testing-patterns
Vercel AI SDK的测试模式,包括模拟提供商、流测试、工具调用测试、快照测试和测试覆盖策略。在实现测试、创建测试套件、模拟AI提供商,或用户提及测试、模拟、测试覆盖、AI测试、流测试或工具测试时使用。
📖 中文导读:Vercel AI SDK的测试模式,包括模拟提供商、流测试、工具调用测试、快照测试和测试覆盖策略。在实现测试、创建测试套件、模拟AI提供商,或用户提及测试、模拟、测试覆盖、AI测试、流测试或工具测试时使用。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: testing-patterns description: Testing patterns for Vercel AI SDK including mock providers, streaming tests, tool calling tests, snapshot testing, and test coverage strategies. Use when implementing tests, creating test suites, mocking AI providers, or when user mentions testing, mocks, test coverage, AI testing, streaming tests, or tool testing. allowed-tools: Read, Write, Bash, Grep, Glob
Testing Patterns for Vercel AI SDK
Purpose: Provide comprehensive testing templates, mock providers, and testing strategies for Vercel AI SDK applications.
Activation Triggers:
- Writing tests for AI features
- Creating mock AI providers
- Testing streaming responses
- Testing tool calling functionality
- Implementing test coverage
- Snapshot testing for AI outputs
Key Resources:
templates/mock-provider.ts- Mock language model implementationtemplates/streaming-test.ts- Test streaming responsestemplates/tool-calling-test.ts- Test tool executiontemplates/snapshot-test.ts- Snapshot testing for outputsscripts/generate-test-suite.sh- Generate test scaffoldsscripts/run-coverage.sh- Run tests with coverageexamples/- Complete test suites for different features
Mock Provider Pattern
Template: templates/mock-provider.ts
import { createMockLanguageModelV1 } from 'ai/test'
const mockProvider = createMockLanguageModelV1({
doGenerate: async ({ prompt, mode }) => ({
text: 'Mocked response'
finishReason: 'stop'
usage: { promptTokens: 10, completionTokens: 20 }
})
doStream: async function* ({ prompt, mode }) {
yield { type: 'text-delta', textDelta: 'Mocked ' }
yield { type: 'text-delta', textDelta: 'streaming ' }
yield { type: 'text-delta', textDelta: 'response' }
yield {
type: 'finish'
finishReason: 'stop'
usage: { promptTokens: 10, completionTokens: 20 }
}
}
})
Testing Strategies
1. Unit Tests (Mock Providers)
Test AI functions without real API calls
2. Integration Tests (Real Providers)
Test with real providers in CI (with rate limits)
3. Snapshot Tests
Ensure consistent outputs over time
4. E2E Tests
Test complete user flows with mocks
Test Coverage Goals
- Core Functions: >90%
- Error Handling: >80%
- Tool Calling: 100%
- Streaming: >85%
Resources
Templates:
mock-provider.ts- Complete mock implementationstreaming-test.ts- Streaming test patternstool-calling-test.ts- Tool execution testssnapshot-test.ts- Snapshot testing setup
Scripts:
generate-test-suite.sh- Scaffold testsrun-coverage.sh- Run with coverage
Examples:
complete-test-suite.test.ts- Full test suite example
Testing Frameworks: Vitest, Jest, Node Test Runner SDK Version: Vercel AI SDK 5+ Coverage Tool: c8, nyc, or built-in coverage
Best Practice: Use mock providers for fast, reliable tests