `terminateAll()` … `timelineService.flush()`).
4. **`disposeAppRuntime(runtime, APP_RUNTIME_DISPOSE_TIMEOUT_MS)`** —
closes the runtime scope (interrupts any fiber started via
`runtime.runX` — none long-lived in Phase 11; runs layer finalizers —
none in Phase 11 by I5). Hung → `warn` at the timeout; never rejects.
Budget: 2 s + 2 s inner bounds inside the callers' 5 s outer budgets;
the outer race in `main.ts` remains the last line of defense.
**Rule for future occupants:** anything forked into `AppFiberScope` must
tolerate interruption at any suspension point and must not depend on
resources torn down in step 1; anything that needs a Layer finalizer
must first prove reverse-construction order is compatible with steps 2–3
(I5).
3. **Latches:** `disposed` makes `dispose()` idempotent (two
`before-quit` listeners, tests/ipc dispose+shutdown). `shutdown()` never
touches the runtime or `AppFiberScope`.
4. **Late callers:** `EffectRunner` handles keep working after runtime
dispose (I2), so a stray `tick()`/`scheduleRetry()` after quit cannot
defect. The `ManagedRuntime` is referenced only by `ServiceContainer`
and the `createCoreServices` return value.
5. **Worker `stop()` stays synchronous** (`runner.runSync(Scope.close)`)
because their fibers suspend only on the clock. The engine core will
fork into `AppFiberScope` (step 2.2 awaits it) — the reason both seams
exist now.
6. **Crash paths:** unchanged — `uncaughtException`/SIGKILL run no
finalizers. Finalizers are best-effort; durable state must remain
crash-safe without them (AGENTS.md self-healing rule). Nothing in Phase
11 makes a finalizer the sole guardian of durable state.

## 6. Risk register

| # | Risk | L/I | Mitigation |
|---|---|---|---|
| R1 | A layer body suspends → `runSync` throws at startup | M/H | I1
assert + PR 1 test (b); doc comment; review checklist; entry-point catch
paths verified in PR 1 |
| R2 | Construction-order side effects differ under staged builds | L/H
| I6 audit per moved constructor; explicit `provideMerge` stages; wiring
layers replay today's order; tests/ipc as behavioral gate |
| R3 | Double teardown (`shutdown()` ∥ `dispose()`; dispose+shutdown in
tests) | M/M | `disposed` latch; runtime/AppFiberScope closed only in
`dispose()`; PR 1 test |
| R4 | Late `runtime.runX` after dispose → defect | M/M | I2: services
hold `EffectRunner`, never the ManagedRuntime |
| R5 | TestClock semantics differ from assumptions | M/L | PR 2 pins
them before any suite converts; per-suite fallback to real timers |
| R6 | effect v4 RC churn (`Context`→`ServiceMap`, Layer renames) | M/M
| All `Layer/Context/ManagedRuntime/TestClock` imports confined to
`di/`; exact pin |
| R7 | Startup latency regression (splash) | L/M | `AppRuntime built` ms
+ `initialize` totals vs baseline in sandbox; PR 3 gate |
| R8 | Typecheck slowdown from large requirement unions | L/L | PR 3
gate records `make typecheck` wall time; fallback (C) |
| R9 | Per-request `Effect.provide` of a ~70-entry Context | L/L |
echo-probe diagnostic in PR 1/5 bodies |
| R10 | Spy seams / direct-construction tests break | L/H | I4; optional
trailing params; audit 4; typecheck of tests |
| R11 | CLI roots forget to dispose runtime/scope | M/L | PR 3 wires
both cleanups; `src/cli/*.test.ts` assert the cleanup steps exist |
| R12 | Someone forks long-lived I/O work via `EffectRunner` expecting
dispose to await it | M/M | Doc on `EffectRunner` ("unsupervised"); PR 2
asymmetry test; review audit 1 |

**Rollback:** PRs are stacked; revert in reverse order (6→1). Service
classes are never modified except for optional trailing params, so any
revert restores the previous composition root wholesale with no data or
API implications.

## 7. Dogfooding (per PR; evidence attached to the PR body)

**Environment (headless Coder host, no `DISPLAY`):**
```bash
XUM_LOG_LEVEL=debug DEV_SERVER_SANDBOX_ARGS="--clean-projects" make dev-server-sandbox   # background bash task; prints URL + XUM_ROOT
```
- **Startup correctness:** `<XUM_ROOT>/logs/*.log` shows, in order:
`Loading services...`, `[startup] AppRuntime built {ms}`, `[startup]
ServiceContainer.initialize starting`, six step durations, `[startup]
ServiceContainer.initialize completed {totalMs, stepDurationsMs}`. Paste
baseline (`origin/main`) vs branch numbers.
- **Startup-never-crash parity (once, locally, not committed):** inject
a throwing scratch layer → `xum server` exits non-zero with the existing
logged error and **no** unhandled-rejection trace; for desktop, confirm
by code path (`loadServices()` rejects → `main.ts:1255` dialog) and via
`src/cli/server.test.ts`/ACP tests.
- **UI smoke (agent-browser):** `open <url>` → `snapshot -i` → add a
scratch git repo as a project → create a workspace → send one message →
`screenshot` the loaded app and the response; `attach_file` both.
**Video:** start `agent-browser record` before the flow and stop it with
a hard timeout (`timeout 30 agent-browser record stop`); if stopping
hangs (known), attach the truncated WebM plus the screenshots and say
so.
- **oRPC Effect path:** pin/unpin a memory entry (rides `handlerGen` +
runtime `effect/context`); screenshot before/after; grep logs for
`ManagedRuntime disposed`/defect lines (expect none).
- **Graceful quit:** record the terminal with `script -q
/tmp/<workspace>-shutdown.log` (or `agent-tty` if present), `kill -TERM
<pid>` → expect `[shutdown]` lines, `AppRuntime disposed {ms}`, exit 0,
no force-exit message; attach the typescript. Exercise the timeout
branch once with a scratch hung finalizer → `warn` + timely exit.
- **Electron (best effort):** with `Xvfb`, `make dev` + agent-browser
via CDP (electron skill): screenshot splash → main window, quit via
menu, confirm exit < 5 s; otherwise state that the Electron path is
covered by `tests/e2e` in CI and the shared `dispose()` path exercised
by `server.ts`.

**Gate suites per PR** (plus `make static-check` always):

| PR | Must pass |
|---|---|
| 1 | `src/node/services/di/*`, `serviceContainer.test.ts`,
`src/node/orpc/*`, `memoryMeta*`, `make test-integration` |
| 2 | + `heartbeatService.test.ts`, `idleCompactionService.test.ts`,
`retryManager.test.ts` |
| 3 | + `bun test src/node/services`, `src/cli/*.test.ts`; record PR 4
gate numbers |
| 4 | + `streamManager*.test.ts`, `aiService.test.ts`,
`workspaceService*.test.ts` |
| 5 | + tests/ui via `make test-integration`, `src/cli/server.test.ts`,
`src/cli/cli.test.ts` |
| 6 | converted suites + full `make test-integration` + sandbox
startup/shutdown evidence |

## 8. Non-goals (explicit)

- streamManager ENGINE CORE conversion (first `AppFiberScope` occupant;
separate phase).
- `Schema` at persistence boundaries; OAuth refresh/device-flow workers;
`AgentStatusService` `setInterval` → Effect.
- `initialize()` as a Layer/startup effect (D2); per-service optional
tags (D3); `streamBridge` on the runtime; layer finalizers for existing
`dispose()` steps.
- Any change to persisted data, IPC wire shapes, or oRPC handler bodies
beyond the `effect/context` source.

## 9. Assumptions stated

- `Effect.context<never>()` inside `EffectRunnerLive` returns the
enclosing build context including an upstream `TestClock` entry (PR 2
test; fallback: provide `Clock.Clock` explicitly in the helper).
- `Scope.fork(parent)` inside a `Layer.effect` body yields a child
closed by the runtime's layer scope on `dispose()` (PR 2 `AppFiberScope`
test).
- Layer bodies never need to observe sibling construction order; all
ordering that matters is expressed as `provide`/`provideMerge` stages or
wiring-layer statement order.
- `EffectRunner`'s `R = never` constraint is sufficient for every
lifecycle fork in the three Phase 11 workers and
`StreamManager.schedulePartialWrite` (they only use
`Effect.sleep`/`Schedule`/`Effect.sync`/`Effect.tryPromise` — no service
tags). Verified by typecheck in PR 2/5.
- The desktop tail's teardown remains explicit unless a later RFC proves
reverse-construction order compatible; this plan does not attempt it.

</details>

---

_Generated with `xum` • Model: `anthropic:claude-fable-5-1` • Thinking:
`xhigh`_

<!-- mux-attribution: model=anthropic:claude-fable-5-1 thinking=xhigh
-->
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