feat(enterprise): persist agent events and run heartbeats

backend/package.json

@@ -41,7 +41,7 @@
     "prepack": "npm run build",
     "typecheck": "tsc --noEmit",
     "test": "jest",
-    "test:core": "jest --runInBand --forceExit src/agent/communication/__tests__/agentMessageBus.test.ts src/agent/core/executors/__tests__/strategyExecutor.test.ts src/agent/core/executors/__tests__/hypothesisExecutor.test.ts src/agent/context/__tests__/enhancedSessionContext.test.ts src/tests/adbTools.test.ts src/services/__tests__/sessionLogger.test.ts src/services/__tests__/traceAnalysisSkillConfig.test.ts src/agent/agents/domain/__tests__/registry.test.ts src/agentv3/__tests__/sqlIncludeInjector.test.ts src/agentv3/__tests__/analysisPatternMemory.test.ts src/agentv3/__tests__/claudeRuntimeRuntimeSnapshots.test.ts src/middleware/__tests__/auth.test.ts src/services/__tests__/rbac.test.ts src/routes/__tests__/agentRoutesRbac.test.ts src/routes/__tests__/ownerGuardRoutes.test.ts src/routes/__tests__/requestContextRouteCoverage.test.ts src/middleware/__tests__/legacyApiCompatibility.test.ts src/services/__tests__/enterpriseDb.test.ts src/services/__tests__/enterpriseSchema.test.ts src/services/__tests__/enterpriseRepository.test.ts src/services/__tests__/processRss.test.ts src/services/__tests__/workingTraceProcessor.enterpriseIsolation.test.ts src/services/__tests__/traceProcessorLeaseStore.test.ts src/services/__tests__/traceProcessorLeaseModeDecision.test.ts src/services/__tests__/traceProcessorLeaseProcessorRouting.test.ts src/services/__tests__/traceProcessorSqlWorker.test.ts src/services/__tests__/traceProcessorRamBudget.test.ts src/scripts/__tests__/benchmarkTraceProcessorRss.test.ts src/services/__tests__/enterpriseKnowledgeScope.test.ts src/services/__tests__/enterpriseMigration.test.ts src/services/__tests__/runtimeSnapshotStore.test.ts src/services/providerManager/__tests__/localSecretStore.test.ts src/services/providerManager/__tests__/enterpriseProviderStore.test.ts src/routes/__tests__/enterpriseTraceMetadataRoutes.test.ts src/routes/__tests__/traceProcessorProxyRoutes.test.ts src/routes/__tests__/enterpriseReportRoutes.test.ts src/routes/__tests__/enterpriseRestartPersistence.test.ts",
+    "test:core": "jest --runInBand --forceExit src/agent/communication/__tests__/agentMessageBus.test.ts src/agent/core/executors/__tests__/strategyExecutor.test.ts src/agent/core/executors/__tests__/hypothesisExecutor.test.ts src/agent/context/__tests__/enhancedSessionContext.test.ts src/tests/adbTools.test.ts src/services/__tests__/sessionLogger.test.ts src/services/__tests__/traceAnalysisSkillConfig.test.ts src/agent/agents/domain/__tests__/registry.test.ts src/agentv3/__tests__/sqlIncludeInjector.test.ts src/agentv3/__tests__/analysisPatternMemory.test.ts src/agentv3/__tests__/claudeRuntimeRuntimeSnapshots.test.ts src/middleware/__tests__/auth.test.ts src/assistant/application/__tests__/assistantApplicationService.test.ts src/services/__tests__/rbac.test.ts src/routes/__tests__/agentRoutesRbac.test.ts src/routes/__tests__/ownerGuardRoutes.test.ts src/routes/__tests__/requestContextRouteCoverage.test.ts src/middleware/__tests__/legacyApiCompatibility.test.ts src/services/__tests__/enterpriseDb.test.ts src/services/__tests__/enterpriseSchema.test.ts src/services/__tests__/enterpriseRepository.test.ts src/services/__tests__/processRss.test.ts src/services/__tests__/workingTraceProcessor.enterpriseIsolation.test.ts src/services/__tests__/traceProcessorLeaseStore.test.ts src/services/__tests__/traceProcessorLeaseModeDecision.test.ts src/services/__tests__/traceProcessorLeaseProcessorRouting.test.ts src/services/__tests__/traceProcessorSqlWorker.test.ts src/services/__tests__/traceProcessorRamBudget.test.ts src/scripts/__tests__/benchmarkTraceProcessorRss.test.ts src/services/__tests__/analysisRunStore.test.ts src/services/__tests__/agentEventStore.test.ts src/services/__tests__/enterpriseKnowledgeScope.test.ts src/services/__tests__/enterpriseMigration.test.ts src/services/__tests__/runtimeSnapshotStore.test.ts src/services/providerManager/__tests__/localSecretStore.test.ts src/services/providerManager/__tests__/enterpriseProviderStore.test.ts src/routes/__tests__/enterpriseTraceMetadataRoutes.test.ts src/routes/__tests__/traceProcessorProxyRoutes.test.ts src/routes/__tests__/enterpriseReportRoutes.test.ts src/routes/__tests__/enterpriseRestartPersistence.test.ts",
     "test:watch": "jest --watch",
     "test:coverage": "jest --coverage",
     "test:unit": "jest --testPathPatterns=src/tests",

backend/src/assistant/application/__tests__/assistantApplicationService.test.ts

@@ -0,0 +1,60 @@
+// SPDX-License-Identifier: AGPL-3.0-or-later
+// Copyright (C) 2024-2026 Gracker (Chris)
+// This file is part of SmartPerfetto. See LICENSE for details.
+
+import { describe, expect, it, jest } from '@jest/globals';
+import {
+  AssistantApplicationService,
+  type ManagedAssistantSession,
+} from '../assistantApplicationService';
+
+function session(overrides: Partial<ManagedAssistantSession> = {}): ManagedAssistantSession {
+  return {
+    sessionId: 'session-a',
+    status: 'running',
+    createdAt: 1_777_000_000_000,
+    lastActivityAt: 1_777_000_000_000,
+    sseClients: [],
+    ...overrides,
+  };
+}
+
+describe('AssistantApplicationService cleanup', () => {
+  it('lets callers keep abandoned non-terminal sessions when an external run heartbeat is fresh', () => {
+    const service = new AssistantApplicationService<ManagedAssistantSession>();
+    const managed = session();
+    const onCleanup = jest.fn();
+    service.setSession(managed.sessionId, managed);
+
+    const removed = service.cleanupIdleSessions({
+      now: 1_777_000_010_000,
+      terminalMaxIdleMs: 1_000,
+      nonTerminalMaxIdleMs: 1_000,
+      shouldCleanup: (_sessionId, _session, context) => {
+        expect(context.isAbandonedNonTerminal).toBe(true);
+        return false;
+      },
+      onCleanup,
+    });
+
+    expect(removed).toEqual([]);
+    expect(onCleanup).not.toHaveBeenCalled();
+    expect(service.getSession(managed.sessionId)).toBe(managed);
+  });
+
+  it('still removes stale abandoned non-terminal sessions when the cleanup predicate allows it', () => {
+    const service = new AssistantApplicationService<ManagedAssistantSession>();
+    const managed = session();
+    service.setSession(managed.sessionId, managed);
+
+    const removed = service.cleanupIdleSessions({
+      now: 1_777_000_010_000,
+      terminalMaxIdleMs: 1_000,
+      nonTerminalMaxIdleMs: 1_000,
+      shouldCleanup: () => true,
+    });
+
+    expect(removed).toEqual([managed.sessionId]);
+    expect(service.getSession(managed.sessionId)).toBeUndefined();
+  });
+});

backend/src/assistant/application/assistantApplicationService.ts

@@ -24,6 +24,16 @@ export interface SessionCleanupOptions<T extends ManagedAssistantSession> {
   terminalMaxIdleMs: number;
   nonTerminalMaxIdleMs: number;
   now?: number;
+  shouldCleanup?: (
+    sessionId: string,
+    session: T,
+    context: {
+      now: number;
+      idleMs: number;
+      isTerminal: boolean;
+      isAbandonedNonTerminal: boolean;
+    },
+  ) => boolean;
   onCleanup?: (sessionId: string, session: T) => void;
 }
 
@@ -99,6 +109,14 @@ export class AssistantApplicationService<T extends ManagedAssistantSession> {
         (isTerminal && idle > options.terminalMaxIdleMs) ||
         (isAbandonedNonTerminal && idle > options.nonTerminalMaxIdleMs)
       ) {
+        if (options.shouldCleanup?.(sessionId, session, {
+          now,
+          idleMs: idle,
+          isTerminal,
+          isAbandonedNonTerminal,
+        }) === false) {
+          continue;
+        }
         options.onCleanup?.(sessionId, session);
         this.sessions.delete(sessionId);
         removed.push(sessionId);

backend/src/routes/__tests__/agentRoutesRbac.test.ts

@@ -11,6 +11,14 @@ import request from 'supertest';
 import { ENTERPRISE_FEATURE_FLAG_ENV } from '../../config';
 import { ENTERPRISE_DB_PATH_ENV } from '../../services/enterpriseDb';
 import { ENTERPRISE_DATA_DIR_ENV, writeTraceMetadata } from '../../services/traceMetadataStore';
+import {
+  persistSerializedAgentEvent,
+  resetAgentEventStoreForTests,
+} from '../../services/agentEventStore';
+import {
+  getAnalysisRunLifecycle,
+  resetAnalysisRunStoreForTests,
+} from '../../services/analysisRunStore';
 import {
   getTraceProcessorLeaseStore,
   setTraceProcessorLeaseStoreForTests,
@@ -64,6 +72,8 @@ afterEach(async () => {
   jest.restoreAllMocks();
   setTraceProcessorServiceForTests(null);
   setTraceProcessorLeaseStoreForTests(null);
+  resetAgentEventStoreForTests();
+  resetAnalysisRunStoreForTests();
   if (originalApiKey === undefined) {
     delete process.env.SMARTPERFETTO_API_KEY;
   } else {
@@ -222,4 +232,108 @@ describe('agent route RBAC', () => {
       await fs.rm(tmpDir, { recursive: true, force: true });
     }
   });
+
+  it('replays persisted terminal SSE events before falling back to the in-memory buffer', async () => {
+    const tmpDir = await fs.mkdtemp(path.join(os.tmpdir(), 'smartperfetto-agent-event-replay-'));
+    let leaseStore: ReturnType<typeof getTraceProcessorLeaseStore> | null = null;
+    try {
+      const traceId = 'trace-agent-event-replay';
+      const tracePath = path.join(tmpDir, `${traceId}.trace`);
+      await fs.writeFile(tracePath, 'trace bytes');
+      delete process.env.SMARTPERFETTO_API_KEY;
+      process.env.SMARTPERFETTO_SSO_TRUSTED_HEADERS = 'true';
+      process.env[ENTERPRISE_FEATURE_FLAG_ENV] = 'true';
+      process.env[ENTERPRISE_DB_PATH_ENV] = path.join(tmpDir, 'enterprise.sqlite');
+      process.env[ENTERPRISE_DATA_DIR_ENV] = path.join(tmpDir, 'data');
+      process.env.UPLOAD_DIR = path.join(tmpDir, 'uploads');
+
+      await writeTraceMetadata({
+        id: traceId,
+        filename: `${traceId}.trace`,
+        size: 11,
+        uploadedAt: new Date().toISOString(),
+        status: 'ready',
+        path: tracePath,
+        tenantId: 'tenant-a',
+        workspaceId: 'workspace-a',
+        userId: 'analyst-user',
+      });
+      setTraceProcessorServiceForTests({
+        getOrLoadTrace: jest.fn(async () => ({
+          id: traceId,
+          filename: `${traceId}.trace`,
+          size: 11,
+          filePath: tracePath,
+          uploadTime: new Date(),
+          status: 'ready',
+        })),
+        getTrace: jest.fn(() => ({
+          id: traceId,
+          filename: `${traceId}.trace`,
+          size: 11,
+          filePath: tracePath,
+          uploadTime: new Date(),
+          status: 'ready',
+        })),
+        ensureProcessorForLease: jest.fn(async () => undefined),
+        runWithLease: jest.fn(() => new Promise<unknown>(() => undefined)),
+        query: jest.fn(async () => ({ columns: [], rows: [], durationMs: 1 })),
+      } as any);
+
+      const analyzeRes = await analystHeaders(request(makeApp()).post('/api/agent/v1/analyze'))
+        .send({ traceId, query: 'analyze this trace' });
+
+      expect(analyzeRes.status).toBe(200);
+      const { sessionId, runId } = analyzeRes.body;
+      const persistedRun = getAnalysisRunLifecycle({
+        tenantId: 'tenant-a',
+        workspaceId: 'workspace-a',
+        userId: 'analyst-user',
+      }, runId);
+      expect(persistedRun).toEqual(expect.objectContaining({
+        id: runId,
+        status: 'running',
+      }));
+      expect(persistedRun?.heartbeatAt).toEqual(expect.any(Number));
+      persistSerializedAgentEvent({
+        tenantId: 'tenant-a',
+        workspaceId: 'workspace-a',
+        userId: 'analyst-user',
+        sessionId,
+        runId,
+        traceId,
+        query: 'analyze this trace',
+      }, {
+        cursor: 99,
+        eventType: 'analysis_completed',
+        eventData: JSON.stringify({
+          type: 'analysis_completed',
+          data: { reportUrl: '/api/reports/report-from-db' },
+        }),
+        createdAt: 1_777_000_002_000,
+      });
+
+      const streamRes = await analystHeaders(
+        request(makeApp())
+          .get(`/api/agent/v1/${sessionId}/stream`)
+          .set('Last-Event-ID', '98')
+          .set('Accept', 'text/event-stream'),
+      );
+
+      expect(streamRes.status).toBe(200);
+      expect(streamRes.text).toContain('id: 99');
+      expect(streamRes.text).toContain('event: analysis_completed');
+      expect(streamRes.text).toContain('/api/reports/report-from-db');
+      leaseStore = getTraceProcessorLeaseStore();
+      expect(leaseStore.listLeases({
+        tenantId: 'tenant-a',
+        workspaceId: 'workspace-a',
+        userId: 'analyst-user',
+      }, { traceId })).toHaveLength(1);
+    } finally {
+      leaseStore?.close();
+      setTraceProcessorLeaseStoreForTests(null);
+      await fs.rm(tmpDir, { recursive: true, force: true });
+    }
+  });
 });