Python API Reference
High-level orchestration
Primary entrypoint:
t2smetrics.run_experiments.run(...)
Core parameters:
jsonl_evals: list of input JSONL paths or foldersmetrics_list: metric names or metric instancesdataset: dataset label for exportsexecution_backend_graph_pathorexecution_backend_endpoint_urlparallel: run multiple systems in parallelper_query: include per-query values in result export
Experiment internals
JsonlEval: iteratesQueryCaserows from JSONLExperiment: wraps context, engine, and aggregationEvaluationEngine: executes metrics with requirement checks
Metric registry
t2smetrics.metrics.metrics_utils.get_metric_mapping()returns registered metrics.t2smetrics.metrics.metrics_utils.str_to_metric(name)resolves a metric name to an instance.