NCP - Natural Context Provider

# Scheduler Tests This directory contains tests for the NCP scheduler system. ## Test Files ### Unit Tests (Fast, Always Run) - **`job-manager.test.ts`** - Tests for legacy job management (backward compatibility) - **`task-manager.test.ts`** - Tests for new timing group and task management - **`cron-manager.test.ts`** - Tests for cron expression parsing and OS scheduler integration - **`natural-language-parser.test.ts`** - Tests for natural language schedule parsing - **`process-isolation.test.ts`** - Tests for timing groups and process isolation architecture ### Integration Tests - **`scheduler-integration.test.ts`** - End-to-end tests with mocked orchestrator - **`real-execution.test.ts`** - Real MCP execution tests (skipped by default, requires setup) ## Running Tests ```bash # Run all scheduler tests npm test -- tests/scheduler # Run specific test file npm test -- tests/scheduler/process-isolation.test.ts # Run with coverage npm test -- --coverage tests/scheduler ``` ## Test Architecture ### Unit Tests - Use isolated test directories (no shared state) - Mock file system paths - Fast execution (< 2 seconds total) - Run on every commit ### Integration Tests - Test with mocked MCP orchestrator - Verify data flow and error handling - Medium speed (< 5 seconds) - Run on every commit ### Real Execution Tests - **Location**: `real-execution.test.ts` - **Status**: Skipped by default (`.skip`) - **Purpose**: Test with actual MCP servers and tool execution - **Requirements**: - MCP servers configured - Actual network/file access - Longer timeouts - **Enable**: Remove `.skip` from test descriptions ## Manual Testing For testing with real MCP tool calls: 1. **Quick Test**: See `TESTING_WITH_REAL_MCPS.md` 2. **Guided Manual Test**: See `MANUAL_TESTING.md` ### Recommended Manual Test Flow ```bash # 1. Build npm run build # 2. Create a test task using built-in scheduler MCP node dist/index.js scheduler:create-task \ --name "Self test" \ --schedule "in 1 minute" \ --tool "scheduler:list-tasks" \ --parameters '{}' # 3. Wait and verify sleep 65 node dist/index.js scheduler:list-executions ``` ## Test Helpers `test-helpers.ts` provides utilities: - `createTestDirectory()` - Create isolated temp directory - `cleanupTestDirectory()` - Clean up after tests - `mockSchedulerEnvironment()` - Set up scheduler directory structure - `createMockJob()` - Create test job objects - `MockCrontab` - Mock OS scheduler for testing ## Process Isolation Testing The `process-isolation.test.ts` file verifies: 1. **Timing Groups**: Multiple tasks share one OS schedule 2. **Task Data Structure**: Tasks have metadata, timings group them 3. **Isolation Architecture**: Child processes prevent cascading failures 4. **Timing Cleanup**: Empty timing groups are removed These are unit tests that verify the **architecture**, not actual execution. For real execution testing, use `TESTING_WITH_REAL_MCPS.md`. ## What to Test Before Releasing ### Unit Tests (Automated) - [x] All unit tests pass - [x] Process isolation architecture verified - [x] Task and timing CRUD operations - [x] Natural language parsing ### Integration Tests (Automated) - [x] Job lifecycle works - [x] Validation catches errors - [x] Execution recording works ### Manual Tests (Before Release) - [ ] Create task with real MCP - [ ] Task executes on schedule - [ ] Multiple tasks run in parallel - [ ] Failed task doesn't crash others - [ ] Paused task doesn't execute - [ ] Completed task stops executing - [ ] Timing group cleanup works - [ ] OS scheduler integration works (macOS/Linux) ## Common Issues ### Tests fail with "Task already exists" - Tests aren't isolated - Make sure `createTestDirectory()` returns unique dirs - Check `beforeEach` sets up fresh environment ### Tests timeout - Increase timeout: `jest.setTimeout(30000)` - Check for hanging promises - Verify cleanup in `afterEach` ### "Cannot find module" errors - Run `npm run build` first - Check import paths use `.js` extension - Verify TypeScript compiled successfully ## Test Coverage Goals - Unit tests: > 90% - Integration tests: > 80% - Manual testing: 100% of user-facing features ## Future Test Improvements - [ ] Add performance benchmarks - [ ] Test with multiple MCP servers simultaneously - [ ] Test with long-running tasks (hours) - [ ] Test timing precision (execution within X seconds of schedule) - [ ] Test cleanup of old executions - [ ] Test migration from V1 to V2 storage