Demonstrates execution metrics tracking (tokens, cost, latency) in evaluations.
- Automatic token usage tracking (input, output, cached)
- Cost tracking in USD
- Execution duration in milliseconds
- Using metrics in code graders for performance evaluation
- Metrics available in evaluation results
# From repository root
cd examples/features
bun agentv eval execution-metrics/evals/dataset.eval.yaml --target mock_metrics_agentCreate .env in examples/features/:
EXECUTION_METRICS_DIR=/absolute/path/to/examples/features/execution-metricsevals/dataset.eval.yaml- Test cases showing metrics collection- Mock agent automatically returns realistic execution metrics