Skip to content

Provenance

Plugins that persist provenance for agentic harness executions.

  • BASELINE_CONDITION
  • CONDITION_KEY

Collect and persist metadata about LLM prompt/response pairs.

__init__(artifact_name: 'str' = 'llm_calls.json') → None

property PromptProvenancePlugin.calls_by_condition

Section titled “property PromptProvenancePlugin.calls_by_condition”

after_run(experiment, result) → None

after_step(experiment, step, data, ctx) → None

method PromptProvenancePlugin.before_replicate

Section titled “method PromptProvenancePlugin.before_replicate”
before_replicate(experiment, ctx) → None

before_run(experiment) → None

Persist an evidence bundle linking claim, spec, code, tests and verdicts.

__init__(filename: 'str' = 'bundle.json') → None

after_run(experiment, result) → None