fix(gateway): defer embedded runner imports

CHANGELOG.md

@@ -50,6 +50,7 @@ Docs: https://docs.openclaw.ai
 - Gateway/startup: lazy-load plugin HTTP route dispatch when active plugin routes exist so no-plugin Gateway boot skips plugin route runtime scope setup. Thanks @vincentkoc.
 - Gateway/startup: move chat run/subscriber registries onto a lightweight state module and defer chat/session event projection until the first event so Gateway boot skips session IO imports. Thanks @vincentkoc.
 - Gateway/startup: keep node session runtime on a lightweight JSON parser instead of importing gateway method validation helpers during boot. Thanks @vincentkoc.
+- Gateway/startup: read embedded-run activity from a lightweight shared state module so restart deferral no longer imports the embedded runner during Gateway boot. Thanks @vincentkoc.
 - CLI/Gateway: use a parse-only config snapshot for plain `gateway status` reads and reuse same-path service config context so status no longer spends tens of seconds in full config validation before printing. Thanks @vincentkoc.
 - Lobster/Gateway: memoize repeated Ajv schema compilation before loading the embedded Lobster runtime so scheduled workflows and `llm.invoke` loops stop growing gateway heap on content-identical schemas. Fixes #71148. Thanks @cmi525, @vsolaz, and @vincentkoc.
 - Codex harness: normalize cached input tokens before session/context accounting so prompt cache reads are not double-counted in `/status`, `session_status`, or persisted `sessionEntry.totalTokens`. Fixes #69298. Thanks @richardmqq.

src/agents/pi-embedded-runner/run-state.ts

@@ -0,0 +1,68 @@
+import {
+  getActiveReplyRunCount,
+  listActiveReplyRunSessionIds,
+} from "../../auto-reply/reply/reply-run-registry.js";
+import { resolveGlobalSingleton } from "../../shared/global-singleton.js";
+
+export type EmbeddedPiQueueHandle = {
+  kind?: "embedded";
+  queueMessage: (text: string) => Promise<void>;
+  isStreaming: () => boolean;
+  isCompacting: () => boolean;
+  cancel?: (reason?: "user_abort" | "restart" | "superseded") => void;
+  abort: () => void;
+};
+
+export type ActiveEmbeddedRunSnapshot = {
+  transcriptLeafId: string | null;
+  messages?: unknown[];
+  inFlightPrompt?: string;
+};
+
+export type EmbeddedRunModelSwitchRequest = {
+  provider: string;
+  model: string;
+  authProfileId?: string;
+  authProfileIdSource?: "auto" | "user";
+};
+
+export type EmbeddedRunWaiter = {
+  resolve: (ended: boolean) => void;
+  timer: NodeJS.Timeout;
+};
+
+const EMBEDDED_RUN_STATE_KEY = Symbol.for("openclaw.embeddedRunState");
+
+const embeddedRunState = resolveGlobalSingleton(EMBEDDED_RUN_STATE_KEY, () => ({
+  activeRuns: new Map<string, EmbeddedPiQueueHandle>(),
+  snapshots: new Map<string, ActiveEmbeddedRunSnapshot>(),
+  sessionIdsByKey: new Map<string, string>(),
+  waiters: new Map<string, Set<EmbeddedRunWaiter>>(),
+  modelSwitchRequests: new Map<string, EmbeddedRunModelSwitchRequest>(),
+}));
+
+export const ACTIVE_EMBEDDED_RUNS =
+  embeddedRunState.activeRuns ??
+  (embeddedRunState.activeRuns = new Map<string, EmbeddedPiQueueHandle>());
+export const ACTIVE_EMBEDDED_RUN_SNAPSHOTS =
+  embeddedRunState.snapshots ??
+  (embeddedRunState.snapshots = new Map<string, ActiveEmbeddedRunSnapshot>());
+export const ACTIVE_EMBEDDED_RUN_SESSION_IDS_BY_KEY =
+  embeddedRunState.sessionIdsByKey ??
+  (embeddedRunState.sessionIdsByKey = new Map<string, string>());
+export const EMBEDDED_RUN_WAITERS =
+  embeddedRunState.waiters ??
+  (embeddedRunState.waiters = new Map<string, Set<EmbeddedRunWaiter>>());
+export const EMBEDDED_RUN_MODEL_SWITCH_REQUESTS =
+  embeddedRunState.modelSwitchRequests ??
+  (embeddedRunState.modelSwitchRequests = new Map<string, EmbeddedRunModelSwitchRequest>());
+
+export function getActiveEmbeddedRunCount(): number {
+  let activeCount = ACTIVE_EMBEDDED_RUNS.size;
+  for (const sessionId of listActiveReplyRunSessionIds()) {
+    if (!ACTIVE_EMBEDDED_RUNS.has(sessionId)) {
+      activeCount += 1;
+    }
+  }
+  return Math.max(activeCount, getActiveReplyRunCount());
+}

src/agents/pi-embedded-runner/runs.ts

@@ -1,10 +1,8 @@
 import {
   abortActiveReplyRuns,
   abortReplyRunBySessionId,
-  getActiveReplyRunCount,
   isReplyRunActiveForSessionId,
   isReplyRunStreamingForSessionId,
-  listActiveReplyRunSessionIds,
   queueReplyRunMessage,
   resolveActiveReplyRunSessionId,
   waitForReplyRunEndBySessionId,
@@ -14,64 +12,26 @@ import {
   logMessageQueued,
   logSessionStateChange,
 } from "../../logging/diagnostic.js";
-import { resolveGlobalSingleton } from "../../shared/global-singleton.js";
 import { normalizeOptionalString } from "../../shared/string-coerce.js";
+import {
+  ACTIVE_EMBEDDED_RUNS,
+  ACTIVE_EMBEDDED_RUN_SESSION_IDS_BY_KEY,
+  ACTIVE_EMBEDDED_RUN_SNAPSHOTS,
+  EMBEDDED_RUN_MODEL_SWITCH_REQUESTS,
+  EMBEDDED_RUN_WAITERS,
+  getActiveEmbeddedRunCount,
+  type ActiveEmbeddedRunSnapshot,
+  type EmbeddedPiQueueHandle,
+  type EmbeddedRunModelSwitchRequest,
+  type EmbeddedRunWaiter,
+} from "./run-state.js";
 
-export type EmbeddedPiQueueHandle = {
-  kind?: "embedded";
-  queueMessage: (text: string) => Promise<void>;
-  isStreaming: () => boolean;
-  isCompacting: () => boolean;
-  cancel?: (reason?: "user_abort" | "restart" | "superseded") => void;
-  abort: () => void;
-};
-
-export type ActiveEmbeddedRunSnapshot = {
-  transcriptLeafId: string | null;
-  messages?: unknown[];
-  inFlightPrompt?: string;
-};
-
-type EmbeddedRunWaiter = {
-  resolve: (ended: boolean) => void;
-  timer: NodeJS.Timeout;
-};
-
-export type EmbeddedRunModelSwitchRequest = {
-  provider: string;
-  model: string;
-  authProfileId?: string;
-  authProfileIdSource?: "auto" | "user";
-};
-
-/**
- * Use global singleton state so busy/streaming checks stay consistent even
- * when the bundler emits multiple copies of this module into separate chunks.
- */
-const EMBEDDED_RUN_STATE_KEY = Symbol.for("openclaw.embeddedRunState");
-
-const embeddedRunState = resolveGlobalSingleton(EMBEDDED_RUN_STATE_KEY, () => ({
-  activeRuns: new Map<string, EmbeddedPiQueueHandle>(),
-  snapshots: new Map<string, ActiveEmbeddedRunSnapshot>(),
-  sessionIdsByKey: new Map<string, string>(),
-  waiters: new Map<string, Set<EmbeddedRunWaiter>>(),
-  modelSwitchRequests: new Map<string, EmbeddedRunModelSwitchRequest>(),
-}));
-const ACTIVE_EMBEDDED_RUNS =
-  embeddedRunState.activeRuns ??
-  (embeddedRunState.activeRuns = new Map<string, EmbeddedPiQueueHandle>());
-const ACTIVE_EMBEDDED_RUN_SNAPSHOTS =
-  embeddedRunState.snapshots ??
-  (embeddedRunState.snapshots = new Map<string, ActiveEmbeddedRunSnapshot>());
-const ACTIVE_EMBEDDED_RUN_SESSION_IDS_BY_KEY =
-  embeddedRunState.sessionIdsByKey ??
-  (embeddedRunState.sessionIdsByKey = new Map<string, string>());
-const EMBEDDED_RUN_WAITERS =
-  embeddedRunState.waiters ??
-  (embeddedRunState.waiters = new Map<string, Set<EmbeddedRunWaiter>>());
-const EMBEDDED_RUN_MODEL_SWITCH_REQUESTS =
-  embeddedRunState.modelSwitchRequests ??
-  (embeddedRunState.modelSwitchRequests = new Map<string, EmbeddedRunModelSwitchRequest>());
+export {
+  getActiveEmbeddedRunCount,
+  type ActiveEmbeddedRunSnapshot,
+  type EmbeddedPiQueueHandle,
+  type EmbeddedRunModelSwitchRequest,
+} from "./run-state.js";
 
 function setActiveRunSessionKey(sessionKey: string | undefined, sessionId: string): void {
   const normalizedSessionKey = sessionKey?.trim();
@@ -216,16 +176,6 @@ export function resolveActiveEmbeddedRunSessionId(sessionKey: string): string |
   );
 }
 
-export function getActiveEmbeddedRunCount(): number {
-  let activeCount = ACTIVE_EMBEDDED_RUNS.size;
-  for (const sessionId of listActiveReplyRunSessionIds()) {
-    if (!ACTIVE_EMBEDDED_RUNS.has(sessionId)) {
-      activeCount += 1;
-    }
-  }
-  return Math.max(activeCount, getActiveReplyRunCount());
-}
-
 export function getActiveEmbeddedRunSnapshot(
   sessionId: string,
 ): ActiveEmbeddedRunSnapshot | undefined {

src/gateway/server-reload-handlers.ts

@@ -1,6 +1,6 @@
 import { resetModelCatalogCache } from "../agents/model-catalog.js";
 import { disposeAllSessionMcpRuntimes } from "../agents/pi-bundle-mcp-tools.js";
-import { getActiveEmbeddedRunCount } from "../agents/pi-embedded-runner/runs.js";
+import { getActiveEmbeddedRunCount } from "../agents/pi-embedded-runner/run-state.js";
 import { getTotalPendingReplies } from "../auto-reply/reply/dispatcher-registry.js";
 import type { CliDeps } from "../cli/deps.types.js";
 import { resolveAgentMaxConcurrent, resolveSubagentMaxConcurrent } from "../config/agent-limits.js";

src/gateway/server.impl.ts

@@ -1,4 +1,4 @@
-import { getActiveEmbeddedRunCount } from "../agents/pi-embedded-runner/runs.js";
+import { getActiveEmbeddedRunCount } from "../agents/pi-embedded-runner/run-state.js";
 import { getTotalPendingReplies } from "../auto-reply/reply/dispatcher-registry.js";
 import type { CanvasHostServer } from "../canvas-host/server.js";
 import { type ChannelId, listChannelPlugins } from "../channels/plugins/index.js";

src/gateway/server.reload.test.ts

@@ -216,6 +216,16 @@ vi.mock("../agents/pi-embedded-runner/runs.js", async () => {
   };
 });
 
+vi.mock("../agents/pi-embedded-runner/run-state.js", async () => {
+  const actual = await vi.importActual<typeof import("../agents/pi-embedded-runner/run-state.js")>(
+    "../agents/pi-embedded-runner/run-state.js",
+  );
+  return {
+    ...actual,
+    getActiveEmbeddedRunCount: () => hoisted.activeEmbeddedRunCount.value,
+  };
+});
+
 vi.mock("../auto-reply/reply/dispatcher-registry.js", async () => {
   const actual = await vi.importActual<typeof import("../auto-reply/reply/dispatcher-registry.js")>(
     "../auto-reply/reply/dispatcher-registry.js",