Demonstrates tool trajectory evaluation combined with expected output validation.
- Static trace file evaluation (pre-recorded traces)
- Tool trajectory validation with full conversation structure
expected_outputfor comprehensive validation- Production-style multi-turn agent workflows
- Input/output validation for tool calls
# From repository root
cd examples/features
bun agentv eval tool-trajectory-advanced/evals/trace-file-demo.eval.yaml --target static_traceCreate .env in examples/features/:
TOOL_TRAJECTORY_DIR=/absolute/path/to/examples/features/tool-trajectory-advancedcat-trace.ts- CLI that reads static trace filesstatic-trace.json- Pre-captured trace dataevals/trace-file-demo.eval.yaml- Test cases with expected_output validation