feat: add replay/parser performance guardrails

.github/workflows/bench.yml

@@ -13,10 +13,6 @@ concurrency:
   group: bench-${{ github.head_ref }}
   cancel-in-progress: true
 
-permissions:
-  contents: read
-  pull-requests: write
-  issues: write
 
 jobs:
   bench:
@@ -34,33 +30,65 @@ jobs:
           path: base
 
       - uses: oven-sh/setup-bun@v2
+        with:
+          bun-version: 1.3.14
+
+      - name: Install PR benchmark dependencies
+        working-directory: pr
+        run: bun install --frozen-lockfile
+
 
-      - name: Run bench on base
+      - name: Detect base replay/parser benchmark
+        id: base-benchmark
+        run: |
+          if [ -f base/test/benchmarks/replay-parser.bench.ts ]; then
+            echo "available=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "available=false" >> "$GITHUB_OUTPUT"
+          fi
+
+      - name: Run benchmarks on base
         working-directory: base
         run: |
-          bun install --frozen-lockfile
+          bun install --no-save --ignore-scripts
           bun run test/benchmarks/pipeline.bench.ts
           mv bench-results.json ../base-results.json
+          if [ -f test/benchmarks/replay-parser.bench.ts ]; then
+            bun test/benchmarks/replay-parser.bench.ts --no-enforce --output=../base-telemetry-results.json
+          fi
 
-      - name: Run bench on PR
+      - name: Run benchmarks on PR
         working-directory: pr
         run: |
-          bun install --frozen-lockfile
           bun run test/benchmarks/pipeline.bench.ts
           mv bench-results.json ../pr-results.json
+          bun test/benchmarks/replay-parser.bench.ts --output=../pr-telemetry-results.json
 
       - name: Compare
         id: compare
         working-directory: pr
         run: |
-          bun scripts/quality/bench-compare.ts ../base-results.json ../pr-results.json --threshold=5 > ../comparison.md
-          cat ../comparison.md
+          bun scripts/quality/bench-compare.ts ../base-results.json ../pr-results.json --threshold=5 > ../pipeline-comparison.md
+          if [ -f ../base-telemetry-results.json ]; then
+            bun scripts/quality/replay-parser-bench-compare.ts ../base-telemetry-results.json ../pr-telemetry-results.json > ../telemetry-comparison.md
+          else
+            {
+              echo "## Replay/parser benchmark budgets"
+              echo
+              echo "Base commit has no replay/parser benchmark. Current hard memory budgets passed; same-machine throughput comparison starts after this benchmark lands on the base branch."
+            } > ../telemetry-comparison.md
+          fi
+          {
+            cat ../pipeline-comparison.md
+            echo
+            cat ../telemetry-comparison.md
+          } | tee ../comparison.md
+          cat ../comparison.md >> "$GITHUB_STEP_SUMMARY"
           {
             echo 'body<<EOF'
             cat ../comparison.md
             echo 'EOF'
           } >> "$GITHUB_OUTPUT"
-
       - name: Find existing comment
         id: find
         uses: peter-evans/find-comment@v3
@@ -75,3 +103,23 @@ jobs:
           comment-id: ${{ steps.find.outputs.comment-id }}
           body: ${{ steps.compare.outputs.body }}
           edit-mode: replace
+
+      - name: Enforce replay/parser budgets
+        if: steps.base-benchmark.outputs.available == 'true'
+        working-directory: pr
+        run: bun scripts/quality/replay-parser-bench-compare.ts ../base-telemetry-results.json ../pr-telemetry-results.json --fail-on-regression
+
+      - name: Upload benchmark reports
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: benchmark-reports
+          path: |
+            base-results.json
+            pr-results.json
+            base-telemetry-results.json
+            pr-telemetry-results.json
+            pipeline-comparison.md
+            telemetry-comparison.md
+            comparison.md
+          retention-days: 7

.gitignore

@@ -32,6 +32,7 @@ laps.json
 client/debug-storybook.log
 client/src/stories/__snapshots__/results/
 bench-results.json
+telemetry-benchmark-results.json
 client/storybook-static/
 .superset/
 

package.json

@@ -31,6 +31,8 @@
     "test:ai": "bun test test/ai/evals/ai-quality.ai-eval.ts --timeout 120000",
     "ai:baseline": "bun scripts/quality/ai-baseline.ts",
     "bench": "bun test/benchmarks/pipeline.bench.ts",
+    "bench:telemetry": "bun scripts/quality/replay-parser-bench-local.ts",
+    "bench:telemetry:current": "bun test/benchmarks/replay-parser.bench.ts",
     "extract:tracks": "bun scripts/games/fm-2023/extract-tracks.ts && bun scripts/games/f1-2025/extract-tracks.ts",
     "extract:tracks:forza": "bun scripts/games/fm-2023/extract-tracks.ts",
     "extract:tracks:f1": "bun scripts/games/f1-2025/extract-tracks.ts",

scripts/quality/README.md

@@ -6,9 +6,11 @@ Capture and compare quality/performance measurements and exercise update install
 |---|---|
 | `bun scripts/quality/ai-baseline.ts` | Run AI evaluation fixtures and write a SHA/model baseline under `test/ai-fixtures/baselines`. Requires Gemini API credentials. |
 | `bun scripts/quality/bench-compare.ts <baseline.json> <current.json> [--threshold=5] [--fail-on-regression]` | Emit benchmark markdown diff; optionally fail when regressions exceed threshold. |
+| `bun scripts/quality/replay-parser-bench-compare.ts <baseline.json> <current.json> [--fail-on-regression]` | Compare compatible same-machine reports; enforce relative throughput/memory and hard memory budgets when requested. |
+| `bun scripts/quality/replay-parser-bench-local.ts [--base=main]` | Benchmark base ref and current worktree on same machine; bootstrap hard memory only when base predates benchmark. |
 | `bun scripts/quality/test-updater.ts` | Build/reuse local installer and start dev server with forced-update variables. |
 
-Inputs: AI credentials and fixtures, Mitata JSON result files, or local package/installer state. Outputs: baseline JSON, markdown on stdout, or update-test process logs/status.
+Inputs: AI credentials and fixtures, same-machine Mitata or replay/parser benchmark JSON result files, or local package/installer state. Outputs: baseline JSON, markdown on stdout, or update-test process logs/status.
 
 Boundary: measurement and local verification workflows only. Scripts do not own production builds, test fixtures, benchmark generation, or release publication.