Expand description
Running the evaluation corpus against the real engine.
The harness exercises the same code path a live session would: a plan is derived from the utterance, queries are fused, and the assembler produces a budgeted snapshot. Nothing is stubbed, so a regression in ranking or budget shows up here rather than in production.
Structs§
- Ingestion
Case Result - Per-case ingestion outcome.
- Ingestion
Report - Aggregate ingestion report.
- Retrieval
Case Result - Per-case retrieval outcome.
- Retrieval
Report - Aggregate retrieval report.
Functions§
- run_
ingestion_ eval - Run the ingestion evaluation.
- run_
retrieval_ eval - Run the retrieval evaluation.
- user
- The evaluation user, re-exported for callers driving the harness directly.