fix: set benchmark baselines from ubuntu-latest CI means

Author: clean6378-max-it

benchmarks/README.md

@@ -40,7 +40,7 @@ The `benchmarks` job on **ubuntu-latest** runs pytest-benchmark, then `scripts/c
 
 ## Refresh baselines
 
-After intentional performance work on ubuntu (same OS as CI):
+After intentional performance work, capture on **ubuntu-latest** (same OS as the gated CI job):
 
 ```bash
 make update-baselines
@@ -53,4 +53,4 @@ pytest tests/benchmarks/ --benchmark-only --benchmark-json=benchmarks/_raw.json
 python scripts/reduce_baselines.py benchmarks/_raw.json benchmarks/baselines.json
 ```
 
-Use `--slack 1.25` on `reduce_baselines.py` when capturing on a faster host than CI to absorb cross-machine variance.
+Do not capture baselines on Windows/macOS for commit — ubuntu runners are slower and the gate will fail. Download `benchmark-results.json` from a green CI artifact to seed baselines if needed.

benchmarks/baselines.json

@@ -1,20 +1,20 @@
 {
-  "_note": "CI gates the ubuntu benchmarks job when mean exceeds baseline by >20%.",
-  "updated": "2026-06-17T20:33:56Z",
-  "machine": "Windows",
+  "_note": "Means captured from ubuntu-latest CI post-cache (PR #90). Refresh via make update-baselines on ubuntu.",
+  "updated": "2026-06-17T21:00:00Z",
+  "machine": "Linux",
   "groups": {
     "parse": {
-      "test_parse_session_small": 8.711556942772768e-05,
-      "test_parse_session_medium": 0.0019041118800302193,
-      "test_parse_session_large": 0.018555900882518687
+      "test_parse_session_small": 0.000105,
+      "test_parse_session_medium": 0.002956,
+      "test_parse_session_large": 0.029678
     },
     "export": {
-      "test_bulk_export_session_count[sessions-10]": 0.0034216649980517108,
-      "test_bulk_export_session_count[sessions-50]": 0.017290590000629893,
-      "test_bulk_export_session_count[sessions-100]": 0.03397804391996553
+      "test_bulk_export_session_count[sessions-10]": 0.004278,
+      "test_bulk_export_session_count[sessions-50]": 0.021144,
+      "test_bulk_export_session_count[sessions-100]": 0.042003
     },
     "search": {
-      "test_search_full_corpus": 0.002600985144447307
+      "test_search_full_corpus": 0.001092
     }
   }
 }