feat(test): add bench mark persist

[ErnestM1234]

Motivation

The monorepo has middleware benchmarks (unit via vitest bench, E2E via Playwright) that only run locally — performance regressions can ship undetected. We need a CI system that automatically compares benchmark results against known baselines and surfaces regressions on PRs, without blocking merges while thresholds are being tuned.

Changes

• .github/workflows/benchmark.yml — New workflow triggered on PRs (with change detection for packages/next/ and the middleware app), repository_dispatch (post-release baseline updates), and workflow_dispatch (manual trigger for external PRs). Uses github-action-benchmark with customSmallerIsBetter format, 150% alert threshold, and informational-only alerts. Baselines are stored on the gh-pages branch and only persisted on release dispatches (never on PRs, for security).
• .github/workflows/release.yml — Added a repository_dispatch step that fires a release-published event with publishedPackages when changesets publishes, so benchmarks only run and baselines only update when gt-next is actually released.
• tests/apps/next/middleware/benchmarks/transform-results.mjs — New script that converts vitest bench JSON and Playwright E2E performance JSON into github-action-benchmark's custom format, tagging each data point with the package name and version.
• tests/apps/next/middleware/benchmarks/e2e-performance.spec.ts — Added BENCH_OUTPUT_NAME env var support for predictable output filenames in CI.
• tests/apps/next/middleware/package.json — Added bench:unit:json convenience script for running vitest bench with JSON output.

Greptile Summary

Implements automated benchmark tracking for middleware performance with CI integration via GitHub Actions. The system runs unit (vitest) and E2E (Playwright) benchmarks on PRs to detect regressions, and updates baselines on releases. Key highlights:

• Security-conscious design: Benchmark data is only persisted to gh-pages on release events, never on PRs (prevents malicious data injection from forks)
• Smart change detection: Benchmarks only run when packages/next/ or the middleware test app changes
• Comprehensive coverage: Tracks factory creation latency, per-request execution, TTFB, DOM load times, redirect chains, and locale switching
• Well-integrated: Release workflow triggers benchmark baseline updates automatically when gt-next is published via repository dispatch
• Future-ready: Includes TODO comments for generalizing the system to support multiple packages

The transform script converts vitest and Playwright outputs into github-action-benchmark format with package/version metadata. Workflow uses 150% alert threshold with informational-only alerts (non-blocking).

Confidence Score: 5/5

• This PR is safe to merge with minimal risk
• Clean implementation with strong security practices (no PR data persistence), well-structured workflow logic with proper change detection, comprehensive documentation, and no critical issues found. The transform script could benefit from error handling, but this is acceptable for CI tooling.
• No files require special attention

Important Files Changed

Filename	Overview
.github/workflows/benchmark.yml	New workflow for running and persisting benchmarks on PRs and releases with proper security controls
.github/workflows/release.yml	Added repository dispatch to trigger benchmark baseline updates after package releases
packages/next/src/middleware-dir/tests/middleware-latency.bench.ts	Moved unit benchmark file from test app to package, updated test names to include 'middleware:' prefix
scripts/transform-bench-results.mjs	New script to transform vitest and Playwright benchmark outputs into github-action-benchmark format
tests/apps/next/middleware/benchmarks/e2e-performance.spec.ts	Added BENCH_OUTPUT_NAME env var support and updated test result keys to include 'middleware:' prefix

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[PR/Release Event] --> B{Event Type?}
    B -->|PR| C[Check Changed Files]
    B -->|Release Dispatch| D[Check if gt-next Published]
    B -->|Manual Trigger| E[Always Run]
    
    C --> F{Should Run?}
    D --> F
    E --> F
    
    F -->|Yes| G[Setup Environment]
    F -->|No| Z[Skip]
    
    G --> H[Build Packages]
    H --> I[Run Unit Benchmarks]
    I --> J[Run E2E Benchmarks]
    J --> K[Transform Results]
    K --> L[Store & Compare with Baseline]
    
    L --> M{Event Type?}
    M -->|PR| N[Compare Only - No Persist]
    M -->|Release| O[Update Baseline on gh-pages]
    
    N --> P[Post Alert if Regression]
    O --> P
    
    style O fill:#f9f,stroke:#333
    style N fill:#bbf,stroke:#333
    style P fill:#bfb,stroke:#333

Loading

_{Last reviewed commit: c9d6ec8}

[ErnestM1234]

@greptileai plz review

[fernando-aviles]

lgtm