Performance confidence for every merge

Catch performance regressions before they merge.

Functional tests can pass while latency climbs, memory grows, and throughput falls. Benchmark Guardian turns those hidden shifts into a deterministic release decision and an actionable Codex handoff.

Analyze benchmarks

Deterministic analysisRepeatable, explainable results

Protected evaluatorPython remains the source of truth

Codex-ready handoffFrom signal to focused fix

Analysis workspace

Compare benchmark evidence

Inputs stay in memory and are evaluated by the existing Python engine.

Sample data

Example scenario

Standard sample — 2 regressions

A 10% threshold flags both the 25% latency increase and the 15% throughput decrease.

Required input

Baseline

Valid JSON object

Required input

Current

Valid JSON object

Optional input

Directions

Valid JSON object

Need example JSON?

Inspect or save the evidence

Choose a scenario above, download the current inputs, or browse the committed files in examples/scenarios/. PR #20 files are permanent snapshots of real GitHub Actions benchmark evidence.

View JSON on GitHub

Regression threshold

A metric is flagged when its harmful change reaches or exceeds this percentage. Lower thresholds catch smaller regressions; higher thresholds report only larger changes.

At 10%: latency_ms and throughput_rps are flagged. At 25%: only latency_ms is flagged.The dashboard default is 10%; a repository or CI workflow may enforce another policy. This repository's protected PR gate uses 25%.

Fallback direction

Your benchmark decision will appear here

Provide baseline and current metrics, set the regression policy, then run the deterministic analysis.

1. Add JSON2. Configure threshold3. Review readiness

Live verified replay

Try the regression-to-fix loop

Replay the exact benchmark snapshots produced by PR #20. Each run sends the committed evidence through the current /api/analyze endpoint and is evaluated by the same Python engine used for uploads.

The protected GitHub Actions workflow generates baseline.json, current.json, and provenance data. Permanent PR #20 snapshots are committed under examples/scenarios/pr-20/ so they remain available after workflow artifacts expire.

Historical benchmark evidence · evaluated liveView pull request

Regression detected