docs(bench): refresh benchmark results

Author: Arkptz

docs/benchmarks.md

@@ -2,21 +2,21 @@
 
 Generated by the [benchmark workflow](.github/workflows/bench.yml).
 
-Results are refreshed weekly or on-demand via `workflow_dispatch`. Latest numbers
-are committed automatically to this file.
+# Benchmark results
 
-Methodology: `hyperfine --warmup N --runs M` runs each tool's full pipeline
-(discover + generate for Rust; 2x invocation for Python, since the tool requires
-two passes to emit templates then the spec). RSS measured via `/usr/bin/time -f %M`.
+_Run: 2026-04-22 22:31 UTC, commit `22ef2faa`, runner: Linux 6.17.0-1011-azure_
 
-Fixture:
-- **Synthetic** (~80 MB, 40k requests): generated via `httpbin + mitmdump + hey`
-  during the workflow run — 8 endpoint shapes, 5k requests each. Represents a
-  realistic mid-sized capture; small fixtures are dominated by Python's interpreter
-  boot cost and don't reflect real-world processing performance.
+Fixture: 89 MB, 40k requests across 8 endpoint shapes (`bench-fixtures-v1`).
 
-Run the benchmark manually: [Actions → Benchmark → Run workflow](../../actions/workflows/bench.yml).
+## Timing
 
----
+| Command | Mean [s] | Min [s] | Max [s] | Relative |
+|:---|---:|---:|---:|---:|
+| `Python mitmproxy2swagger` | 44.757 ± 0.219 | 44.384 | 44.965 | 16.80 ± 0.26 |
+| `Rust mitm2openapi` | 2.663 ± 0.039 | 2.618 | 2.712 | 1.00 |
 
-_(Seeded file — actual numbers appear after the first scheduled / manual run.)_
+## Peak RSS
+| Tool | RSS |
+|------|----:|
+| Python mitmproxy2swagger | 46104 KB |
+| Rust mitm2openapi        | 6168 KB |