Module deterministic 

Helpers for pinning generation parameters to reproducible values.

By default every bench run forces temperature=0.0 on the configured LLM provider so that two runs with identical inputs produce the same output. Pass --no-deterministic on the CLI (or no_deterministic = true to apply_deterministic_overrides) to opt out.

Functions

apply_deterministic_overrides

Optionally apply deterministic generation overrides to an AnyProvider.

deterministic_overrides

Build the GenerationOverrides that pin temperature to 0.0.