Enhanced ChatGPT Clone: Features Agents, MCP, Skills, DeepSeek, Anthropic, AWS, OpenAI, Responses API, Azure, Groq, o1, GPT-5, Mistral, OpenRouter, Vertex AI, Gemini, Artifacts, AI model switching, message search, Code Interpreter, langchain, DALL-E-3, OpenAPI Actions, Functions, Secure Multi-User Auth, Presets, open-source for self-hosting. Active
TypeScript aianthropicartifactsawsazurechatgptchatgpt-cloneclaudeclonedeepseekgeminigooglegpt-5librechatmcpo1openairesponses-apivisionwebui
🪪 fix: Preserve Detached Subagent Owner Context (#15015)
Danny Avilaec5283452e41
diff --git a/packages/api/src/agents/subagentThreads.spec.ts b/packages/api/src/agents/subagentThreads.spec.tsindex 649bb8a010..e513f12e2a 100644--- a/packages/api/src/agents/subagentThreads.spec.ts+++ b/packages/api/src/agents/subagentThreads.spec.ts@@ -2,8 +2,15 @@ import mongoose from 'mongoose';import { randomUUID } from 'node:crypto';import { MongoMemoryServer } from 'mongodb-memory-server';import { Constants, EModelEndpoint } from 'librechat-data-provider';-import { createMethods, createModels, logger } from '@librechat/data-schemas';import { AIMessage, HumanMessage } from '@librechat/agents/langchain/messages';+import {+ createMethods,+ createModels,+ getTenantId,+ getUserId,+ logger,+ tenantStorage,+} from '@librechat/data-schemas';import type {SubagentTaskClaim,SubagentTaskControlCommand,@@ -376,6 +383,50 @@ describe('SubagentThreadTaskStore', () => {expect(await methods.getConvo(userId, requireThreadId(started))).not.toBeNull();});+ it('reconstructs trusted owner context after the admitting request has ended', async () => {+ const userId = 'detached-context-user';+ const tenantId = 'detached-context-tenant';+ const parentConversationId = randomUUID();+ await tenantStorage.run({ tenantId, userId }, async () =>+ saveParent(userId, parentConversationId, { tenantId }),+ );++ const observedContexts: Array<{ tenantId?: string; userId?: string }> = [];+ const observeContext = () => {+ observedContexts.push({ tenantId: getTenantId(), userId: getUserId() });+ };+ const store = new SubagentThreadTaskStore(methods, {+ isOwnerActive: async () => {+ observeContext();+ return true;+ },+ });+ const config = buildSubagentThreadTaskConfig(store, {+ userId,+ tenantId,+ parentConversationId,+ });+ const defaultRun = taskRequest(config.scopeId).run;+ const run = jest.fn(async (...args: Parameters<typeof defaultRun>) => {+ observeContext();+ return defaultRun(...args);+ });++ /** `start` deliberately runs outside `tenantStorage.run`: the detached task+ * owns only its serialized host scope once the HTTP request has returned. */+ const started = store.start(taskRequest(config.scopeId, { run }));+ await waitForSettled(store, config.scopeId, started);++ expect(run).toHaveBeenCalledTimes(1);+ expect(observedContexts.length).toBeGreaterThan(0);+ expect(observedContexts).toEqual(observedContexts.map(() => ({ tenantId, userId })));+ const messages = await tenantStorage.run({ tenantId, userId }, async () =>+ methods.getMessages({ user: userId, conversationId: requireThreadId(started) }),+ );+ expect(messages).toHaveLength(2);+ expect(messages.every((message) => message.tenantId === tenantId)).toBe(true);+ });+it('fails without leaving an orphan when parent persistence rejects', async () => {const userId = 'parent-gate-failure-user';const parentConversationId = randomUUID();diff --git a/packages/api/src/agents/subagentThreads.ts b/packages/api/src/agents/subagentThreads.tsindex becac41104..eb3f21fcac 100644--- a/packages/api/src/agents/subagentThreads.ts+++ b/packages/api/src/agents/subagentThreads.ts@@ -1,6 +1,6 @@import { randomUUID } from 'node:crypto';-import { logger } from '@librechat/data-schemas';import { InMemorySubagentTaskStore } from '@librechat/agents';+import { logger, tenantStorage } from '@librechat/data-schemas';import { EModelEndpoint, Constants } from 'librechat-data-provider';import {mapChatMessagesToStoredMessages,@@ -63,6 +63,7 @@ export interface SubagentCancellationPlan {}/** Three missed 10-second transport heartbeats retire a crashed owner. */const DEFAULT_TASK_ROUTING_TTL_MS = 30_000;+const SLOW_PREPARATION_WARN_MS = 5_000;const MAX_TRANSCRIPT_BYTES = 12 * 1024 * 1024;const TRANSCRIPT_SELECT ='messageId parentMessageId text createdAt +subagentTranscript +subagentTask';@@ -380,6 +381,21 @@ function safeErrorMessage(error: unknown): string {return `Subagent task failed: ${publicFailureDetail(error).slice(0, 2_000)}`;}+async function observeSlowPreparation<T>(+ operation: Promise<T>,+ context: { stage: string; taskId: string; threadId: string },+): Promise<T> {+ const warning = setTimeout(() => {+ logger.warn('[subagentThreads] Child-thread preparation is still waiting', context);+ }, SLOW_PREPARATION_WARN_MS);+ warning.unref?.();+ try {+ return await operation;+ } finally {+ clearTimeout(warning);+ }+}+/** Persists view-only logical child threads with owner-routed controls and a shared execution fence. */export class SubagentThreadTaskStore extends InMemorySubagentTaskStore {readonly supportsThreadContinuation = true;@@ -519,128 +535,129 @@ export class SubagentThreadTaskStore extends InMemorySubagentTaskStore {started = super.start({...request,threadId,- run: async (runtime: SubagentTaskRuntime) => {- lease.taskId = runtime.taskId;- lease.running = true;- const detachedUsage: UsageMetadata[] = [];- let prepared: PreparedThread | undefined;- try {- if (runtime.signal.aborted) {- throw runtime.signal.reason ?? new Error('Subagent task was cancelled.');- }- /** Publish the owner address before any provider work: a child running- * while unaddressable cannot be polled, controlled, or cancelled, and its- * side effects would already have happened by the time a heartbeat- * republished it. A failed registration fails the task closed instead. */- await this.taskControlTransport?.registerTask(- request.scopeId,- runtime.taskId,- this.taskRoutingTtlMs,- );- await parentReady;- prepared = await this.prepareThread(- request.scopeId,- scope,- threadId,- isContinuation,- request,- runtime.taskId,- lease,- );- await this.registerTaskWakeup(scope, prepared.conversation.conversationId, request, {- taskId: prepared.replay?.taskId ?? runtime.taskId,- parentRunId: prepared.replay?.parentRunId ?? request.parentRunId,- createdAt: prepared.taskCreatedAt,- });- if (runtime.signal.aborted) {- throw runtime.signal.reason ?? new Error('Subagent task was cancelled.');- }- if (prepared.replay != null) {- if (prepared.replay.status === 'completed') {- return { content: prepared.replay.content };+ run: (runtime: SubagentTaskRuntime) =>+ this.runWithOwnerContext(scope, async () => {+ lease.taskId = runtime.taskId;+ lease.running = true;+ const detachedUsage: UsageMetadata[] = [];+ let prepared: PreparedThread | undefined;+ try {+ if (runtime.signal.aborted) {+ throw runtime.signal.reason ?? new Error('Subagent task was cancelled.');}- throw new SubagentThreadPublicError(prepared.replay.content);- }- if (!(await this.renewSharedLease(scope, threadId, lease))) {- throw new SubagentThreadPublicError(- 'This child thread is already being continued by another run.',+ /** Publish the owner address before any provider work: a child running+ * while unaddressable cannot be polled, controlled, or cancelled, and its+ * side effects would already have happened by the time a heartbeat+ * republished it. A failed registration fails the task closed instead. */+ await this.taskControlTransport?.registerTask(+ request.scopeId,+ runtime.taskId,+ this.taskRoutingTtlMs,);- }- const preparedThread = prepared;- const result = await runWithDetachedSubagentUsage(detachedUsage, () =>- request.run(runtime, preparedThread.initialMessages),- );- if (runtime.signal.aborted) {- throw runtime.signal.reason ?? new Error('Subagent task was cancelled.');- }- if (!(await this.renewSharedLease(scope, threadId, lease))) {- throw new SubagentThreadPublicError(- 'This child thread is already being continued by another run.',+ await parentReady;+ prepared = await this.prepareThread(+ request.scopeId,+ scope,+ threadId,+ isContinuation,+ request,+ runtime.taskId,+ lease,+ );+ await this.registerTaskWakeup(scope, prepared.conversation.conversationId, request, {+ taskId: prepared.replay?.taskId ?? runtime.taskId,+ parentRunId: prepared.replay?.parentRunId ?? request.parentRunId,+ createdAt: prepared.taskCreatedAt,+ });+ if (runtime.signal.aborted) {+ throw runtime.signal.reason ?? new Error('Subagent task was cancelled.');+ }+ if (prepared.replay != null) {+ if (prepared.replay.status === 'completed') {+ return { content: prepared.replay.content };+ }+ throw new SubagentThreadPublicError(prepared.replay.content);+ }+ if (!(await this.renewSharedLease(scope, threadId, lease))) {+ throw new SubagentThreadPublicError(+ 'This child thread is already being continued by another run.',+ );+ }+ const preparedThread = prepared;+ const result = await runWithDetachedSubagentUsage(detachedUsage, () =>+ request.run(runtime, preparedThread.initialMessages),+ );+ if (runtime.signal.aborted) {+ throw runtime.signal.reason ?? new Error('Subagent task was cancelled.');+ }+ if (!(await this.renewSharedLease(scope, threadId, lease))) {+ throw new SubagentThreadPublicError(+ 'This child thread is already being continued by another run.',+ );+ }+ lease.settling = true;+ await this.persistResult(+ scope,+ request,+ runtime.taskId,+ prepared,+ result,+ detachedUsage,+ );+ return result;+ } catch (error) {+ /** A replay is already terminal in Mongo. A temporary wakeup-queue+ * outage must not overwrite that canonical result with a new error. */+ if (prepared?.replay != null) {+ throw error;+ }+ const mayPersist =+ lease.shared == null || (await this.renewSharedLease(scope, threadId, lease));+ const terminalTask = this.get(request.scopeId, runtime.taskId);+ if (runtime.signal.aborted && terminalTask?.status === 'cancelled') {+ if (mayPersist) {+ await this.persistCancellation(+ scope,+ threadId,+ request,+ runtime.taskId,+ detachedUsage,+ ).catch((persistError) => {+ logger.error(+ '[subagentThreads] Failed to persist child-thread cancellation',+ persistError,+ );+ });+ }+ throw error;+ }+ logger.error(+ '[subagentThreads] Child-thread execution failed',+ publicFailureDetail(error),);- }- lease.settling = true;- await this.persistResult(- scope,- request,- runtime.taskId,- prepared,- result,- detachedUsage,- );- return result;- } catch (error) {- /** A replay is already terminal in Mongo. A temporary wakeup-queue- * outage must not overwrite that canonical result with a new error. */- if (prepared?.replay != null) {- throw error;- }- const mayPersist =- lease.shared == null || (await this.renewSharedLease(scope, threadId, lease));- const terminalTask = this.get(request.scopeId, runtime.taskId);- if (runtime.signal.aborted && terminalTask?.status === 'cancelled') {if (mayPersist) {- await this.persistCancellation(+ await this.persistFailure(scope,threadId,request,runtime.taskId,+ error,detachedUsage,).catch((persistError) => {logger.error(- '[subagentThreads] Failed to persist child-thread cancellation',+ '[subagentThreads] Failed to persist child-thread failure',persistError,);});}- throw error;- }- logger.error(- '[subagentThreads] Child-thread execution failed',- publicFailureDetail(error),- );- if (mayPersist) {- await this.persistFailure(- scope,- threadId,- request,- runtime.taskId,- error,- detachedUsage,- ).catch((persistError) => {- logger.error(- '[subagentThreads] Failed to persist child-thread failure',- persistError,- );- });- }- throw new Error(publicFailureDetail(error));- } finally {- await this.stopAndReleaseSharedLease(scope, threadId, lease);- if (this.activeThreads.get(lockKey) === lease) {- this.activeThreads.delete(lockKey);+ throw new Error(publicFailureDetail(error));+ } finally {+ await this.stopAndReleaseSharedLease(scope, threadId, lease);+ if (this.activeThreads.get(lockKey) === lease) {+ this.activeThreads.delete(lockKey);+ }}- }- },+ }),});} catch (error) {if (ownsLease && this.activeThreads.get(lockKey) === lease) {@@ -657,6 +674,20 @@ export class SubagentThreadTaskStore extends InMemorySubagentTaskStore {return started;}+ /** Detached tasks intentionally outlive the HTTP request that admitted them.+ * Reconstruct only the trusted owner identity carried by the opaque task scope+ * so tenant-isolated database reads and lazy child initialization do not depend+ * on request AsyncLocalStorage remaining alive after the parent turn returns. */+ private runWithOwnerContext<T>(scope: SubagentThreadScope, run: () => Promise<T>): Promise<T> {+ return tenantStorage.run(+ {+ userId: scope.userId,+ ...(scope.tenantId == null ? {} : { tenantId: scope.tenantId }),+ },+ run,+ );+ }+/*** Claims locally when possible, otherwise asks the registered owning replica.*@@ -1566,13 +1597,22 @@ export class SubagentThreadTaskStore extends InMemorySubagentTaskStore {/** Account deletion can fence the owner after the optimistic probe but before* this lease exists. Once the lease is visible, revalidate so deletion either* observes and drains us or wins before any provider work can begin. */- if (!(await this.isOwnerActive(scope.userId))) {+ if (+ !(await observeSlowPreparation(this.isOwnerActive(scope.userId), {+ stage: 'owner_recheck',+ taskId,+ threadId,+ }))+ ) {throw new SubagentThreadDeletedError('The thread owner is unavailable.');}- const allMessages = (await this.methods.getMessages(- { conversationId: threadId, user: scope.userId },- TRANSCRIPT_SELECT,- { sort: { createdAt: 1, _id: 1 } },+ const allMessages = (await observeSlowPreparation(+ this.methods.getMessages(+ { conversationId: threadId, user: scope.userId },+ TRANSCRIPT_SELECT,+ { sort: { createdAt: 1, _id: 1 } },+ ),+ { stage: 'transcript_read', taskId, threadId },)) as ThreadMessage[];const attemptKey = createSubagentAttemptKey(scopeId, request.idempotencyKey);const requestFingerprint = normalizedRequestFingerprint(request);@@ -1685,25 +1725,28 @@ export class SubagentThreadTaskStore extends InMemorySubagentTaskStore {break;}}- const savedUserMessage = await this.methods.saveMessage(- { userId: scope.userId },- {- messageId: userMessageId,- conversationId: threadId,- parentMessageId,- sender: 'User',- text: request.input,- endpoint: EModelEndpoint.agents,- isCreatedByUser: true,- subagentTask: {- attemptKey,- parentRunId: request.parentRunId,- ...(requestFingerprint == null ? {} : { requestFingerprint }),- status: 'running',+ const savedUserMessage = await observeSlowPreparation(+ this.methods.saveMessage(+ { userId: scope.userId },+ {+ messageId: userMessageId,+ conversationId: threadId,+ parentMessageId,+ sender: 'User',+ text: request.input,+ endpoint: EModelEndpoint.agents,+ isCreatedByUser: true,+ subagentTask: {+ attemptKey,+ parentRunId: request.parentRunId,+ ...(requestFingerprint == null ? {} : { requestFingerprint }),+ status: 'running',+ },+ ...retentionFields(conversation),},- ...retentionFields(conversation),- },- { context: 'SubagentThreadTaskStore.prepareThread' },+ { context: 'SubagentThreadTaskStore.prepareThread' },+ ),+ { stage: 'seed_write', taskId, threadId },);if (savedUserMessage == null) {throw new Error('Unable to persist the child-thread input.');