run_experiment

run_experiment(
    spec,
    *,
    task_root=None,
    output_root='outputs',
    metrics_output_root=None,
    timestamp=None,
    pull_model=True,
    start_ollama_if_needed=True,
    quantization_type=None,
)

Run one end-to-end experiment and evaluate it against ground truth.

Parameters

Name	Type	Description	Default
spec	ExperimentSpec	Concrete experiment specification to execute.	required
task_root	str \| Path \| None	Optional directory containing user-defined task folders with `codebook.json` and `ground-truth.csv`. If omitted, bundled example tasks are used when the task name matches one shipped with the package.	`None`
output_root	str \| Path	Root directory where per-run outputs should be created.	`'outputs'`
metrics_output_root	str \| Path \| None	Directory for aggregate metrics CSV files. Defaults to `output_root / "metrics"`.	`None`
timestamp	str \| None	Optional timestamp string to control output folder naming.	`None`
pull_model	bool	If `True`, run `ollama pull` for `spec.model` before execution. Defaults to `True` so experiment runs ensure the requested model is available.	`True`
start_ollama_if_needed	bool	If `True`, try to auto-start a local Ollama server when the default local host is not already reachable. Defaults to `True` so experiment runs can bring up local Ollama automatically when needed.	`True`
quantization_type	str \| None	Optional metadata field written to the metrics CSV.	`None`

Returns

Name	Type	Description
		`ExperimentRunResult` containing both annotation and metrics results.