0.2.0-beta
Pre-release
Pre-release
AgentEval Public Beta Release 🎉
This release marks the transition from alpha to beta. The framework is now feature-complete for core scenarios and ready for community feedback.
Added
- Codecov Badge - Coverage visibility in README.md
- NuGet Consumer Sample (
samples/AgentEval.NuGetConsumer/) - Standalone project showcasing all major features- Tool chain assertions (HaveCalledTool, WithArgument, BeforeTool, AfterTool)
- Performance assertions (Duration, TTFT, Cost, Token limits)
- Behavioral policies (NeverCallTool, MustConfirmBefore, NeverPassArgumentMatching)
- Response assertions (Contain, NotContain, length validation)
- Mock testing with FakeChatClient
- Stochastic testing examples
- Model comparison patterns
- Agentic metrics overview
- Works offline with mock data - no Azure OpenAI required
- Custom Domain - AgentEval.dev documentation site with GitHub Pages
- Comprehensive Documentation - 25+ documentation pages with zero DocFX warnings
- Security Scanning - Enhanced pipeline with secret detection and dependency scanning
Changed
- Updated README test count badge to 3000+ (reflecting 1000+ tests × 3 TFMs)
- Documentation navigation reorganized with improved feature grouping
- Security scanning patterns refined to reduce false positives
- Version bumped from 0.1.3-alpha to 0.2.0-beta signaling production readiness
Documentation
- Getting Started, Assertions, Metrics Reference, Model Comparison guides
- Trace Record & Replay, Stochastic Testing, Benchmarks documentation
- CI/CD Integration guide with GitHub Actions examples
- Migration guide for Python/Node.js developers