Module engine 

Replay engine.

Given a baseline trace and an LlmBackend, produce a candidate trace with the same request order but LLM-derived responses freshly generated by the backend. See SPEC §10 for the algorithm.

Structs

FixedClock

A clock that returns a fixed incrementing counter. Used by tests.

Enums

ReplayError

Errors from run_replay.

Traits

Clock

Clock abstraction so tests can pin timestamps.

Functions

run_replay

Run a replay: walk baseline, dispatch every chat_request to backend, and produce a fresh trace with the same structure.