chanl-eval drops multi-turn agent testing engine

The industry is moving from single-turn chatbots to autonomous agents, making traditional prompt-response evaluations insufficient for production-grade reliability. Dynamic persona simulation allows developers to stress-test agents against hostile or impatient customers, while comprehensive tool mocking enables verification of business logic without live API dependencies. The engine also includes built-in red-teaming presets for security-critical issues like prompt injection and PII leakage. Finally, exportable transcripts facilitate a continuous improvement loop between testing results and model fine-tuning.