test(10-02): session rotation during execution edge case tests (D-05.3)
- Rotation triggered when run count exceeds maxRuns threshold - Rotation triggered when token count exceeds maxTokens threshold - Handoff markdown generated with task title and run history - New session after rotation has clean counters (0 runs, 0 tokens) - Claude connector exempt from rotation (nativeContextManagement) - Codex connector exempt from rotation (nativeContextManagement) - Handoff markdown available for next run context after rotation
This commit is contained in:
ZacharyZhang-NY committed
1 parent
2f80c319f9
commit
c395dbe8bf
1 file changed
+365
@@ -0,0 +1,365 @@
|
||||
import { join } from "node:path";
|
||||
import { tmpdir } from "node:os";
|
||||
import { mkdtemp, rm } from "node:fs/promises";
|
||||
import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
|
||||
|
||||
vi.mock("electron", () => ({
|
||||
safeStorage: {
|
||||
isEncryptionAvailable: () => false,
|
||||
getSelectedStorageBackend: () => "basic_text",
|
||||
encryptString: (value: string) => Buffer.from(value, "utf8"),
|
||||
decryptString: (value: Buffer) => value.toString("utf8"),
|
||||
},
|
||||
}));
|
||||
|
||||
import { AppDatabase } from "../../src/main/database";
|
||||
import { createSessionManagementService, DEFAULT_THRESHOLDS } from "../../src/main/session-management-service";
|
||||
import type { ConnectorId } from "@shared/types";
|
||||
|
||||
// ---------------------------------------------------------------------------
|
||||
// Helpers
|
||||
// ---------------------------------------------------------------------------
|
||||
|
||||
function scaffold(db: AppDatabase, connectorId: ConnectorId = "gemini_local") {
|
||||
const companyId = db.saveCompany({
|
||||
name: "Session Edge Case Co",
|
||||
description: "Session rotation edge case tests (D-05.3)",
|
||||
status: "active",
|
||||
});
|
||||
|
||||
// Always save gemini_local (needed for most tests)
|
||||
db.saveConnector({
|
||||
id: "gemini_local",
|
||||
command: "gemini",
|
||||
model: "gemini-pro",
|
||||
envBindingText: "",
|
||||
notes: "Test connector - no native context management",
|
||||
});
|
||||
|
||||
const workspaceId = db.saveWorkspace({
|
||||
companyId,
|
||||
name: "Primary",
|
||||
localPath: "/tmp/session-edge-test",
|
||||
repoUrl: "",
|
||||
repoRef: "",
|
||||
isPrimary: true,
|
||||
});
|
||||
const agentId = db.saveAgent({
|
||||
companyId,
|
||||
name: "Worker Agent",
|
||||
role: "developer",
|
||||
title: "Developer",
|
||||
status: "active",
|
||||
reportsTo: null,
|
||||
connectorId,
|
||||
workspaceId,
|
||||
model: null,
|
||||
capabilities: "Executes tasks",
|
||||
budgetMonthlyUsd: 100,
|
||||
});
|
||||
const goalId = db.saveGoal({
|
||||
companyId,
|
||||
title: "Ship feature",
|
||||
description: "Build session rotation edge case test",
|
||||
ownerAgentId: agentId,
|
||||
status: "active",
|
||||
});
|
||||
const projectId = db.saveProject({
|
||||
companyId,
|
||||
goalId,
|
||||
name: "Session Project",
|
||||
description: "Testing session rotation during execution",
|
||||
leadAgentId: agentId,
|
||||
status: "in_progress",
|
||||
targetDate: null,
|
||||
});
|
||||
const taskId = db.saveTask({
|
||||
companyId,
|
||||
projectId,
|
||||
goalId,
|
||||
title: "Implement session rotation feature",
|
||||
description: "Build the session rotation logic for task execution edge cases.",
|
||||
assigneeAgentId: agentId,
|
||||
workspaceId,
|
||||
priority: "high",
|
||||
status: "in_progress",
|
||||
});
|
||||
return { companyId, workspaceId, agentId, goalId, projectId, taskId };
|
||||
}
|
||||
|
||||
// ---------------------------------------------------------------------------
|
||||
// Tests -- Edge Case: Session Rotation During Execution (D-05.3)
|
||||
// ---------------------------------------------------------------------------
|
||||
|
||||
describe("Edge Case: Session Rotation During Execution (D-05.3)", () => {
|
||||
let tempDir: string;
|
||||
let db: AppDatabase;
|
||||
const mockLogger = { info: vi.fn(), warn: vi.fn(), error: vi.fn() };
|
||||
|
||||
beforeEach(async () => {
|
||||
tempDir = await mkdtemp(join(tmpdir(), "agentcompany-session-edge-"));
|
||||
db = new AppDatabase(tempDir);
|
||||
db.init();
|
||||
mockLogger.info.mockClear();
|
||||
mockLogger.warn.mockClear();
|
||||
mockLogger.error.mockClear();
|
||||
});
|
||||
|
||||
afterEach(async () => {
|
||||
db.close();
|
||||
await rm(tempDir, { recursive: true, force: true });
|
||||
});
|
||||
|
||||
it("rotation triggered when run count exceeds maxRuns threshold", () => {
|
||||
const { companyId, agentId, taskId } = scaffold(db);
|
||||
const sessionService = createSessionManagementService({ db, logger: mockLogger });
|
||||
|
||||
// Create a session
|
||||
const sessionId = db.createNewSession({
|
||||
taskId,
|
||||
companyId,
|
||||
agentId,
|
||||
adapterType: "gemini_local",
|
||||
sessionParamsJson: null,
|
||||
sessionDisplayId: null,
|
||||
});
|
||||
|
||||
// Simulate runs until we reach the maxRuns threshold (20)
|
||||
for (let i = 0; i < DEFAULT_THRESHOLDS.maxRuns; i++) {
|
||||
db.incrementSessionUsage(sessionId, 1000, 500, `run-${i}`);
|
||||
}
|
||||
|
||||
// Get the updated session from DB
|
||||
const session = db.getActiveTaskSession(taskId);
|
||||
expect(session).not.toBeNull();
|
||||
expect(session!.runCount).toBe(DEFAULT_THRESHOLDS.maxRuns);
|
||||
|
||||
// Evaluate rotation -- should trigger
|
||||
const result = sessionService.evaluateRotation(session!, "gemini_local");
|
||||
expect(result.shouldRotate).toBe(true);
|
||||
expect(result.reason).toBe("max_runs");
|
||||
});
|
||||
|
||||
it("rotation triggered when token count exceeds maxTokens threshold", () => {
|
||||
const { companyId, agentId, taskId } = scaffold(db);
|
||||
const sessionService = createSessionManagementService({ db, logger: mockLogger });
|
||||
|
||||
const sessionId = db.createNewSession({
|
||||
taskId,
|
||||
companyId,
|
||||
agentId,
|
||||
adapterType: "gemini_local",
|
||||
sessionParamsJson: null,
|
||||
sessionDisplayId: null,
|
||||
});
|
||||
|
||||
// Accumulate tokens past the threshold in one big call
|
||||
// DEFAULT_THRESHOLDS.maxTokens = 500_000 (total input + output)
|
||||
db.incrementSessionUsage(sessionId, 300000, 250000, "big-run");
|
||||
|
||||
const session = db.getActiveTaskSession(taskId);
|
||||
expect(session).not.toBeNull();
|
||||
const totalTokens = session!.totalInputTokens + session!.totalOutputTokens;
|
||||
expect(totalTokens).toBe(550000);
|
||||
|
||||
const result = sessionService.evaluateRotation(session!, "gemini_local");
|
||||
expect(result.shouldRotate).toBe(true);
|
||||
expect(result.reason).toBe("max_tokens");
|
||||
});
|
||||
|
||||
it("handoff markdown generated before next run starts", () => {
|
||||
const { companyId, agentId, taskId, workspaceId } = scaffold(db);
|
||||
const sessionService = createSessionManagementService({ db, logger: mockLogger });
|
||||
|
||||
const sessionId = db.createNewSession({
|
||||
taskId,
|
||||
companyId,
|
||||
agentId,
|
||||
adapterType: "gemini_local",
|
||||
sessionParamsJson: null,
|
||||
sessionDisplayId: null,
|
||||
});
|
||||
|
||||
// Accumulate enough to trigger rotation
|
||||
for (let i = 0; i < DEFAULT_THRESHOLDS.maxRuns; i++) {
|
||||
db.incrementSessionUsage(sessionId, 1000, 500, `run-${i}`);
|
||||
}
|
||||
|
||||
// Create a completed run for history (so generateHandoff has run data)
|
||||
const run = db.createRun({
|
||||
companyId,
|
||||
taskId,
|
||||
agentId,
|
||||
workspaceId,
|
||||
connectorId: "gemini_local" as ConnectorId,
|
||||
});
|
||||
db.finishRun({
|
||||
runId: run.id,
|
||||
status: "completed",
|
||||
summary: "Completed initial implementation",
|
||||
errorMessage: null,
|
||||
exitCode: 0,
|
||||
signal: null,
|
||||
model: "gemini-pro",
|
||||
sessionDisplayId: null,
|
||||
costUsd: 0.05,
|
||||
unattributedCost: false,
|
||||
});
|
||||
|
||||
const session = db.getActiveTaskSession(taskId)!;
|
||||
const handoff = sessionService.generateHandoff(taskId, session);
|
||||
|
||||
// Verify handoff markdown content
|
||||
expect(handoff).toContain("# Session Handoff");
|
||||
expect(handoff).toContain("Implement session rotation feature"); // task title
|
||||
expect(handoff).toContain("## Recent Run History");
|
||||
|
||||
// Archive the old session with handoff
|
||||
db.archiveSession(sessionId, "max_runs", handoff);
|
||||
|
||||
// Verify archived session has handoff stored
|
||||
const archivedRows = db.queryAll<{ handoffMarkdown: string | null }>(
|
||||
"SELECT handoff_markdown as handoffMarkdown FROM task_sessions WHERE id = ?",
|
||||
sessionId,
|
||||
);
|
||||
expect(archivedRows[0].handoffMarkdown).toContain("# Session Handoff");
|
||||
});
|
||||
|
||||
it("new session created after rotation with clean counters", () => {
|
||||
const { companyId, agentId, taskId } = scaffold(db);
|
||||
|
||||
// Create initial session
|
||||
const oldSessionId = db.createNewSession({
|
||||
taskId,
|
||||
companyId,
|
||||
agentId,
|
||||
adapterType: "gemini_local",
|
||||
sessionParamsJson: JSON.stringify({ resumeKey: "abc" }),
|
||||
sessionDisplayId: "gemini-session-1",
|
||||
});
|
||||
|
||||
// Add usage to the old session
|
||||
db.incrementSessionUsage(oldSessionId, 300000, 250000, "run-final");
|
||||
|
||||
// Archive the old session
|
||||
db.archiveSession(oldSessionId, "max_tokens", "# Session Handoff\nPrevious work summary.");
|
||||
|
||||
// Create new session
|
||||
const newSessionId = db.createNewSession({
|
||||
taskId,
|
||||
companyId,
|
||||
agentId,
|
||||
adapterType: "gemini_local",
|
||||
sessionParamsJson: JSON.stringify({ resumeKey: "abc" }),
|
||||
sessionDisplayId: "gemini-session-2",
|
||||
});
|
||||
|
||||
// Get the new active session
|
||||
const newSession = db.getActiveTaskSession(taskId);
|
||||
expect(newSession).not.toBeNull();
|
||||
expect(newSession!.id).toBe(newSessionId);
|
||||
expect(newSession!.id).not.toBe(oldSessionId);
|
||||
|
||||
// Verify clean counters
|
||||
expect(newSession!.runCount).toBe(0);
|
||||
expect(newSession!.totalInputTokens).toBe(0);
|
||||
expect(newSession!.totalOutputTokens).toBe(0);
|
||||
});
|
||||
|
||||
it("claude connector is exempt from rotation", () => {
|
||||
const { companyId, agentId, taskId } = scaffold(db, "gemini_local");
|
||||
const sessionService = createSessionManagementService({ db, logger: mockLogger });
|
||||
|
||||
const sessionId = db.createNewSession({
|
||||
taskId,
|
||||
companyId,
|
||||
agentId,
|
||||
adapterType: "claude_local",
|
||||
sessionParamsJson: null,
|
||||
sessionDisplayId: null,
|
||||
});
|
||||
|
||||
// Accumulate way past all thresholds
|
||||
for (let i = 0; i < 25; i++) {
|
||||
db.incrementSessionUsage(sessionId, 30000, 20000, `run-${i}`);
|
||||
}
|
||||
|
||||
const session = db.getActiveTaskSession(taskId)!;
|
||||
// Per Pitfall 6: claude_local has nativeContextManagement: true, exempt from rotation
|
||||
const result = sessionService.evaluateRotation(session, "claude_local");
|
||||
expect(result.shouldRotate).toBe(false);
|
||||
expect(result.reason).toBeNull();
|
||||
});
|
||||
|
||||
it("codex connector is exempt from rotation", () => {
|
||||
const { companyId, agentId, taskId } = scaffold(db, "gemini_local");
|
||||
const sessionService = createSessionManagementService({ db, logger: mockLogger });
|
||||
|
||||
const sessionId = db.createNewSession({
|
||||
taskId,
|
||||
companyId,
|
||||
agentId,
|
||||
adapterType: "codex_local",
|
||||
sessionParamsJson: null,
|
||||
sessionDisplayId: null,
|
||||
});
|
||||
|
||||
// Accumulate way past all thresholds
|
||||
for (let i = 0; i < 25; i++) {
|
||||
db.incrementSessionUsage(sessionId, 30000, 20000, `run-${i}`);
|
||||
}
|
||||
|
||||
const session = db.getActiveTaskSession(taskId)!;
|
||||
// Per Pitfall 6: codex_local has nativeContextManagement: true, exempt from rotation
|
||||
const result = sessionService.evaluateRotation(session, "codex_local");
|
||||
expect(result.shouldRotate).toBe(false);
|
||||
expect(result.reason).toBeNull();
|
||||
});
|
||||
|
||||
it("handoff markdown available for next run context after rotation", () => {
|
||||
const { companyId, agentId, taskId, workspaceId } = scaffold(db);
|
||||
const sessionService = createSessionManagementService({ db, logger: mockLogger });
|
||||
|
||||
// Create session and add usage
|
||||
const sessionId = db.createNewSession({
|
||||
taskId,
|
||||
companyId,
|
||||
agentId,
|
||||
adapterType: "gemini_local",
|
||||
sessionParamsJson: null,
|
||||
sessionDisplayId: null,
|
||||
});
|
||||
db.incrementSessionUsage(sessionId, 300000, 200000, "run-1");
|
||||
|
||||
// Create a completed run for run history
|
||||
const run = db.createRun({
|
||||
companyId,
|
||||
taskId,
|
||||
agentId,
|
||||
workspaceId,
|
||||
connectorId: "gemini_local" as ConnectorId,
|
||||
});
|
||||
db.finishRun({
|
||||
runId: run.id,
|
||||
status: "completed",
|
||||
summary: "Previous work done",
|
||||
errorMessage: null,
|
||||
exitCode: 0,
|
||||
signal: null,
|
||||
model: "gemini-pro",
|
||||
sessionDisplayId: null,
|
||||
costUsd: 0.01,
|
||||
unattributedCost: false,
|
||||
});
|
||||
|
||||
// Rotate the session (archives old, creates new)
|
||||
const oldSession = db.getActiveTaskSession(taskId)!;
|
||||
sessionService.rotateSession(oldSession);
|
||||
|
||||
// After rotation, the handoff should be retrievable for the next run
|
||||
const handoff = sessionService.getHandoffContext(taskId);
|
||||
expect(handoff).not.toBeNull();
|
||||
expect(handoff).toContain("# Session Handoff");
|
||||
expect(handoff).toContain("Implement session rotation feature"); // task title from scaffold
|
||||
});
|
||||
});
|
||||
Reference in new issue
Block a user