Status: active
Primary bead: bd-1lsy.8.2
Machine-readable contract: docs/rgc_statistical_validation_pipeline_v1.json
This contract defines deterministic statistical validation for benchmark promotion decisions. It ensures variance, confidence, and effect-size controls are enforced before release-facing performance claims proceed.
The pipeline is fail-closed:
- rejects incomplete benchmark metadata,
- quarantines high-variance and low-confidence runs,
- fails on significant regression threshold breaches,
- emits replay-stable artifacts for each gate run.
schema_version:franken-engine.rgc-statistical-validation-pipeline.v1contract_version:1.0.0policy_id:policy-rgc-statistical-validation-pipeline-v1
Validation policy includes:
max_cv_millionthswarning_regression_millionthsfail_regression_millionthsmax_p_value_millionthsmin_effect_size_millionthsconfidence_level_millionths
All thresholds are deterministic and enforced in millionths to avoid float-only policy ambiguity.
Each workload evaluation emits an event with stable keys:
trace_iddecision_idpolicy_idcomponenteventscenario_idworkload_idoutcomeerror_code
Gate entrypoint:
scripts/run_rgc_statistical_validation_pipeline.sh
Replay wrapper:
scripts/e2e/rgc_statistical_validation_pipeline_replay.sh
Modes:
check,test,clippy,ci
Strict mode is fail-closed and requires remote execution for heavy cargo
operations (rch only, no local fallback).
Validation targets:
crates/franken-engine/tests/rgc_statistical_validation_pipeline.rscrates/franken-engine/tests/performance_statistical_validation_integration.rs
Each run emits:
run_manifest.jsonevents.jsonlcommands.txttrace_ids.jsonsummary.mdenv.jsonrepro.lockstep_logs/support_bundle/stats_verdict_report.json
under artifacts/rgc_statistical_validation_pipeline/<UTC_TIMESTAMP>/.
jq empty docs/rgc_statistical_validation_pipeline_v1.json
rch exec -- env CARGO_TARGET_DIR="$PWD/target_rch_rgc_statistical_validation_pipeline_verify" \
cargo test -p frankenengine-engine \
--test rgc_statistical_validation_pipeline \
--test performance_statistical_validation_integration
./scripts/run_rgc_statistical_validation_pipeline.sh ci
./scripts/e2e/rgc_statistical_validation_pipeline_replay.sh ci